Computing, Information Systems & Development

www.cisdijournal.net Vol. 3. No 2,

Computing, Information Systems & Development Informatics Journal Vol. 3. No.2. May, 2012 An International Journal Publication of the Creative Research & Technology Evaluation Network in Conjunction with the African Institute of Development Informatics & Policy, Accra, Ghana and the Cyber Systems Division, International Centre for Information Technology & Development, Baton Rouge, USA.

© All Rights Reserved Published in the USA by: Trans-Atlantic Management Consultants, 2580 Fairham Drive Baton Rouge, LA, USA 70816

May, 2012

Computing, Information Systems & Development Informatics Journal Vol 3. No. 2, May, 2012

CONTENTS i.

Table of Contents

ii.

Editorial Board

iii.

Preface

1-8

On the Classification of Gasoline-fuelled Engine Exhaust Fume Related Faults Using Electronic Nose and Principal Component Analysis Arulogun, O.T., Waheed, M. A., Fakolujo, O.A & Olaniyi, O.M.

9-26

GIS-based Decision Support Systems Applied to Study Climate Change Impacts on Coastal Systems and Associated Ecosystems. Iyalomhe F., Rizzi J., Torresan S., Gallina V., Critto A. & Marcomini A

27-34

Data Mining Technique for Predicting Telecommunications Industry Customer Churn Using both Descriptive and Predictive Algorithms. Kolajo, T & Adeyemo, A.B.

35-42

Lecture Attendance System Using Radio Frequency Identification and Facial Recognition Olaniyi, O.M, Adewumi D.O, Shoewu O. & Sanda O.W

43-50

Employee’s Conformity to Information Security Policies In Nigerian Business Organisations – The Case of Data Engineering Services PLC Adedara, O., Karatu, M.T. & Lagunju, A.

51-56

Policing the Cyber Space – Is the Peel Theory of Community Policing Applicable? Wada, F.J.

57-66

An Intelligent System for Detecting Irregularities in Electronic Banking Transactions Adeyiga,, J.A, Ezike, J.O.J & Adegbola, O.M.

67-74

Management Issues and Facilities Management In Geographic Information System The Case of the Activities of the Lagos Area Metropolitan Transport Authority (LAMATA). Nwambuonwo. J.O & Mughele.E.S

75-81

Strategies for Effective Adoption of Electronic Banking In Nigeria Orunsolu A.A, Bamgboye O., Aina-David O.O & Alaran M.A

82

Call for Papers

83

CISDI Journal Publication Template

ii


Editorial Board Editor-in-Chief

Associate Editors

Prof. Stella S. Chiemeke University of Benin Benin City, Nigeria

Dr. Friday Wada Nelson Mandela Sch. of Public Policy Southern University Baton Rouge, LA, USA

Co-Editor In-Chief

Prof. MAntonio .L. Llorens Gomez

Dr. Richard Boateng University of Ghana Legon, Ghana

Universidad Del Este Carolina, Peurto Rico, USA. Dr. Yetunde Folajimi University of Ibadan, Ibadan, Nigeria

Editorial Advisory Board Prof. C.K. Ayo Covenant University Ota, Nigeria Prof. Adenike Osofisan University of Ibadan Ibadan, Nigeria Prof. Lynette Kvasnny Penn. State University Pennsylvania, USA Prof. Bamidele Oluwade Salem University Lokoja, Nigeria. Dr. Istance Howell DeMontfort University Leceister, United Kingdom

Azeez Nureni Ayofe University of Western Cape Bellville, Cape Town, South Africa Dr. John Effah University of Ghana Business School University of Ghana, Legon Accra Colin Thakur Durban University of Technology South Africa Makoji Robert Stephen Salford Business School Greater Manchester, United Kingdom Dr Akeem Ayofe Akinwale Department of Social Sciences, Landmark University, Omu Aran, Nigeria

Prof. Victor Mbarika The ICT University United State of America

Prof. Maritza .I. Espina Universidad Del Este Carolina, Peurto Rico, USA Managing/Production Editor

Prof. Damien Ejigiri Nelson Mandela Sch. of Public Policy Southern University, USA Dr. Abel Usoro University of the West of Scotland Paisley, Scotland Dr. Onifade O.F.W. Nancy 2 Université France

Dr. Longe Olumide PhD Int. Centre for Information Technology & Development Cyber Systems Division College of Business Southern University and A & M College, Baton Rouge, USA Nigeria Contact Department of Computer Science University of Ibadan, Ibadan, Nigeria

iii


Preface to the CISDI Journal Vol 3, No. 2, May, 2012 This volume of the Computing, Information Systems and Development Informatics Journal (CISDI) provides a distinctive international perspective on theories, issues, frameworks and practice at the nexus of computing, information systems Developments Informatics and policy. A new wave of multidisciplinary research efforts is required to provide pragmatic solution to most of the problems the world faces today. With Computing and Information Technology (IT) providing the needed momentum to drive growth and development in different spheres of human endeavours, there is a need to create platforms through which breakthrough research and research findings that cuts across different discipline can be reported. Such dissemination to a global audience will in turn support future discoveries, sharpen the understanding of theoretical underpinnings and improve practices. This is exactly what the CISDI Journal aims to achieve with timely publications of research, cases and findings from practices in the domain of Computing, Information Technology, Information System/Science and Development Informatics. Articles in this volume cover a broad spectrum of issues that reflects on cases, practices, theories and design. Case studies on using ICTs to support social intervention among asylum seekers was reported from Ireland. eCollaboration for Tertiary Education Using Mobile Systems, eCollaboration for law enforcement agencies as well as factors affecting the use and adoption of Open-Source Software were presented. Other papers covers topics such as social and enterprise informatics, inclusion criteria and instructional technology design, climate change and protocols for improving transactional support in interoperable service oriented application systems. We encourage you to read through this volume and consider submitting articles that reports cutting edge research in computing and short communications/reviews in development informatics research that appropriate design, localization, development, implementation and usage of information and communication technologies (ICTs) to achieve development goals. The CISDI Journal accept articles that promote policy research by employing established (and proposed) legal and social frameworks to support the achievement of development goals through ICTs - particularly the millennium development goals. We will also consider for acceptance, academically robust papers, empirical research, case studies, action research and theoretical discussions which advances learning within the Journal scope and target domains. Extended versions and papers with approved copyright release previously presented at conferences, workshops and seminars will also be accepted for publication. We welcome feedbacks and rejoinders Enjoy your reading Thank you

Longe Olumide Babatope PhD Managing Editor Computing, Information Systems and Development Informatics Journal Fulbright SIR Fellow International Centre for Information Technology & Development Southern University System Southern University Baton Rouge Louisiana, USA E-mail: [email protected] [email protected]

Computing, Information Systems & Development Informatics Journal Vol 3. No. 2, May , 2012

Computing, Information Systems & Development Informatics Journal Volume 3. No. 2. May, 2012

On the Classification of Gasoline-fuelled Engine Exhaust Fume Related Faults Using Electronic Nose and Principal Component Analysis Arulogun, O.T. & Omidiora, E. O. Computer Science and Engineering Department Ladoke Akintola University of Technology P.M.B. 4000 Ogbomoso, Nigeria Waheed, M.A., Mechanical Engineering Department Ladoke Akintola University of Technology P.M.B. 4000 Ogbomoso, Nigeria Fakolujo, O.A., Electrical and Electronic Engineering Department University of Ibadan Ibadan, Nigeria Olaniyi, O.M. Electronic and Electrical Engineering Department, Bells University of Technology Ota, Nigeria

Reference Format: Arulogun, O.T., Waheed, M. A., Fakolujo, O.A & Olaniyi, O.M. (2012). On the Classification of Gasoline-fuelled Engine Exhaust Fume Related Faults Using Electronic Nose and Principal Component Analysis . Computing, Information Systems & Development Informatics Journal. Vol 3, No.2. pp 1-8

1


On the Classification of Gasoline-fuelled Engine Exhaust Fume Related Faults Using Electronic Nose and Principal Component Analysis Arulogun, O.T., Waheed, M. A., Fakolujo, O.A & Olaniyi, O.M.

ABSTRACT The efficiency and effectiveness of every equipment or system is of paramount concern to both the manufacturers and the end users, which necessitates equipment condition monitoring schemes. Intelligent fault diagnosis system using pattern recognition tools can be developed from the result of the condition monitoring. A prototype electronic nose that uses array of broadly tuned Taguchi metal oxide sensors was used to carry out condition monitoring of automobile engine using its exhaust fumes with principal component analysis (PCA) as pattern recognition tool for diagnosing some exhaust related faults. The results showed that the following automobile engine faults; plug-not-firing faults and loss of compression faults were diagnosable from the automobile exhaust fumes very well with average classification accuracy of 91%. Key words: Electronic nose, Condition Monitoring, Automobile, Fault, Diagnosis, PCA.

1. INTRODUCTION The engine is one of the most critical and complex subsystems in the automobile. It is more prone to fault because of its electromechanical nature. It has been shown that early detection of the malfunctions and faults in automobiles as well as their compensation is crucial both for maintenance and mission reliability of vehicles [2]. There are two major approaches that are employed in detecting or predicting faults in any automobile engine, namely: physical observation and electronic condition monitoring approaches. While the first approach uses human senses such as hearing, sight, and smell, the second approach deploys electronic sensors to monitor some conditions such as thermal, vibration, acoustic emission, torque, speed, voltage, current, flow rate, power and so on. The latter approach is more desirable because it avoids human errors when properly implemented. In addition, it predicts with high level of accuracy the real status of the system to which it is deployed when employed with intelligent pattern recognition tools. Condition monitoring consists of methods by which small variations in the performance of equipment can be detected and used to indicate the need for maintenance and the prediction of failure [11]. It can be used to appraise the current state and estimate the future state using real time measurements and calculations. Reference [6] pointed out that a contributing factor in providing ongoing assurance of acceptable plant condition is the use of condition monitoring techniques. Its technologies, such as vibration analysis, infra-red thermal imaging, oil analysis, motor current analysis and ultra-sonic flow detection along with many others have been widely used for detecting imminent equipment failures in various industries [5]. Its techniques have been applied in various fields for the purpose of fault detection and isolation. Reference [17] developed a condition monitoring based diesel engine cooling system model.

The developed model was experimented on a real life diesel engine powered electricity generator to simulate detection of fan fault, thermostat fault and pump fault using temperature measurements. Reference [1] used micro-acoustic viscosity sensors to carry out on-line condition monitoring of lubricating oils in order to monitor the thermal aging of automobile engine oils so as to predict the appropriate time for engine oil change. Electronic noses are technology implementation of systems that are used for the automated detection and classification of odours, vapours and gases [3]. Electronic nose utilizes an instrument, which comprises two main components; an array of electronic chemical sensors with partial specificity and an appropriate patternrecognition system, capable of recognizing simple or complex odours [3]. The main motivation for the implementation of electronic noses is the development of qualitative low cost real-time and portable methods to perform reliable, objective and reproducible measures of volatile compounds and odours [16]. Reference [7] reported the use of electronic nose for the discrimination of odours from trim plastic materials used in automobiles. Reference [9] used electronic nose to quantify the amount of carbon monoxide and methane in humid air. A method for determination of the volatile compounds present in new and used engine lubricant oils was reported by [15]. The identification of the new and used oils was based on the abundance of volatile compounds in headspace above the oils that were detectable by electronic nose. The electronic nose sensor array was able to correlate and differentiate both the new and the used oils by their increased mileages. In [3], electronic nose-based condition monitoring scheme consisting of array of broadly tuned Taguchi metal oxide sensors (MOS) was used to acquire and characterize the exhaust fume smell prints of three gasoline-powered engines operating under induced faults. Reference [8] applied high temperature electronic nose sensors to exhaust gases from modified automotive engine for the purpose of emission control.

2


The array included a tin-oxide-based sensor doped for nitrogen oxide (NOx) sensitivity, a SiC-based hydrocarbon (C xHy ) sensor, and an oxygen (O 2) sensor. The results obtained showed that the electronic nose sensors were adequate to monitor different aspect of the engine's exhaust chemical components qualitatively. In this present study, a prototype of an electronic nosebased condition monitoring scheme using array of ten broadly tuned Taguchi metal oxide sensors (MOS) was used to acquire the exhaust fume of a gasoline-powered engine operating with induced faults. The acquired exhaust fume data were analysed by PCA to diagnose some exhaust related faults.

2. MATERIALS AND METHOD 2.1 The Automobile Engine Automobile engine is a mechanical system where combustion takes place internally. The parts of an engine vary depending on the engine’s type and the manufacturer. Fig. 1 shows some of the basic parts of the internal combustion engine. The system is a heat engine in which combustion occurs in a confined space called a combustion chamber. In a gasoline fuelled engine, a mixture of gasoline and air is sprayed into a cylinder and the mixture is compressed by a piston. The ignition system produces a high-voltage electrical charge and transmits it to the spark plugs via ignition wires. The hot gases that are contained in the cylinder possess higher pressure than the air-fuel mixture so this drives the piston down [13].

Fig. 1: Basic parts of an internal combustion engine [13]

In a perfectly operating engine with ideal combustion conditions, the following chemical reaction would take place in the presence of the following components of basic combustion namely air, fuel and spark: 1. 2. 3.

Hydrocarbons (H xCy ) would react with oxygen to produce water vapour (H2O) and carbon dioxide (CO2) and Nitrogen (N2) would pass through the engine without being affected by the combustion process

In any case of variations in the components of basic combustion or loss of compression due to worn piston rings or high operating temperature the composition of the exhaust gases will change to H2O, CO2, N2, NOX, CO, HxC y and O2. Measurements of exhaust gases such as CO2, CO, NOx, and O 2 can provide information on what is going on inside the combustion chamber and other things going on in the remaining engine units. For example, CO2 is an excellent indicator of efficient combustion: The higher the CO 2 measurement, the higher the efficiency of the engine. High HxC y indicates poor combustion that can be caused by ignition misfire (ignition system failures), insufficient cylinder compression.

3


The gasoline fuelled spark ignition automobile engine considered was a test bed automobile engine. Table 1 gives the specification of the engine, while Fig. 2 shows the snapshot of the test bed engine used in this study.

Samples of the exhaust fumes of the engine operating in normal and various induced faulty conditions were collected for analysis using electronic nose system that consisted of array of ten broadly tuned chemical sensors.

Fig. 2: Snapshot of the Gasoline Fuelled Engine

Table 1: The Engine specification

S/N

Item

Value

1.

Track (rear axle)

50.6 in

2.

Kerb weight

900 Kg.

3.

Engine capacity

1.61 L

4.

Number of valves

8

5.

Number of cylinder

4

6.

Bore/Stroke ratio

1.21

7.

Displacement

96.906 Cu in

8.

Compression ratio

9.5:1

9.

Maximum output

78.3 kW

10.

Maximum rpm coolant

11.

Top gear ratio

Water bhp/litre 0.86

2.2 Chemical sensors The chemical sensor is usually enclosed in an air tight chamber or container with inlet and outlet valves to allow volatile odour in and out of the chamber. The most popular sensors used to develop electronic noses are; semiconductor metal oxide chemo resistive sensors, quartzresonator sensors and conducting polymers. Semiconductor metal oxide chemo resistive sensors types were used in this study because they are quite sensitive to combustible materials such as alcohols but are less efficient at detecting sulphur or nitrogen based odours [4]. The overall sensitivity of these types of sensors is quite good. They are relatively resistant to humidity and to ageing, and are made of particularly strong metals [12].

66.1

Taguchi metal oxide semiconductor (Figaro Sensor, Japan) TGS 813, TGS 822, TGS 816, TGS 2602, TGS 5042, TGS 2104 and TGS 2201 were used based on their broad selectivity to some exhaust gases such as CO2, N2, NOX, CO, uncombusted HxCy, and some other gases such as H2, methane, ethanol, benzene.

4


2.3 Induced Fault Conditions Faults may take time to develop in an automobile engine, hence the need to induce the faults to be investigated. The major faults classes under consideration in this work are plug-not-firing faults and worn piston ring (loss of compression). (a) Plug-not-firing faults: When any of the plugs is malfunctioning, the air-fuel mixture will not be properly ignited but will only be compressed by the piston thereby producing unburnt hydrocarbon with lean quantity of carbon dioxide and more carbon monoxide. Different ignition faults considered were the one-plug-firing, two-plug-firing and the three-plugfiring faults. The faults were inducted into the engines by removing the cables connected to the spark plugs one after the other. (b) Worn piston ring faults: The piston ring prevents engine oil from the oil sump to mix with gasoline-air mixture in the engine combustion chamber and to maintain the engine compression at optimum level. When this ring wears out, the engine oil escapes and mixes with the gasoline-air mixture thereby increases the amount of unburnt hydrocarbon that comes out of the combustion chamber via the exhaust valve. The worn piston ring fault was induced by mixing the gasoline and engine oil in various proportional ratios as 90:10, 80:20, 70:30, 60:40, 50:50 and 40:60. The following calibration was used for the loss of compression faults: a 90:10 fuel mixture will correspond to a 1st degree worn ring and 80:20, 70:30, 60:40, 50:50 and 40:60 will correspond to 2nd, 3rd, 4th, 5th and 6th degree worn ring respectively. The higher the percentage of engine oil that mixes with the gasoline, the higher the degree of wearing of the piston ring which adversely affect the efficiency of the engine. 2.3 Data Acquisition The required exhaust fumes of the gasoline fuelled engine operating in various induced fault conditions were obtained from the engine exhaust tail pipe in the absence of a catalytic converter as specimens into 1000ml Intravenous Injection Bags (IIB). Drip set was used to connect each of the IIB containing the exhaust gases to a confined chamber that contained the array of the selected Taguchi MOS sensors. Static headspace analysis odour handling and sampling method was used to expose the exhaust fume samples to the plastic chamber because the exhaust fume tends to diffuse upwards in clean air due to its lighter weight thus there was no need for elaborate odour handling and sampling method. Readings were taken from the sensors 60 seconds after the introduction of each exhaust fume sample into the air tight plastic chamber so as to achieve odour saturation of the headspace. The digitized data were collected continuously for 10 minutes using Pico ADC 11/10 data acquisition system into the personal computer for storage and further analysis. 1400 x 10 data samples (1 dataset) for each of the ten (10) fault classes making a total of 14000 x 10 data samples (10 datasets) were collected from the test bed engine in the first instance and were designated as training datasets.

The sensors were purged after every measurement so that they can return to their respective default states known as baseline with the use of compressed air. The baseline reading was taken as the unknown fault data. These measurement procedures were repeated five more times to have five samples for each fault class as testing datasets. All data collection were done with the engine speed maintained at 1000 revolutions per second except for 5th degree worn ring, 6th degree worn ring and 3 plugs bad fault conditions that were collected at engine speed of 2000 revolutions per second. 2.4 Data Analysis Principal Components Analysis (PCA) is a technique of linear statistical predictors that has been applied in various fields of sciences especially in process applications [18]. The primary objectives of PCA are data summarization, classification of variables, outlier detection and early indication of abnormality in data structure. PCA has been successfully applied to reduce the dimensionality of a problem by forming a new set of variables. PCA seeks to find a series of new variables in the data with a minimal loss of information [5]. Let X = x1, x2, x3 …………, xm be an m-dimensional observation vector describing the process or machine variables under consideration. A number of observation vectors (obtained or measured at different times) constitute data matrix X. The PCA decomposes the data matrix, X, as T

T

T

m

X = TPT = t1 p1 + t2 p2 ........+ tm pm = ∑ti pi

T

(1)

i =1

Where Pi is an eigenvector of the covariance matrix of X. P is defined as the Principal Components (PC) loading matrix and T is defined to be the matrix of PC scores. The loading provides information as to which variables contribute the most to individual PCs. That is, they are the coefficients in the PC model, whilst information on the clustering of the samples and the identification of transition between different operating conditions is obtained from the score. PCA transforms correlated original variables into a new set of uncorrelated variables using the covariance matrix or correlation matrix [18]. The expectation from conducting PCA is that the correlation among the original variables is large enough that the first few new variables or PCs account for most of the variance [5]. If this holds, no essential insight is lost by applying the first few PCs for further analysis and decision making. If the original variables are collinear, k PCs (k ≤ m) will explain the majority of the variability [5]. In general, k will normally be much smaller than the number of variables in the original data. Consequently, it is desirable to exclude higher-order PCs and retain a small number of PCs.

5


Equation (1) can then be expressed as 3. RESULTS AND DISCUSSION K

X = TP

T

+ E =

∑

ti p iT + E

(2)

I =1

Where E represents residual error matrix [19]. For instance, if the first three PCs represent a large part of the total variance, the residual error matrix will be:

[

E = X − t1 p 1

T

+ t2 p2

T

+ t3 p 3

T

]

(3 )

Typically, in the literature, it is emphasized that the first few PCs contain all the important information [5]. In this study, the singular value decomposition (SVD) technique was used to implement the PCA. In SVD, data matrix X is decomposed into three products by the following equation X = UλP

T

(4)

where U are Eigenvectors, λ are Eigen values and PT is the loading matrix. The main virtue of SVD is that all three matrices are obtained in one operation without having to obtain a covariance matrix as in conventional PCA method [5]. Loading matrices obtained in this method were used to establish the initial PCA models of the system which were based on normal condition and faulty condition data. New observations (measurements) PCA models were projected onto the initial PCA models. Discrepancy or residual between the initial PCA models and new measurement PCA models were detected by calculating the Euclidean distances of the new observations PCA models to initial PCA models.

This study was conducted with various numbers of faulty conditions and normal datasets with each condition having its own developed PCA model. The first ten PCs were used for the purpose of fault classification using euclidean distance metric for to discrimination between PCA models obtained from the training datasets and the testing datasets. Table 2 shows the summary of results of testing the new PCA models against initial PCA models. In Table 2, the number in the squared brackets represents the fault number while the number of times classification was done is shown in bold typeface. Testing of each fault class was done five times. Results of testing of the PCA models with new data samples Compression fault with 1st degree worn ring was classified correctly four out of five times and was incorrectly classified once as compression fault with 3rd degree worn ring. One plug bad fault and compression faults with 4th and 6th degree worn ring were also not classified correctly during the testing. Compression faults with 2nd, 3rd, 5th degree worn ring, oneplug-not-firing fault, two-plugs-not-firing fault, unknown fault and normal condition were correctly classified five out five times. Out of 55 testing samples, 5 were inaccurately classified while 50 were correctly classified. The average classification accuracy of 91% was achieved from the testing. Out of the five inaccurately classified classifications, three were classified as a subset of the same fault class while the other two were truly misclassified to wrong classes as shown in Table 2

The new measurement was classified as any of the existing PCA models with the minimum Euclidean distance or as unknown fault. Eleven initial PCA models were created from the training datasets collected. These PCA models corresponded to the following engine conditions 1st , 2nd, 3rd, 4th, 5thand 6th degree worn ring, one-plug, two-plug, three-plug, and unknown faults, normal conditions. Five different PCA models were developed for each engine condition from the testing datasets.

6


Table 2: Results of testing of the PCA models with new data samples

D ata sample

Cla ssification [5 ] [6] [7 ] [8 ]

[1 ]

[2]

[3]

[4]

[9]

[10]

[11 ]

4

0

1

0

0

0

0

0

0

0

0

0

5

0

0

0

0

0

0

0

0

0

0

0

5

0

0

0

0

0

0

0

0

1

0

0

3

0

0

0

0

0

1

0

0

0

0

0

5

0

0

0

0

0

0

C ompre ssion fa ul t With 1 st degree w orn ring C ompre ssion fa ul t With2nd deg ree w orn ring C ompre ssion fa ul t With 3 rd degree wor n ring C ompre ssion fa ul t With 4 th degree worn ring C ompre ssion fa ul t With 5 th degree worn ring C ompre ssion fa ul t With 6 th degree worn ring One-plug-not-firing f ault

0

0

0

0

0

4

0

0

0

1

0

[6 ] [7]

0

0

0

0

0

0

5

0

0

0

0

Tw o-plug-no t-firi ng fa ult

[ 8]

0

0

0

0

0

0

0

5

0

0

0

Three-pl ug-not-firing fault [9]

0

0

0

0

0

0

0

1

4

0

0

U nkno wn engine fault

[10 ]

0

0

0

0

0

0

0

0

0

5

0

N orma l engine

[1 1]

0

0

0

0

0

0

0

0

0

0

5

[1 ] [2] [3 ] [4 ] [5 ]

4. CONCLUSION An electronic nose-based condition monitoring scheme prototype comprising of ten broadly tuned Taguchi metal oxide sensors (MOS) was used to acquire the exhaust fume of a gasoline-powered engine operating with induced faults. The acquired exhaust fume data were analysed by PCA to diagnose the exhaust related faults. The testing of the PCA algorithm on the exhaust fume data showed a good performance with regards to automobile engine fault diagnosis. The developed system is capable of classifying the plug-not-firing faults and worn piston ring faults from the exhaust fumes very well

[4] Bartlett, P. N., Elliott, J. M. and Gardner, J. W. (1997): Electronic Noses and their Applications in the Food Industry, Food Technology, Vol.51, No 12, pp 44-48. [5] Baydar, N., Ball, A. And Payne, B. (2002): Detection of Incipient Gear Failure Using Statistical Techniques, IMA Journal Of Management Mathematics, Vol. 13, Pp 71-79. [6] Beebe, R. (2003): “Condition monitoring of steam turbines by performance analysis”, Journal of Quality in Maintenance Engineering, Vol. 9, No. 2, pp. 102112. [7] Guadarrama, A., Rodriguez-Mendez, M.L., and De Saja, J.A. (2002): “Conducting polymer-based array for the discrimination of odours from trim plastic materials used in automobiles”, Analytica Chimica Acta., Vol.455, pp. 41–47.

REFERENCES [1] Agoston, A., Otsch, C. and Jakoby, B. (2005): “Viscosity sensors for engine oil condition monitoring: application and interpretation of results”, Sensors and Actuators A, Vol.121, pp.327–332. [2] Alessandri, A. A., Hawkinson, T. Healey, A. J., and Veruggio, G. (1999): Robust Model-Based Fault Diagnosis for Unmanned Underwater Vehicles Using Sliding Mode-Observers, 11th International Symposium on Unmanned Untethered Submersible Technology (UUST'99). [3] Arulogun, O.T., Fakolujo, O.A., Olatunbosun, A., Waheed, M.A., Omidiora, E. O. and Ogunbona, P. O. (2011): “Characterization of Gasoline Engine Exhaust Fumes Using Electronic Nose Based Condition Monitoring”, Global Journal of Researches In Engineering (GJRE-D), Vol. 11, No. 5

[8] Hunter, G. W., Chung-Chiun, L, and Makel, D. B.(2002): Microfabricated Chemical Sensors for Aerospace Applications, The MEMS Handbook, Mohamed Gad-el-Hak, ed., CRC Press, Boca Raton, FL, pp. 22-1--22-24. [9] Huyberechts, G., Szecowka, P., Roggen, J., and Licznerski, B.W. (1997): “Simultaneous Quantification of Carbon Monoxide and Methane in Humid Air Using A Sensor Array and an Artificial Neural Network”, Sensors And Actuators B, Vol. 45, Pp.123-130.

7


[10] Martini, E. B., Morris, A.J. and Zhang, J. (1996): Process Performance Monitoring Using Multivariate Statistical Process Control, I.E.E. Process Control Applications., Vol. 43 (2). [11] Massuyes, T. L. and Milne, R., (1997): “Gas-Turbine Condition Monitoring Using Qualitative Model based Diagnosis”, IEEE EXPERT, Vol. 12, pp 22-31. [12] Meille, P. (1996): ‘Electronic noses’: Towards the objective instrumental characterization of food aroma, Trends in Food Science and Technology, Vol.7, pp. 432-438. [13] NASA (2007): “Internal Combustion Engine”, National Aeronautical and Space Administration website. Retrieved on 10th October, 2008 from http://www.grc.nasa.gov/WWW/K12/airplane/combst1.html

[16] Shilbayeh, N. and Iskandarani, M. (2004): “Quality Control of Coffee Using an Electronic Nose System”, American Journal of Applied Sciences Vol.1, No.2, pp. 129-135. [17] Twiddle, J.A. (1999): “Fuzzy Model Based Fault Diagnosis of a Diesel Engine Cooling System”, Department of Engineering, University of Leicester, Report No. 99-1. Retrieved 7th February 2007 (http://www.le.ac.uk/engineering/mjp9/li1.pdf) [18] Venkatasubramanian, V., Raghunathan, R., Kewen, Y., Surya, N.K. (2003): “A Review of Process Fault Detection and Diagnosis”, Part 1, Computers and Chemical Engineering, Vol. 27, pp. 293-311. [19] Wise, B. M. and Gallagher, N.B. (1996): “The Process Chemometrics Approach to Process and Fault Detection”, Journal of Process Control, Vol. 6, Pp 329-348.

[14] Pöyhönen, S., Jover, P., Hyötyniemi, H. (2004): “Signal Processing of Vibrations for Condition Monitoring of an Induction Motor”, Proc. of the 1st IEEE-EURASIP Int. Symp. on Control, Communications, and Signal Processing, ISCCSP 2004, pp. 499-502, Hammamet, Tunisia. [15] Sepcic, K., Josowicz, M., Janata, J. and Selbyb, T. (2004): “Diagnosis of used engine oil based on gas phase analysis”, Analyst, Vol. 129, pp 1070 – 1075

8



On the Application of GIS-based Decision Support Systems to study climate change impacts on coastal systems and associated ecosystems Iyalomhe F., Rizzi J. & Critto A. University Ca’ Foscari Venice Department of Environmental Sciences, Informatics and Statistics Calle Larga S. Marta 2137, 30123 Venezia (Italy) [email protected]

Torresan S. , Gallina V & Marcomini A. Euro Mediterranean Centre for Climate Change CMCC, Lecce Via Augusto Imperatore, 16 – 73100 Italy

Reference Format:

Iyalomhe F., Rizzi J., Torresan S., Gallina V., Critto A. & Marcomini A. (2012). GIS-based Decision Support Systems applied to study climate change impacts on coastal systems and associated ecosystems. Computing, Information Systems & Development Informatics Journal. Vol 3, No.2. pp 9-26

9


On the Application of GIS-based Decision Support Systems to study climate change impacts on coastal systems and associated ecosystems Iyalomhe F., Rizzi J., Torresan S., Gallina V., Critto A. & Marcomini A

ABSTRACT One of the most remarkable achievements by scientists in the field of global change in recent years is the improved understanding of climate change issues. Its effects on human environments, particularly coastal zones and associated water systems, are now a huge challenge to environmental resource managers and decision makers. International and regional regulatory frameworks have been established to guide the implementation of interdisciplinary methodologies, useful to analyse water-related systems issues and support the definition of management strategies against the effects of climate change. As a response to these concerns, several decision support systems (DSS) have been developed and applied to address climate change through geographical information systems (GIS) and multi-criteria decision analysis (MCDA) techniques; linking the DSS objectives with specific functionalities leading to key outcomes, and aspects of the decision making process involving coastal and waters resources. An analysis of existing DSS focusing on climate change impacts on coastal and related ecosystems was conducted by surveying the open literature. Consequently, twenty DSS were identified and are comparatively discussed according to their specific objectives and functionalities, including a set of criteria (general technical, specific technical and applicability) in order to better inform potential users and concerned stakeholders through the evaluation of a DSS’ actual application. Key words: Climate change, Decision support, GIS, regulations, Environment 1. INTRODUCTION One of the most remarkable achievements by scientists in the field of global change in recent years is the improved understanding of climate change issues, whose effects have been linked to the increase in global average temperature according to the IPCC emission scenarios [11]. Resulting ocean thermal expansion is expected to generate significant impacts via sea level rise, seawater intrusion into coastal aquifers, enhanced coastal erosion and storm surge flooding, while increasing population in coastal cities, especially megacities on islands and deltas, further aggravates major impacts of climate change on marine coastal regions. The latter include transitional environments such as estuaries, lagoons, low lying lands and lakes, which are particularly vulnerable because of their geographical location and intensive socio-economic activities [12,13]. Accordingly, several environmental resource regulations have already included the need to assess and manage negative impacts derived from climate change through their implementation. For instance, the European Commission approved the Green and White papers [1415], the Water Framework Directive (WFD) [16], which represent an integrated and sound approach for the protection and management of water-related resources in both inland and coastal zones. They also signed the protocol for Integrated Coastal Zone Management (ICZM) [17], useful in the promotion of the integrated management of coastal areas in relation to local, regional, national and international goals. Moreover, the principles of Integrated Water Resources Management (IWRM) aimed to address typical water quality and quantity concerns with the optimisation of water management and sustainability in collaboration with WFD policy declarations [18].

Likewise, relevant national legislations like Shoreline Management Planning (SMP) in the United Kingdom [19], Hazard Emergency Management (HEM) in the United States [20] and Groundwater Resources Management (GRM) in Bangladesh and India [21] were ratified and further endorse the assessment and management of coastal communities in relation to climate change impacts. Decision Support System (DSSs) is computer-based software that can assist decision makers in their decision process, supporting rather than replacing their judgment and, at length, improving effectiveness over efficiency [1]. Environmental DSS are models based tools that cope with environmental issues and support decision makers in the sustainable management of natural resources and in the definition of possible adaptation and mitigation measures [2]. DSS have been developed and used to address complex decision-based problems in varying fields of research. For instance, in environmental resource management, DSS are generally classified into two main categories: Spatial Decision Support Systems (SDSS) and Environmental Decision Supports Systems (EDSS) [3-5]. SDSS provide the necessary platform for decision makers to analyse geographical information in a flexible manner, while EDSS integrate the relevant environmental models, database and assessment tools – coupled within a Graphic User Interface (GUI) – for functionality within a Geographical Information System (GIS) [1-4-6]. In some detail, GIS is a set of computer tools that can capture, manipulate, process and display spatial or geo-referenced data [7] in which the enhancement of spatial data integration, analysis and visualization can be conducted [89]. These functionalities make GIS-tools useful for efficient development and effective implementation of DSS within the management process. For this purpose they are used either as data managers (i.e. as a spatial geodatabase tool) or as an end in itself (i.e. media to communicate information to decision makers) [8].

10


At present the increasing trends of industrialisation, urbanisation and population growth has not only resulted in numerous environmental problems but has increased the complexity in terms of uncertainty and multiplicity of scales. Accordingly, there is a consensus on the consideration of several perspectives in order to tackle environmental problems, particularly, climate change related impacts in coastal zones which are characterised by the dynamics and interactions of socio-economic and biogeophysical phenomena. There is the need to develop and apply relevant tools and techniques capable of processing not only the numerical aspects of these problems but also knowledge from experts, to assure stakeholder participation which is essential in the decision making process [5] and to guarantee the overall effectiveness of assessment and management of coastal environments – including related inland watersheds (i.e. surface and groundwater affected by, and affecting, coastal waters). The scientific community projected that climate change would further exacerbate environmental problems due to natural and anthropogenic impacts – with specific emphasis in coastal areas [10]. This data, nevertheless, depends on global and regional policy measures especially in sectors such as energy, economy and agriculture which seem to be a major threat to global sustainable development. As a response to this, mitigation and adaptation measures are already identified through intense research activities, yet these may not limit the projected effects of climate change over the next few decades On one side there is the influence of socio-economic development and environmental response while on the other there is the significant uncertainty still associated with present climatic predictive models. Thus, model inputs need to take into account scenarios highly affected by present and future policy measures in order to further reduce uncertainty in their predictions and thereby guarantee robust adaptation strategies. In addition, climate change effects have been linked to the increase in global average temperature according to the IPCC emission scenarios [11]. Resulting ocean thermal expansion is expected to generate significant impacts via sea level rise, seawater intrusion into coastal aquifers, enhanced coastal erosion and storm surge flooding, while increasing population in coastal cities, especially megacities on islands and deltas, further aggravates major impacts of climate change on marine coastal regions. The latter include transitional environments such as estuaries, lagoons, low lying lands, lakes, which are particularly vulnerable because of their geographical location and intensive socio-economic activities [12-13].

Within this context, the development of innovative tools is needed to implement regulatory frameworks and the decision making process required to cope with climate related impacts and risks. To this end, DSS are advocated as one of the principal tools for the described purposes. This work will attempt to examine GIS-based DSS resulting from an open literature survey. It will highlight major features and applicability of each DSS in order to help the reader in the selection of DSS tailored on his specific application needs. 2. DESCRIPTION OF THE EXAMINED DECISION SUPPORT SYSTEMS (DSS) The literature survey led to identify twenty DSS designed to support the decision making-process related to climate change and environmental issues in coastal environments – including inland watersheds. The identified DSS are listed in Table 1 with the indication of the developer, development years, and literature reference. In order to provide a description of major features and an evaluation of the applicability of the 20 examined DSS, the work adopted the sets of criteria reported in Table 2 and grouped them within three different categories: general technical criteria, specific technical criteria, and availability and applicability criteria. The general technical criteria underline relevant general features related to each DSS, which include: the target coastal regions and ecosystems domain; the regulatory frameworks and specific legislations supported by each DSS; the considered climate change impacts and related scenarios, as well as the objectives of the examined systems. The specific technical aspects include the main functionalities, analytical methodologies and inference engine (i.e. structural elements) of the systems. A final set of criteria concerned applicability, i.e. scale and study areas, flexibility, status and availability of the examined systems. Within the following sections the identified DSS, listed in Table 1, will be presented discussed according to these criteria

11


Table 1. List of existing DSS on coastal waters and related inland watersheds. Name

Developer

CLIME: Climate and Lake Impacts decision support system CORAL: Coastal Management Decision Support Modelling for Coral Reef Ecosystem COSMO: Coastal zone Simulation MOdel Coastal Simulator decision support system. CVAT: Community Vulnerability Assessment Tool DESYCO: Decision Support SYstem for COastal climate change impact assessment DITTY: Information technology tool for the management of Southern European lagoons DIVA: Dynamic Interactive Vulnerability Assessment ELBE: Elbe river basin Decision Support System GVT:Groundwater Vulnerability Tool

Helsinki University of Technology, Finland

IWRM: Integrated Water Resources Management Decision Support System KRIM decision support system MODSIM decision support systems RegIS-Regional Impact Simulator RAMCO: Rapid Assessment Module Coastal Zone Management SimLUCIA: Simulator model for St LUCIA

Year of Development 1998-2003

[22] http://clime.tkk.fi

Within a World Bank funded Project :LA3EU

1994-1995

[23]

Coastal Zone Management Centre, Hague Tyndall Centre for Climate Change Research, UK. National Oceanic and Atmospheric Administration, US. Euro-Mediterranean Centre for Climate Change, (CMCC) Italy.

1992

[24]

2000-2009

[25]

1999

[20] www.csc.noaa.gov/products/nchaz/startup.htm

2005-2010

[2]

Within the European region project: DITTY

2002- 2005

[26]

Potsdam Institute for Climate Impact Research, Germany

2003-2004

[27] http://www.dinas-coast.net.

Research Institute of Knowledge SystemRIKS, Netherland University of Thrace and Water Resource Management Authority, Greece. Institute of Water Modelling, Bangladesh

2000-2006

[28] www.riks.nl/projects/Elbe-DSS

2003-2004

[29]

2002-2010

[21] www.iwmbd.org

Within the KRIM Project in Germany. Labadie of Colorado State University, US Cranfield University, UK Research Institute of Knowledge SystemRIKS, Netherland

2001-2004

[30] www.krim.uni-bremen.de

1970

[31-32] www.modsim.engr.colostate.edu

2003-2010 1996-1999

[33]http://www.cranfield.ac.uk/sas/naturalresources /research/projects/regis2.html [34-35] http://www.riks.nl/projects/RAMCO

Research Institute of Knowledge SystemRIKS within the UNEP Project, Netherland

1988-1996

[36] http://www.riks.nl/projects/SimLUCIA

12

Reference Source


SimCLIM: Simulator model System for Climate Change Impacts and Adaptation STREAM: Spatial Tools for River Basins and Environment and Analysis of Management Options TaiWAP: Taiwan Water Resources Assessment Program to Climate Change WADBOS: decision support systems

University of Waikato and CLIMsystem limited, New Zealand.

2005

[37] www.climsystems.com

Vrije Universiteit Amsterdam and Coastal Zone Management Centre, Hague

1999

[38] http://www.geo.vu.nl/users/ivmstream/

National Taiwan University, Taiwan

2008

[39]

Research Institute of Knowledge SystemRIKS, Netherland

1996-2002

[40-41] www.riks.nl/projects/WADBOS

Table 2. List of criteria used for the description of existing DSS. Categories General technical criteria

Specific technical criteria

Availability and applicability

Criteria • Coping with regulatory framework. This indicates the particular legislation or policy, the DSS refers to and which phase of the decision-making process is supported at the National, Regional and Local level (e.g., EU WFD, ICZM, IWRM, SMP, GRM, and HEM). • Study/ field of application area. The coastal zones where this DSS has been applied and tested (e.g., coastal zone, lakes, river basin, lagoon, groundwater aquifer etc.) • Objective. It specifies the main aims of the DSS. • Climate change impacts. This refers to relevant impacts due to climate change on the system (e.g., sea-level rise, coastal flooding, erosion, water quality). • Climate Change Scenarios. The kind of scenarios considered by the DSS, which are relevant to the system analysis and connected to climate change (e.g., emission, sea level rise, climatic scenarios). • Functionalities. These indicate relevant functionalities (key outcomes) of the system useful to the decision process: environmental status evaluation, scenarios import (climate change and socioeconomic scenarios) and analysis, measure identification and/or evaluation, relevant pressure identification and indicators production. • Methodological tools/ (analytical tools). These indicate the methodologies included in the system such as risks analysis, scenarios construction and/or analysis, integrated vulnerability analysis, MultiCriteria Decision Analysis (MCDA), socio-economic analysis, uncertainty analysis, ecosystem-based approach etc. • Structural elements. The three major components of the DSS: dataset (i.e., the typology of data), models (e.g., economic, ecological, hydrological and morphological), interface (i.e., addressing if it’s user-friendly and desktop or web-based). • Scale and area of application. This specifies the spatiality of the system (e.g., local, regional, national, supra-national and global) within the case study areas. • Flexibility. The characteristics of the system to be flexible, in terms of change of input parameters, additional modules or models and functionalities. It is also linked to the fact that it can be apply on different coastal regions or case study areas. • Status and Availability. This specifies if the system is under development or already developed and ready for use, and if it is restricted to the developer and case study areas only or the public can access it too and the website where information about the DSS can be found.

13


3 GENERAL TECHNICAL CRITERIA As far the application domain, the considered DSS focus on coastal zones and related ecosystems (e.g. lagoons, groundwater, river basins, estuaries, and lakes), specifically thirteen DSS are on coastal zones, seven concern coastal associated ecosystems and four focuses on both (Table 3). As far as regulatory frameworks (i.e. ICZM, WFD, and IWRM) and national legislations are concerned, the examined DSS reflect the assessment and management aspects of the related decision making process. Within the coastal, marine and river basin environments, the assessment phase of these frameworks consists of the analysis of environmental, social, economic and regulatory conditions, while the management phase looks at the definition and implementation of management plans.

The main objective of the examined DSS is the analysis of vulnerability, impacts and risks, and the identification and evaluation of related management options, in order to guarantee robust decisions required for sustainable management of coastal and inland water resources. Specifically, the objectives of the examined DSS are concerned with three major issues: (1) the assessment of vulnerability to natural hazards and climate change (four DSS: CVAT, GVT, SimLUCIA, TaiWAP); (2) the evaluation of present and potential climate change impacts and risks on coastal zones and linked ecosystems, in order to predict how coastal regions will respond to climate change (nine DSS); (3) the evaluation or analysis of management options for the optimal utilisation of coastal resources and ecosystems through the identification of feasible measures and adequate coordination of all relevant users/stakeholders (seven DSS: WADBOS, COSMO CORAL, DITTY, ELBE, MODSIM, RAMCO).

Accordingly, support is provided by each DSS to the implementation of one or two frameworks in the assessment and/or management phase in relation to specific objectives and application domain. Specifically, the investigated DSS can provide the evaluation of ecosystem pressures, the assessment of climate change hazard, vulnerability and risks, the development and analysis of relevant policies, and the definition and evaluation of different management options. Eight out of the twenty examined DSS provide support for the ICZM implementation through an integrated assessment involving regional climatic, ecological and socio-economic aspects (Table 3, second column). With respect to the WFD (i.e. six DSS) and IWRM (i.e. seven DSS), the main focus is on the assessment of environmental or ecological status of coastal regions and related ecosystems and on the consideration of anthropogenic impacts and risks on coastal resources. These two groups of DSS consider also the river basins management via evaluation of adaptation options, which is essential for the management phase of the WFD and IWRM implementation. Particularly interesting are the approaches adopted by three DSS: CLIME, STREAM and COSMO. CLIME supports both the assessment and management phases of WFD through the analysis of present and future climate change impacts on ecosystems and the socio-economic influence on water quality of the European lakes. STREAM evaluates climate change and land use effects on the hydrology of a specific river basin, in order to support the management phase of IWRM and WFD via the identification of water resources management measures. Lastly, COSMO provides support for the ICZM through the identification and evaluation of feasible management strategies for climate change and anthropogenic impacts relevant for coastal areas. Moreover, RegIS, Coastal Simulator, CVAT and GVT specifically support the implementation of national legislations through the consideration of socio-economic and technological issues relevant for identifying suitable mitigation actions. To this purpose, these DSS promote the involvement of stakeholders through participatory processes.

14


Table 3. List of the examined DSSs according to the general technical criteria (ND: Not Defined). Name Application Regulatory Objective Climate change Climate change domain Framework of impacts scenarios reference addressed generating impacts WFD for To explore the potential CLIME • Emission • Lakes. • Water quality. environmental impacts of climate change scenarios. assessment. on European lakes • Temperature dynamics linked coast. scenarios. IWRM and ICZM Sustainable management CORAL • Coral reef • ND • ND both for of coastal ecosystems in environmental particular, coral reef. assessment and management. ICZM for To evaluate coastal COSMO • Coastal • Sea-level rise. • Sea-level rise environmental management options zones. scenarios. management. considering anthropic (human) forcing and climate change impacts. National Effects of climate change • Storm surge Coastal • Coastal • Emission legislation for /management decisions on Simulator zones. flooding. scenarios. environmental the future dynamics of the • Coastal • Sea-level rise assessment and coast. erosion. scenarios. management. National To assess hazards, CVAT • Coastal • Storm surge • Past observations legislation for vulnerability and risks zones. flooding. environmental related to climate change • Coastal assessment and and support hazard erosion. management. mitigation options. • Cyclone. • Typhoon. • Extreme events ICZM for To assess risks and DESYCO • Coastal • Sea-level rise. • Emission environmental impacts related to climate zones. scenarios. • Relative seaassessment and change and support the • Coastal level rise • Sea level rise management. definition of adaptation • Storm surge Lagoons scenarios. measures. flooding. • Coastal erosion. • Water quality IWRM and WFD To achieve sustainable and • ND DITTY • ND for environmental rational utilization of • Coastal management. resources in the southern Lagoons. European lagoons by taking into account major anthropogenic impacts. ICZM for To explore the effects of DIVA • Coastal • Sea-level rise. • Emission environmental climate change impacts on • Coastal zones. scenarios. assessment and coastal regions. erosion. • Sea-level rise management. • Storm surge scenarios. flooding. WFD for To improve the general ELBE • River • Precipitation • Emission environmental status of the river basin basin. and scenarios. usage and provide • Catchment. management. temperature sustainable protection variation. measure within coast. National To describe the GVT • Coastal • Groundwater • Sea-level rise legislation for vulnerability of zones. quality. scenarios. environmental groundwater resources to • Saltwater assessment. pollution in a particular intrusion. coastal region.

15


IWRM

• Coastal zones. • River basin

IWRM for environmental assessment and management.

To explore potential risks on coastal resources due to climate and water management policies.

• Sea-level rise. • Coastal erosion.

KRIM

• Coastal zones.

ICZM for environmental assessment.

To determine how coastal systems reacts to climate change in order to develop modern coastal management strategies.

MODSIM

• River basin.

IWRM for environmental management.

RegIS

• Coastal zones.

SMP and Habitats regulation (UK) for environmental assessment and management.

To improve coordination and management of water resources in a typical river basin. To evaluate the impacts of climate change, and adaptation options.

• Sea-level rise. • Extreme events. • Coastal erosion. • ND

RAMCO

• River basin. • Coastal zones. • Coastal zones.

WFD and ICZM for environmental assessment and management.

For effective and sustainable management of coastal resources at the regional and local scales.

• ND

National legislation for environmental assessment.

To assess the vulnerability of low lying areas in the coastal zones and island to sea-level rise due to climate change.

• Sea-level rise scenarios.

SimCLIM

• Coastal zones.

ICZM for environmental assessment and management.

To explore present and potential risks related to climate change and natural hazards (e.g. erosion, flood).

STREAM

• River basin. • Estuaries.

IWRM and WFD for environmental management.

TaiWAP

• River basin.

IWRM for environmental assessment.

• Water quality variations.

• Emission scenarios.

WADBOS

• River basin. • Coastal zones.

WFD and ICZM for environmental assessment and management.

To integrate the impacts of climate change and landuse on water resources management. To assess vulnerability of water supply systems to impacts of climate change and water demand. To support the design and analysis of policy measures in order to achieve an integrated and sustainable management.

• Sea-level rise. • Coastal erosion. • Storm surge flooding. • Sea-level rise. • Coastal flooding. • Coastal erosion. • Water quality variation. • Salt intrusion.

• ND

• ND

SimLUCIA

According to the climate change impacts considered by the examined DSS, the review highlights that fifteen out of the 20 DSS applications regard the assessment of climate change impacts and related risks (CC-DSS). These DSS consider climate change impacts relative to sea level rise, coastal erosion, and storm surge flooding and water quality. In particular, DESYCO also consider relative sea level rise in coastal regions where there are records of land subsidence, whereas KRIM and CVAT assess impacts related to extreme events and natural hazards (e.g. typhoon, cyclone, etc.) respectively. Moreover, GVT is specifically devoted to groundwater quality variations.

• Coastal and river flooding. • Sea level rise

• Sea-level rise scenarios. • Emission scenarios. • Sea-level rise scenarios. • Extreme events scenarios. • ND

• Emission scenarios • Socio-economic scenarios • Sea level rise scenarios • ND

• Sea-level rise scenarios.

• Emission scenarios.

The relevant climate change related scenarios considered by the examined DSS refer to emission of greenhouse gases, temperature increase, sea level rise and occurrence of extreme events. In addition, CVAT used previous observations as baseline scenarios for the assessment of natural hazards; while RegIS considered scenarios related to coastal and river flooding along with socio-economic scenarios in order to estimate their potential feedback on climate change impacts. Although most of these CC-DSS applications used sea level rise scenarios, only DIVA used global sea level rise scenarios to estimate related impacts like coastal erosion and storm surge flooding. KRIM is the only DSS considering extreme events scenarios in its analysis to support the development of robust coastal management strategies.

16


4. SPECIFIC TECHNICAL CRITERIA The criteria related to the specific technical aspects are reported in Table 4. As far as the functionalities are concerned (Table 4, first column), the ones implemented by DESYCO, COSMO, SimCLIM, KRIM and RegIS include the identification and prioritisation of impacts, targets and areas at risk from climate change, sectorial evaluation of impacts or integrated assessment approach, and vulnerability evaluation and problem characterisation.

These are to effectively differentiate and quantify impacts and risks at the regional scale. Moreover, they also support the definition and evaluation of management options through GIS-based spatial analysis. Other DSS, i.e. DIVA, SimCLIM and KRIM, implement scenarios import and generation, environmental status evaluation, impacts and vulnerability analysis and evaluation of adaptation strategies to adequately achieve a sustainable state of coastal resources and ecosystems.

Table 4. List of the examined DSSs according to the specific technical criteria. Name CLIME

CORAL

COSMO

Coastal Simulator

CVAT

Functionalities • Identification of pressure generated by climatic variables. • Environmental status evaluation. • Water quality evaluation related to climate change. • Socio-economic evaluation. • Spatial analysis (GIS). • Evaluation of management strategies • Spatial analysis (GIS).

Analytical methodologies • Scenarios construction and analysis. • Probabilistic Bayesian network. • Uncertainty analysis.

Structural elements • Climatic, hydrological, chemical, geomorphological data. • Climate, ecological and hydrological models. • Web-based user interface

• Scenarios construction and analysis. • Cost-effectiveness analysis. • Ecosystem-based.

• Problem characterization (e.g. water quality variation, coastal erosion etc.) • Impact evaluation of different development and protection plans. • Indicator production. • Spatial analysis (GIS). • Environmental status evaluation. • Management strategies identification and evaluation. • Indicator production. • Spatial analysis (GIS).

• Scenarios construction and analysis. • MCDA. • Ecosystem-based

• Environmental, socioeconomic, ecological, biological data. • Economic and ecological models. • Desktop user interface. • Socio-economic, climatic, environmental, hydrological data. • Ecological, economic and hydrological models. • Desktop user friendly interface

• Environmental status evaluation. • Hazard identification. • Indicators production. • Mitigation options identification and evaluation. • Spatial analysis (GIS).

• • • • • •

• Scenarios construction and analysis. • Uncertainty analysis. • Risk analysis. • Ecosystem-based.

Hazard analysis. Critical facilities analysis. Society analysis. Economic analysis. Environmental analysis. Mitigation options analysis.

17

• Climatic, socio-economic, environmental, hydrological, geomorphological data. • Ecological, morphological climatic and hydrological models. • Desktop user interface. • Environmental and socioeconomic data. • Hydrological model. • Desktop user friendly interface


DESYCO

DITTY

DIVA

ELBE

• Prioritization of impacts, targets and areas at risk from climate change. • Impacts, vulnerability and risks identification. • Indicators production. • Adaptation options definition • Spatial analysis (GIS). • Management options evaluation • Indicator production. • Spatial analysis (GIS).

• Scenarios generation and analysis. • Environmental status evaluation. • Indicators production. • Adaptation options evaluation. • Spatial analysis (GIS). • Environmental status evaluation. • Protection measures identification. • End-user involvement. • Spatial analysis (GIS).

• Regional Risk Assessment methodology. • Scenarios construction and analysis. • MCDA. • Risk analysis.

• Climatic, biophysical, socio-economic, geomorphological, hydrological data. • Desktop automated user interface.

• Scenarios construction and analysis. • Uncertainty analysis. • MCDA. • Social cost and benefits analysis. • DPSIR. • Scenarios construction and analysis. • Cost-benefit analysis. • Ecosystem-based.

• Morphological, social, hydrological, ecological data. • Hydrodynamics, biogeochemical, socioeconomic models. • Desktop user interface.

• Scenarios construction and analysis.

• Hydrological, ecological, socio-economic, morphological data. • Economic, • Hydrological, models. • Desktop complex user interface. • Data (environmental, climatic, hydrological, socioeconomic). Hydrological, socioeconomic and DEM models. • Desktop user interface.

GVT

• Environmental status evaluation. • Indicators production • Spatial analysis (GIS). • Impact and vulnerability evaluation

• Risks analysis. • Fuzzy logic. • MCDA.

IWRM

• Environmental status evaluation. • Indicators production. • Adaptation measures evaluation. • Information for non-technical users. • Spatial analysis (GIS). • Environmental status evaluation. • Adaptation measures evaluation. • Information for non-technical users. • Spatial analysis (GIS).

• Scenarios construction and analysis. • Risk analysis. • Cost-benefit analysis. • Socio-economic analysis.

KRIM

• Scenarios construction and analysis. • Impact and risk analysis. • Ecosystem-based.

18

• Climatic, socio-economic, geography, morphological data. • Economic, ecological, geomorphological, climate models. • Desktop graphical user interface.

• Climatic, environmental, socio-economic, geomorphological data. • Hydrodynamic, climate, economic models. • Desktop user interface. • Climatic, socio-economic, ecological, environmental, hydrological data. • Economic, ecological, hydrodynamic, geomorphological models. • Desktop user interface.


MODSIM

• Environmental status evaluation. • Management measures evaluation. • Spatial analysis (GIS).

• Statistical analysis. • Analysis of policies.

RegIS

• Indicators production • Management measures evaluation. • Information for non-technical users. • sectoral evaluation • Spatial analysis (GIS). • Environmental status evaluation. • Indicators generation. • Management measures evaluation. • Spatial analysis (GIS).

• Scenarios construction and analysis. • Impact analysis. • DPSIR. • Integrated assessment.

SimLUCIA

• Indicators production. • Impact and vulnerability evaluation. • Management and land-use measures evaluation. • Spatial analysis (GIS).

SimCLIM

• Environmental status evaluation. • Impact and vulnerability evaluation. • Adaptation strategies evaluation • Spatial analysis (GIS). • Environmental status evaluation. • Indicators production. • Management measures evaluation spatial analysis (GIS). • Environmental status evaluation.• Indicators production. • Spatial analysis (GIS).

• Cellular Automata. • Scenarios construction and analysis. • Socio-economic analysis. • Bayesian probabilistic networks. • Ecosystem-based. • Scenario construction and analysis. • Statistical analysis. • Risk analysis. • Cost/benefit analysis. • Ecosystem-based.

RAMCO

STREAM

TaiWAP

WADBOS

• Management measures identification and evaluation. • Spatial analysis (GIS).

• Scenarios construction and analysis. • Cellular automata. • Ecosystem-based.

• Scenarios construction and analysis.

• Scenarios construction and analysis. • Impact and vulnerability analysis.

• Scenarios construction and analysis. • Sensitivity analysis. • MCDA.

19

• Administrative, hydrological, socioeconomic, environmental data. • Socio-economic, hydrological models. • Web-based user interface. • Climatic, socio-economic, geomorphological, hydrological data. • Climate and flood metalmodels. • Desktop user interface. • Socio-economic, environmental, climatic data. • Biophysical, socioeconomic and environmental models. • Web-based user interface. • Climatic, environmental, socio-economic data. • Land use, social and economic, climate models. • Web-based user interface.

• Climatic, hydrological, socio-economic data. • Climate, hydrological, economic models. • Desktop user interface.

• Climatic, socio-economic, ecological, hydrological data. • Climate, hydrological models. • Web-based user interface. • Climatic, socio-economic, hydrological data. • Climate, hydrological, water system dynamic models. • Desktop user interface. • Socio-economic, hydrological, environmental, ecological data. • Socio-economic, ecological, landscape models. • Desktop user interface.


In order to effectively support the assessment and management of groundwater resources, GVT and DESYCO estimate indicators in assessing impacts, vulnerability and risks to estimate groundwater quality and coastal environmental quality, respectively. Similarly, STREAM, ELBE, RAMCO and DITTY employ environmental status evaluation, protection measures identification, and spatial analysis to support the management aspects of coastal ecosystems. Moreover, CLIME and CORAL specifically support the assessment and management of lakes and coral reefs via the adoption of management strategies and the evaluation and identification of pressures from climatic variables. In particular, five out of the 20 examined DSS (i.e. CVAT, GVT, Coastal Simulator, SimLUCIA and RegIS) consider hazards identification, impacts and vulnerability evaluation, mitigation/ management options identification and evaluation and sectoral evaluation to achieve a comprehensive and integrated analysis of coastal issues at the local or regional scale. Among all considered DSS, RegIS is the one most oriented to stakeholders. The second column of table 4 shows the methodologies adopted by each DSS. Seventeen out of 20 examined DSS consider scenarios analysis to enable coastal managers, decision makers and stakeholders to anticipate and visualise coastal problems in the foreseeable future, and to better understand which future scenario is most suitable for consideration in the evaluation process. A useful methodology is represented by the Multi-Criteria Decision Analysis (MCDA) technique that is considered by five DSS (i.e. COSMO, DESYCO, DITTY, GVT and WADBOS) in order to compare, select and rank multiple alternatives that involve several attributes based on several different criteria. Moreover, DITTY and RegIS also consider the DPSIR approach as a causal framework to describe the interactions between the coastal system, society and ecosystems to carry out an integrated assessment with the aim to protect the coastal environment, guarantee its sustainable use, and conserve its biodiversity in accordance to the Convention on Biodiversity (2003). An ecosystemic assessment was developed nine DSS (i.e. CORAL, COSMO, Coastal simulator, DIVA, RegIS, KRIM, RAMCO, SimLUCIA, SimCLIM) to support the analysis of the studied region through the representation of relevant processes and their feedbacks. Furthermore KRIM, IWRM, COSMO, SimCLIM and Coastal Simulator employ the risk analysis approach for impacts and vulnerability evaluation and also for general environmental status evaluation. A more detailed approach to risk analysis, through the regional risk assessment methodology (RRA), was adopted by DESYCO, Coastal Simulator and RegIS with huge emphasis on the local or regional scales. Finally, CLIME and SimLUCIA consider the Bayesian probability network to highlight the causal relationship between ecosystems (e.g. lakes) and climate change effects.

With regard to the structure of examined DSS (Table 4, third column), most of them employ analytical models useful to highlight the basic features and natural processes of the examined territory, such as the landscape and ecological models used by the WADBOS, the environmental model employed by RAMCO, the geomorphological model used within KRIM and the flood meta-model which interface other models considered by the RegIS. Moreover, the majority of these DSS utilise numerical models necessary to simulate relevant circulation and geomorphological processes that may influence climate change and related risks. DSS like CLIME, DESYCO, CVAT and TaiWAP adopt models useful to represent specific climatic processes (e.g. hydrological cycle and fate of sediment). More importantly, ten (i.e. WADBOS, SimLUCIA, RAMCO, MODSIM, GVT, ELBE, DIVA, CORAL, DITTY AND SimCLIM) out of the twenty examined DSS consider relevant socioeconomic models outputs in their analysis to critically support the integrated assessment of coastal zones. Finally, the majority of these DSS consider integrated assessment models in order to emphasise the basic relationship among different categories of environmental processes such as physical, morphological, chemical, ecological and socio-economic – and to provide inclusive information about the environmental and socioeconomic processes. As far as the software interfaces are concerned, very few of the examined DSS are applied through webbased interfaces, in spite of the fact that web-based facilities enhance easy access to information within a large network of users. Furthermore, all the reviewed DSS consider GIS tools as basic media to express their results or outputs in order to provide fast and intuitive results representation to non-experts (i.e. decision makers and stakeholders) and empower them for robust decisions. In addition to maps, the outputs produced by each DSS are also graphs, charts, and statistical tables. 5. APPLICABILITY CRITERIA Table 5 shows the implementation of the criteria concerning applicability to the examined DSS. Applicability includes three aspects: scale/study areas, flexibility and status/availability (Table 2). The spatial scales considered were five: global, supranational, national, regional, and local, in order of decreasing size. The study areas are those reported in the literature cited in Table 1. The flexibility derives from the capability of a given DSS to include new modules and models in its structure, thus new input parameters, and the suitability to be used for regionally different case studies. In order to visualize the estimation of the overall flexibility of a system, highly flexible/flexible/moderately-to-no flexible were indicated as +++/++/+. Status and availability refer to different extent of development (e.g. research prototype, commercial software) and public accessibility/last updated version, respectively.

20


Table 5. List of the examined DSSs according to the applicability criteria (+++, highly flexible; ++, flexible; +: moderately to no-flexible). Name

Scale and area of application

Flexibility

Status and availability last updated version (year)

CLIME

•Supra-National, National, Local.

+++

(Northern, western and central part of

Flexible

in

structural

2010.

modification and study area.

Europe). CORAL

Available to the public. Demo.

•Regional, Local.

+++

Not

available

(Coastal areas of Curacao; Jamaica and

Flexible in study area.

Prototype.

to

the

public.

1995.

Maldives). •National, Local.

++

Commercial application.

(Coast of Netherland).


1998.

Coastal

•National, Regional, Local.

+

Available only to the Tyndall

Simulator

(Coast of Norfolk in East Anglia, UK).

COSMO

Research Centre. Prototype. 2009

CVAT

DESYCO

•Regional, Local.

++

Available to public. Prototype.

(New Hanover County, North Carolina).


2002.

•Regional, Local.

++

Not available to the public.

(North Adriatic Sea).


Prototype. 2010.

DITTY

•Supranational, National, Regional.

+++


(Ria Formosa-Portugal; Mar Menor-


2006

+++

Available to the public.


2009

+

Available to the public.

Spain; Etang de Thau-France; Sacca di Goro-Italy, Gera-Greece). DIVA

ELBE

•Global, National.

•Local.

2003

(Elbe river basin Germany). GVT

+

•Regional, Local. (Eastern

Macedonia

and

Not available to the public. 2006

Northern

Greece). IWRM

•Regional, Local.

++


(Halti-Beel, Bangladesh)


Prototype. 2009

21


KRIM

+

• Regional.

Not available to the public. Prototype.

(German North sea Coast, Jade-Weser

2003

area in Germany). MODSIM

• National, Regional.

++

Available to the public online.

(San Diego Water County, Geum river


2006

•Regional, Local.

++

Available online to stakeholders.

(North-West, East Anglia).


Prototype.

basin- Korea). RegIS

2008 RAMCO

•Regional, Local.

++

Not

(South-West Sulawesi coastal zone).

Flexible in the used dataset and

Prototype.

concepts.

1999

+

Available online to the public.

SimLUCIA •Local

available

to

the

public.

Demo.

(St Lucia Island, West India)

1996 SimCLIM

++

•National, Regional, Local. (Rarotonga

Island,

Southeast

Queensland). STREAM

Available to the public. Demo.

Flexible

structural

2009


•Regional, Local.

+++

(Ganges/Brahmaputra river basin, Rhine

Flexible

river basin, Yangtze river basin and

in

Available online to the public. in

structural

Demo.


1999

+

Available to National Taiwan

Amudarya river basin). TaiWAP

•Regional, Local.

University. Prototype.

(Touchien river basin).

2008 WADBOS

•Regional, Local.

+

Available online to the public. Demo.

(Dutch Wadden sea).

2002

As far as the scale of application is concerned, all the examined DSS, except DIVA, have been applied only at the local and regional scales because they were developed for a specific geographical context. Moreover, five out of the 20 examined DSS (i.e. CLIME, CORAL, DITTY, DIVA and STREAM) considered global, supranational, national, regional and local scales during their implementation. Five of the reported DSS are highly flexible systems because they are used to address several impacts related to different case studies.

Although DIVA can be applied to any coastal area around the world, it is sometimes not considered a highly flexible tool in terms of structural modification due to its inability to change its default integrated dataset. Finally, ELBE and WADBOS are identified as moderately-to-no flexible systems because their structure and functionalities were based on the specific needs of particular river basins.

22


The applicability of DSS reflects their ability to be implemented in several contexts (i.e. case study areas and structural modification), for example to include new models and functionalities ensuring common approaches to decision making and the production of comparable results [42]. Finally, concerning the availability and the status of the development, Table 5 shows that nine DSS are available to the public, three are available with a restricted access (i.e. only to stakeholders or to the developers), one is a commercial software (i.e. COSMO) and seven are not available to the public. Sometimes the restriction of the access is due to the fact that results require special skill for their interpretation, so the public can use them only with the support of the developer team. Among examined DSS, only 11 were developed/updated during the last 5 years, and 4 over the previous five years (for a total of 15 during the last 10 years) with the remaining five DSS showing the last version dating back to the ‘90s. The overall content of Table 5, together with the main features of each DSS reported in Tables 3 and 4, allow the reader to undertake a screening evaluation of available DSS in relation to the specific impacts from climate change to be addressed. 6. CLIMATE CHANGE & DSS FUNCTIONALITIES Among the challenges of coastal environmental problems identified by [23, 8, 43 and 8] the paper elicits those related to climate change and categorises them into assessment and management aspects – bearing in mind that scientific solutions to climate change are often based on assessment and management procedures which are very contingent because assessment methodologies or approaches, data and tools could determine the robustness of potential management measures. Thus, the examined DSS functionalities necessary to cope with climate change can be evaluated from an in depth consideration of framed questions intended to reflect the significant coastal systems challenges. Assessment • Does the DSS consider interdisciplinary processes/modelling? • Does the DSS support spatial and temporal dimensions of coastal issues? • Does the DSS consider uncertainty range or incomplete knowledge? • Does the DSS support sensitivity analysis? • Does the DSS predict potential effects of proposed scenarios? Management • Does the DSS consider the integration of science and policy / stakeholders involvement? • Does the DSS support optimisation of management measures? • Does the DSS make complex information understandable / aid visualization of processes? An attempt to answer these questions, the paper synthesised the information elicited from the open literature survey in Table 3, 4 and 5. The results reflect the fact that, none of these tools possess all the functionalities related to both the assessment and management aspects.

However, they all appear to support the spatial and temporal dimensions of coastal processes; prediction of scenarios outcomes; integrated analysis of issues via in-inclusion of several models and approaches and making complex processes understandable through visualisation techniques e.g. GIS, 2D and 3D models etc. It should be noted, none of these DSS prove adequate sensitivity analysis of climate variables. Whereas only three (Coastal Simulator, CLIME and RegIS) partly consider uncertainty range via the application of the Monte Carlo Simulation and climate change projection analysis. RegIS adopts a novel 3D visualisation in order to communicate uncertainty associated with future coastal change modelling [33]. Nine out of the twenty DSS (COSMO, CVAT, DIVA, IWRM, KRIM, RegIS, SimLUCIA, SimCLIM and STREAM) partly support the optimisation of management measures, by considering effects related to different protection plans and, cost-benefit, socio-economic and mitigation options analysis. To a large extent stakeholders’ participation is not fully supported by these tools even though there could be workshops and capacity building during development phases. Nonetheless potential users cannot use these tools effectively; for instance, four out of the twenty systems (ELBE, RegIS, KRIM and IWRM) support the provision of information for non-technical experts among which only RegIS can be used by stakeholders without the intervention of expert. 7. CONCLUSIONS This work should be regarded as a preliminary attempt to describe and evaluate the main features of available DSS for the assessment and management of climate change impacts on coastal area and related inland watersheds. A further and comprehensive evaluation should be based on comparative application in selected and relevant case studies, in order to evaluate the DSS technical performance, especially in relation to datasets availability, that often represents the real limiting factor. Moreover, sensitivity and uncertainty analyses will provide further evidence of the reliability of the investigated DSS. This review highlighted the relevance of developing climate change impact assessment and management at the regional scale (i.e. subnational and local scale), according to the requirements of policy and regulatory frameworks and to the methodological and technical features of the described DSS. In fact, most of the available DSS show a regional to local applicability with a moderate to high flexibility. Indeed climate change impacts are very dependent on regional geographical features, climate and socio-economic conditions and regionally-specific information can assist coastal communities in planning adaptation measures to the effects of climate change. Despite the current situation that shows available DSS mainly focusing on the analysis of specific individual climate change impacts and affected sectors (15 out of the 20 examined DSS), the further developments should aim at the adoption of ecosystem approaches considering the complex dynamics and interactions between coastal systems and other systems closely related to them (e.g. coastal aquifers, surface waters, river basins, estuaries).

23


The adoption of multi-risk approaches in order to consider the interaction among different climate change impacts that affect the considered region should also be a focus. Finally, it is important to remark the need to involve the end users and relevant stakeholders since the initial steps of the development process of these tools, in order to satisfy their actual requirements, especially in the perspective of providing useful climate services, and to avoid the quite often and frustrating situation where time and resource demanding DSS are not used beyond scientific testing exercises.

[10]

IPCC, Climate Change (2007): Impacts, Adaptation and Vulnerability, Summary for Policymakers, Contribution of Working Group II to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Geneva, 2007.

[11]

Nakicenovic, N., Alcamo, J., Davis, G., de Varies, B., Fenhann, J., Gaffin, S., Gregory, K., Grubler, A., Jung, T.Y. and Kram, T. (2000). Special report on emissions scenarios: a special report of Working Group III of the Intergovernmental Panel on Climate Change, Pacific Northwest National Laboratory, Richland, WA (US), Environmental Molecular Sciences Laboratory (US).

[12]

Nicholls, R.J., Cazenave, A., (2010). Sea-level rise and its impact on coastal zones. Science, 328(5985):1517-1520.

[13]

Jiang L., Hardee K.., (2010). How do Recent Population Trends Matter to Climate Change? Popul re policy rev, 30(2):287-312

[14]

EC, COM(2007) 354, 29.06.2007. Green Paper – Adapting to climate change in Europe – options for EU action, Brussels.

[15]

EC, COM(2009) 147, 01.04.2009. White Paper – Adapting to climate change: Towards a European framework for action, Brussels.

[16]

EC (2000) Directive 2000/60/EC ‘‘Directive 2000/60/EC of the European Parliament and of the Council Establishing a Framework for the Community Action in the Field of Water Policy’’ Official Journal (OJ L 327) on 22 December 2000.

[17]

EC (2002) Recommendation the European Parliament and of the Council of 30 May 2002 Concerning the Implementation of Integrated Coastal Zone Management in Europe, 2002/413/EC.

[18]

Fabbri, K.P., (1998). A methodology for supporting decision making in integrated coastal zone management. Ocean cost manage, 39(1-2), pp.51–62. Environmental Systems Research Institute – ESRI, (1992). Arc Version 6.1.2, Redlands, California, USA.

EC (2003) Common Implementation Strategy for the Water Framework Directive (2000/60/CE). Guidance Document n. 11. Planning process. Office for Official Publications of the European Communities, Luxembourg.

[19]

Thumerer, T., A. P. Jones, and D. Brown. (2000) “A GIS Based Coastal Management System for Climate Change Associated Flood Risk Assessment on the East Coast of England.” Int j geogr inf sci 14(3):265–281.

Nobre, A.M. & Ferreira, J.G. (2009). Integration of ecosystem-based tools to support coastal zone management. J costal res, 1676-1670.

[20]

Flax, L.K., Jackson, R.W. & Stein, D.N. (2002). Community vulnerability assessment tool methodology. Natural Hazards Review, 3:163.

ACKNOWLEDGEMENTS The authors gratefully acknowledge the Euro-Mediterranean Centre for Climate Change (CMCC; Lecce, Italy), GEMINA project, for financial support.

REFERENCES [1]

[2]

Janssen, R. (1992). Multiobjective decision support for environmental management. Kluwer Academic. (Dordrecht) Boston. Torresan, S., Zabeo, A., Rizzi, J., Critto, A., Pizzol, L., Giove, S. and Marcomini, A. (2010). Risks assessment and decision support tools for the integrated evaluation of climate change impacts on coastal zones. International Congress on Environmental Modelling and Software Modelling for Environmental Sake, Fifth Biennial Meeting, Ottawa, Canada.

[3]

Matthies, M., Giupponi, C. and Ostendorf, B. (2007). Environmental Decision Support Systems; Current Issues, Methods and tools Environ model soft 22(2):123-28.

[4]

Uran, O. & Janssen, R. (2003). Why spatial decision support systems not used? Some experiences from the Netherlands. Comput environ urban, 27(5):511–526.

[5]

[6]

[7]

[8]

[9]

Poch, M., Comas, J., Rodriguez-Roda, I., Sachez-Marrie, M. and Cortes, U. (2004). Designing and building real environmental decision support systems. Environ modell soft 19:857-873.

Matthies, M., Giupponi, C., Ostendorf, B. (2007). Environmental Decision Support Systems; Current Issues, Methods and tools. Environ modell soft 22(2):123-28.

24


[21]

Zaman, A. M., S. M. M. Rahman, and M. R. Khan. “Development of a DSS for Integrated Water Resources Management in Bangladesh” (2009), 18th World IMACS / MODSIM Congress, Cairns, Australia 13-17 July.

[22]

Jolma, A., Kokkonen, T., Koivusalo, H., Laine, H., Tiits, K. (2010). Developing a Decision Support System for assessing the impact of climate change on lakes. The Impact of Climate Change on European Lakes, 411–435.

[32]

Westmacott, S. (2001). Developing Decision Support Systems for Integrated Coastal Management in the Tropics: Is the ICM Decision-Making Environment Too Complex for the Development of a Useable and Useful DSS? J environ manage, 62(1): 55-74.

Labadie, J.W. (2006). MODSIM: decision support system for integrated river basin management. Summit on Environmental Modeling and Software, the International Environmental Modeling and Software Society, Burlington, VT USA.

[33]

Holman, I.P., Rounsevell, M.D.A., Berry, P.M., Nicholls, R.J. (2008). Development and application of participatory integrated assessment software to support local/regional impacts and adaptation assessment. Journal of Earth and Environmental Science. DOI: 10. 1007/s 10584008-9452-7.

[30]

[31]

[23]

Schirmer, M., B. Schuchardt, B. Hahn, S. Bakkenist & D. Kraft (2003): KRIM: Climate change risk construct and coastal defence. DEKLM German climate research programme. Proceedings, 269-273 Salewicz, K. A. and M, Nakayama. (2004): Development of a web-based decision support system (DSS) for managing large international rivers. Global environ chang, 14:25-37

[24]

Feenstra, J.F., Programme, U.N.E. & Milieuvraagst, V. (1998). Handbook on methods for climate change impact assessment and adaptation strategies, United Nations Environment Programme.

[25]

Nicholls, R., Mokrech, M. & Hanson, S. (2009). An integrated coastal simulator for assessing the impacts of climate change. In IOP Conference Series: Earth and Environmental Science. 6 092020.

[34]

De Kok, J.L., Engelen, G., White, R. and Wind, H. G. (2001). Modelling land-use change in a decision-support system for coastal-zone management. Environ model assess, 6(2):123– 132.

[26]

Agnetis A., Basosi R., Caballero K., Casini M., Chesi G., Ciaschetti G., Detti P., Federici M., Focardi S., Franchi E., Garulli A., Mocenni C., Paoletti S., Pranzo M., Tiribocchi A., Torsello L., Vercelli A., Verdesca D., VicinoA. (2006). Development of a Decision Support System for the management of Southern European lagoons. Centre for Complex Systems Studies University of Siena, Siena, Italy. Technical Report TR2006-

[35]

Uljee, I, Engelen, G. And White, R. (1996) Rapid Assessment Module for Coastal Zone Management (RAMCO). Demo Guide Version 1.0, Work document CZM-C 96.08, RIKS (Research Institute for Knowledge Systems).

[36]

Engelen, G., White, R., Uljee, I. and Wargnies, S. (1995). Vulnerability Assessment of Lowlying Coastal Areas and Small Islands to Climate change and sea-level rise. Report to the United Nations Environment Programme, Caribbean Regional Co-ordinating Unit, Kingston, Jamaica (905000/9379). Maastrict: Research Institute Knowledge Systems (RIKS).

[37]

Warrick, R. A. (2009): Using SimCLIM for modelling the impacts of climate extremes in a changing climate: a preliminary case study of household water harvesting in Southeast Queensland. 18th World IMACS/MODSIM Congress, Cairns, Australia 13-17 July 2009.

[38]

Aerts, J., Kriek, M. & Schepel, M. (1999). STREAM (Spatial tools for river basins and environment and analysis of management options): set up and requirements’. Phys chem earth, Part B: Hydrology, Oceans and Atmosphere, 24(6):591–595.

[27]

Hinkel, J. & Klein, R.J.T. (2009). Integrating knowledge to assess coastal vulnerability to sealevel rise: The development of the DIVA tool. Global environ chang, 19(3):384–395.

[28]

BfG, (2003). Plot phase for the design and development of a decision support system (DSS) for river basin management with the example of the Elbe (Interim Report 2002-2003 (in German). Bundesanstalt fur Gewässer-kunde (German Federal Institute of Hydrology), Koblenz. Germany.

[29]

Gemitzi, A. et al., 2006. Assessment of groundwater vulnerability to pollution: a combination of GIS, fuzzy logic and decision making techniques. Environ geol, 49(5):653– 673.

25


[39]

Liu, M.T., Tung, C.P., Ke, K.Y., Chuang, L.H, Lin, C.Y. (2009): Application and development of decision support systems for assessing water shortage and allocation with climate change. Paddy Water Environ, 7:301-311. DOI 10.1007/s/0333-009-0177-7.

[40]

van Buuren, J.T., Engelen, G. & van de Ven, K. (2002). The DSS WadBOS and EU policies implementation.

[41]

[42]

Agostini, P., Suter, G.W.II, Gottardo, S., Giubilato, E. (2009). Indicators and endpoints for risk-based decision process with decision support systems. In Marcomini, A., Suter, G.W.II, Critto, A. (Eds). Decision Support Systems for Risk Based Management of Contaminated Sites. Springer Verlag, New York.

[43]

Van Kouwen, F., C. Dieperink, P. Schot, and M. Wassen. “Applicability of Decision Support Systems for Integrated Coastal Zone Management.” Coastal Management 36, no. 1 (2007): 19–34.

Engelen G. (2000). The WADBOS Policy Support System: Information Technology to Bridge Knowledge and Choice. Technical paper prepared for the National Institute for Coastal and Marine Management/RIKZ. The Hague, the Netherlands.

26



Data Mining Technique for Predicting Telecommunications Industry Customer Churn Using both Descriptive and Predictive Algorithms

Kolajo Taiwo Computer Science Department Federal College of Education (Technical) Bichi, Kano, Nigeria [email protected]

Adeyemo, A.B. Computer Science Department University of Ibadan, Ibadan, Nigeria [email protected]

Reference Format:

Kolajo, T & Adeyemo, A.B. (2012). Data Mining Technique for Predicting Telecommunications Industry Customer Churn Using both Descriptive and Predictive Algorithms. Computing, Information Systems & Development Informatics Journal. Vol 3, No.2. pp 27-34

Data Mining Technique for Predicting Telecommunications Industry Customer Churn Using both Descriptive and Predictive Algorithms 27


Kolajo, T & Adeyemo, A.B

ABSTRACT As markets have become increasingly saturated, companies have acknowledged that their business strategies need to focus on identifying those customers who are most likely to churn. It is becoming common knowledge in business, that retaining existing customers is the best core marketing strategy to survive in industry. In this research, both descriptive and predictive data mining techniques were used to determine the calling behaviour of subscribers and to recognise subscribers with high probability of churn in a telecommunications company subscriber database. First a data model for the input data variables obtained from the subscriber database was developed. Then Simple K-Means and Expected Maximization (EM) clustering algorithms were used for the clustering stage, while Decision Stump, M5P and RepTree Decision Tree algorithms were used for the classification stage. The best algorithms in both the clustering and classification stages were used for the prediction process where customers that were likely to churn were identified. Keywords: customer churn; prediction; clustering; classification

1. INTRODUCTION The mobile telephony market is one of the fastest growing service segments in telecommunications, and more than 75% of all potential phone calls worldwide can be made through mobile phones and as with the any other competitive markets, the mode of competition has shifted from acquisition to retention of customers [1]. Among all industries that suffer customer churn the telecommunications industry can be considered as being at the top of the list with an approximate annual rate of 30% [2]; [3]. This results in a waste of money and effort and “is like adding water to a leaking bucket” [4]. Considering the fact that the cost to European and US telecommunication companies is US$ 4 billion per year, then it seems reasonable to invest more on churn management rather than acquisition management for mature companies, especially when it is noted that the cost of acquiring new customer is eight times more than that of retaining an existing one [3]. On the other hand, existing subscribers tend to generate more cash flow and profit, since they are less sensitive to price and often lead to sales referrals [5]. Due to the high cost of acquiring new subscribers and considerable benefits of retaining the existing ones, building a churn prediction model to facilitate subsequent churn management and customer retention is critical for the success or bottom-line survival of a mobile telecommunications provider in this greatly compressed market-space. Subscriber churning (often referred to as customer attrition in other industries) in mobile telecommunication refers to the movement of subscribers from one provider to another. Many subscribers frequently churn from one provider to another in search of better rates or services. Churning customers can be divided into two main groups, voluntary churners and non-voluntary churners. Non-voluntary churn is the type of churn in which the service is purposely withdrawn by the company. Voluntary churn is more difficult to determine. This type of churn occurs when a customer makes a conscious decision to terminate his/her service with the provider. This type of churn has been a serious and puzzling problem for service providers.

The varied behaviour of consumers has baffled researchers and market practitioner’s alike [6]. Voluntary churn can be divided into two sub categories, incidental churn and deliberate churn. Incidental churn happens when changes in circumstances prevent the customer from further requiring the provided service and is a small percentage of a company’s voluntary churn [7]. Deliberate churn is the problem that most churn management solutions attempt to identify. This type of churn occurs when a customer decides to move to a competing company due to reasons of dissatisfaction [1]. Deliberate churn within the telecommunications industry is minimised because switching would require a change in the telephone number. In 2003 customers in the United States of America were given the option to switch mobile telephone provider but keep their existing phone number, and as soon as this law came into force 12 million customers immediately churned from their service providers thereby increasing the retention battle [8]. A churn management solution should not target the entire customer base because (i) not all customers are worth retaining, and (ii) customer retention costs money; attempting to retain customers that have no intention of churning is a waste of resources. Nowadays lack of data is no longer a problem, but the inability to extract useful information from data [9]. Due to the constant increase in the amount of data efficiently operable to managers and policy makers through the high speed computers and rapid data communication, there has grown and will continue to grow a greater dependency on statistical methods as a means of extracting useful information from the abundant data sources. To survive or maintain an advantage in an ever-increasing competitive marketplace, many companies are turning to data mining techniques to address churn prediction and management [10]. Data mining is the analysis of (often large) observational data sets to find unsuspected relationships and to summarize the data in novel ways that are both understandable and useful to the data owner [11]. Data mining is an interdisciplinary field bringing together techniques from machine learning, pattern recognition, statistics, databases, and visualization to address the issue of information extraction from large data bases. Data mining techniques include a wide range of choices from many disciplines. These choices include techniques such as support vector machines, correlation, linear regression, nonlinear regression, genetic algorithms, neural networks,

28


and decision trees. The choice of a data mining technique is contingent upon the nature of the problem to be solved and the size of the database. Based on the kind of knowledge which can be discovered from databases, data mining techniques can be broadly classified into several categories including clustering, classification, dependency analysis, data visualization and text mining [12]. Clustering analysis is a process whereby a set of instances (without a predefined class attribute) is partitioned (or grouped) according to some distance metric into several clusters in which all instances in one cluster are similar to each other and different from the instances of other clusters. Classification is a model that induces a model to categorize a set of pre-classified instances (called training examples) into classes. Such a classification model is now used to classify future instances. Clustering is a way to segment data into groups that are not previously defined, whereas classification is a way to segment data by assigning it to groups that are already defined. Dependency analysis discovers dependency patterns (e.g. association rules, sequential patterns, temporal patterns, and episode rules) embedded in data. Data Visualization allows decision makers to view complex patterns in the data as visual objects in three dimensions and colour; it supports advanced manipulation capabilities to slice, rotate or zoom the objects to provide varying levels of details of the patterns observed.

order to utilize them in building the required and targeted features: a. Phone No of each subscriber b. Incoming Calls c. Incoming Start Time d. Incoming Duration e. Outgoing Calls f. Outgoing Start Time g. Outgoing Duration 2.2 Data Mining Figure 1 presents the Data Mining framework developed for this work. Both descriptive and predictive data mining techniques were used. In the descriptive step, the customers were clustered based on their usage behavioural (RFM) feature. K-means and (EM) Expected Maximization clustering methods were used for the clustering. K-means clustering is a partitioning method that treats observations in data as objects having locations and distances from each other. It partitions the objects into K mutually exclusive clusters, such that objects within each cluster are as close to each other as possible, and as far from objects in other clusters as possible. Each cluster is characterized by its centroid, or centre point. EM (Expectation Maximization) assigns a probability distribution to each instance which indicates the probability of it belonging to each of the clusters. EM can decide how many clusters to create by cross validation, or you may specify apriori how many clusters to generate.

In this study both descriptive model and predictive data mining techniques will be used to extract information on the calling behaviour of subscribers and to recognise subscribers with high probability of churn in the future. While some researchers have focused on the use of either descriptive or descriptive algorithms, in this work both algorithms will be combined. First the elements of the dataset will be grouped by clustering them and then classification algorithms will be applied to the clusters of interest so that each cluster's unique "rules" for relating attributes to classes can be determined and thereby more accurately classify the members of each cluster. The dataset used were obtained from a Nigerian Telecommunications service provider. 2. MATERIALS AND METHODS The Customer churn prediction model was developed based on some selected input variables from a Nigerian telecommunications service provider customer database. Fig 1. Data Mining Framework 2.1 Data Selection and Preprocessing The data used were from the call records of subscribers in one of the telecommunications service providers in Nigeria. The total number of records in the dataset is 228,520. The records were for transactions covering a period of 3 months from October 1st – December 31st 2010. The raw data was uploaded into Mysql database for the extraction of necessary features from the raw data. The features that were selected were based on those used by [12], [13] and the RFM (Recency, Frequency, Monetary) related features. These features were chosen due to the nature of pre-paid service providers. The focus is on constructing features that are able to reflect the changes in usage behavior. The final dataset used consisted of 996 subscriber call records and consisted of the following data variables which were selected from the call records in

For the predictive step, classification techniques were utilized. Classification is the process of finding a model (or function) that describes and distinguishes data classes or concepts for the purpose of being able to use the model to predict the class of objects whose class label is unknown. Decision tree was chosen because it is capable of efficiently generating interpretable knowledge in an understandable form. Models from tree classifier (DecisionStump, M5P, and RepTree) were used. DecisionStump is a model consisting of a one-level decision tree. That is, it is a decision tree with one internal node (the root) which is immediately connected to the terminal nodes. A decision stump makes a prediction based on the value of just a single input feature. Sometimes they

29


are also called 1-rules. The algorithm builds simple binary decision ‘stumps’ (1 level decision trees) for both numeric and nominal classification problems. It copes with mission values by extending a third branch from the stump or treating ‘missing’ as a separate attribute value. DecisionStump is usually used in conjunction with a boosting algorithm such as LogitBoost. It does regression (based on mean-squared error) or classification (based on entropy). M5P implements base routines for generating M5 Model trees and rules. A learning technique that consistently yields the best results is M5P regression trees. RepTree is a fast decision tree learner. It builds a decision/regression tree using information gain/variance and prunes it using reduced-error pruning (with backfitting). The algorithm only sorts values for numeric attributes once. Missing values are dealt with by splitting the corresponding instances into pieces. 2.3 Building the Clustering Model Weka was leveraged with Java in building the clustering model. SimpleKmeans wrapped up with MakeDensityBased Clusterer and EM (Expectation Maximization) was used for the clustering. The following set of 12 RFM related variables were constructed by the use of Mysql in order to segment the subscriber based on their calling behavior: 1.

Call Ratio: Proportion of calls which has been made by each subscriber to his/her total number of calls (incoming and outgoing calls). 2. Max Date: The last date in our observed period in which a subscriber has made a call. 3. Min Date: The first date in our observed period in which a subscriber has made a call. 4. Average Call Distance: The average time distance between one’s calls. 5. Life: The period of time in our observed time span in which each subscriber has been active. 6. Max-Distance: The maximum time distance between two calls of a specific subscriber in our observed period. 7. No-of-days: Number of days in which a specific subscriber has made or received a call. 8. Total-no-in: The total number of incoming calls for each subscriber in our observed period. 9. Total-no-out: The total number of outgoing calls for each subscriber in our observed period. 10. Total Cost: The total money that each subscriber has been charged for using the services in the specific time period under study. 11. Total-duration-in: The total duration of incoming calls (in Sec) for a specific subscriber in our observed time span. 12. Total-duration-outgoing: The total duration of outgoing calls (in Sec) for a specific subscriber in our observed time span.

a. b. c.

For every single cluster the following features were extracted: a. MOUinitial: this represents the MOU of a subscriber in the first sub-period. b. FOUinitial: this represents the FOU of a subscriber in the first sub-period. c. SOIinitial: this represents the SOI of a subscriber in the first sub-period. d. ∆MOUs: this represents the change in MOU of a subscriber between the sub-period s – 1 and s (for s=2, n) and is measured by ∆MOUs = (MOUs – MOUs-1+δ)/( MOUs-1+ δ), where MOU1 = MOUinitial and δ is a small positive real number (e.g. 0.01) to avoid the case when MOUs-1 is 0 (i.e. when ∆MOUs cannot be calculated). e. ∆FOUs: this represents the change in FOU of a subscriber between the sub-period s – 1 and s (for s=2,...n) and is calculated as ∆FOUs = (FOUs – FOUs-1+δ)/( FOUs-1+ δ). f. ∆SOIs: this represents the change in SOI of a subscriber between the sub-period s – 1 and s (for s=2,...n) and is calculated as ∆SOIs = (SOIs – SOIs-1+δ)/( SOIs-1+ δ). Using Decision Tree algorithms the predictive models were constructed for each of the clusters. 3. RESULTS AND DISCUSSION The following presents the results of the descriptive and predictive models. The descriptive model was used to describe the calling behaviour of the subscribers while the predictive model was used for prediction of subscribers who are likely to churn. 3.1 Clustering (Descriptive) Model Result In the descriptive model, SimpleKMeans and EM (Expected Maximization) algorithms were used in describing the customer behaviour. EM performed better than the SimpleKMeans algorithm. The performance measure used was the log likelihood. The log likelihood measures how well an algorithm has performed; the closer to zero the better the performance of the algorithm. From the Log likelihood value of both the SimpleKMeans clustering and EM clustering: a. The Log likelihood value of table the SimpleKMeans is -58.56228 b.

2.4 Building the Predictive Model Among the call details maintained in the investigated company, three measures commonly used to describe call patterns of a subscriber by aggregating his/her recall records which are:

Minutes of use (MOU): this refers to the total number of minutes of outgoing calls made be the subscriber over a specific period. Frequency of use (FOU): this refers to the total number of outgoing calls made by the subscriber over a specific period. Sphere of influence (SOI): this refers to the total number of distinctive receivers contacted by the subscriber over a specific period.

The Log likelihood value of table the EM is 58.0476

-

The EM result is better than that of the SimpleKMeans. Hence EM algorithm was used for the clustering model. The analysis of each of the attributes used is presented in a graphical form.

30


3.1.1 Call Ratio Call Ratio is the proportion of calls which has been made by each subscriber to his/her total number of calls (incoming and outgoing calls). Almost all the clusters are having the same call ratio except for cluster 10 with 0.3. The more a subscriber calls the more likelihood he /she is retained and the more turnover for the telecom service provider. Hence the service provider should intensify effort in deploying strategy that will encourage subscribers in cluster 10 to make more calls.

as quality of the service, coverage, price, etc. For instance if a subscriber relocates to a location where there is no network coverage, the subscriber will need to go for another network.

Fig 4. Life for each Cluster using EM (Expected Maximization) 3.1.4 Max Distance Max-Distance is the maximum time distance between two calls of a specific subscriber in our observed period. Clusters with high maximum time distance represents the subscribers that have not been calling regularly. The higher the maximum time distance the more tendencies for the subscribers to churn. As a result retention efforts should be focused on the subscribers that fall into clusters 8, 9 and 10. Fig 2. Call-Ratio for each Cluster using EM (Expected Maximization) 3.1.2 Average Call Distance The Average Call Distance is the average time distance between one’s calls. From figure 3, a cluster with high average call distance implies that the subscribers are not making call regularly. Clusters 8 and 9 fell into this category. This might be due to a number of reasons including getting the same service at lower cost from other service provider, quality of service to mention a few. Hence they are likely to churn in the nearest future. The telecommunications service provider should intensify retention efforts on them so as to win them back.

Fig 5. Max-Distance for each Cluster using EM (Expected Maximization) 3.1.5 No of Days No-of-days stands for the number of days in which a specific subscriber has made or received a call. The total number of days in the observed period was ninety (90) days. The number of days for clusters 8, 9 and 10 is far below average and this implies that they have not been active. For them to be won back retention efforts have to be focused on them otherwise in the nearest future they are likely to churn.

Fig 3. Average-Call Distance for each Cluster using EM (Expected Maximization) 3.1.3 Life Life represents the period of time in our observed time span in which each subscriber has been active. Those subscribers that fall in the category of clusters 8, 9 and 10 are likely to churn which may be due to some factors such

31


Fig 6. No-of-Days for each Cluster using EM (Expected Maximization) 3.1.6 Total No In Total-no-in is the total number of incoming calls for each subscriber in our observed period. When a subscriber stops receiving calls though a network; it points to the fact that the subscriber might not be interested in the network again because if he/she is making calls through that network he will definitely be communicated back through that same network. Subscribers in clusters 2, 6, and especially clusters 8, 9 and 10 should be tracked to know what actually went wrong. From the investigation telecommunications service provider will now be informed of the kind of retention effort to deploy.

Fig 8. Total-No-Out for each Cluster using EM (Expected Maximization) 3.1.8 Total Cost Total Cost is the total money that each subscriber has been charged for using the services in the specific time period under study. The more money spent the likelihood that the subscriber is satisfied with the network services and vice versa. Price is the most determinant factor here because the lower the price the more the total turnover for the service provider and the more calls made by the subscribers. From figure 4.8, the clusters 2, 3, 6, 8, 9 and 10 have the lowest total cost. Retention efforts should be focused on those subscribers that form those clusters.

Fig 7. Total-No-In for each Cluster using EM (Expected Maximization) 3.1.7 Total No Out Total-no-out represents the total number of outgoing calls for each subscriber in our observed period. A subscriber that stops making calls through a network will definitely not be receiving call through that network. The subscribers that are in clusters 2, 3, 6, 8-10 are likely to churn.

Fig 9. Total-Cost for each Cluster using EM (Expected Maximization) 3.1.9 Remarks 1. Subscribers in clusters 8, 9 and 10 have not been responding well and are very likely to churn in the nearest future. Hence they should form the focus of targeted campaign in order to win them back. 2. Subscribers in clusters 2, 3 and 6 are slightly different, their call ratio and no of days attributes were still fairly okay. With demographic data further investigation could be carried out. For instance they could be students in school. Therefore a package could be developed for them in other to encourage them.

32


3.2 The Classification (Predictive) Model Result The DecisionStump, M5P, and RepTree classifier algorithm implemented in WEKA were used. Five performance measures were used in determining the performance of these algorithms on the dataset. These were: i. Correlation coefficient (CC): This measures the degree of correlation or relationship among the attributes. It ranges between 1 for high positive correlation to -1 for high negative correlation, with 0 indicating a purely random relationship. ii. Mean Absolute Error (MAE): This is a quantity used to measure how close forecasts or predictions are to the eventual outcomes. As the name suggests, the mean absolute error is an average of the absolute errors. MAE can range from 0 to ∞. It is a negativelyoriented score: Lower values are better. iii. Root Mean Squared Error (RMSE): The RMSE is a quadratic scoring rule which measures the average magnitude of the error. In other words, it represents the difference between forecast and corresponding observed values are each squared and then averaged over the sample. Finally, the square root of the average is taken. Since the errors are squared before they are averaged, the RMSE gives a relatively high weight to large errors. This means the RMSE is most useful when large errors are particularly undesirable. It can range from 0 to ∞. It is a negatively-oriented score: Lower values are better. iv. Relative Absolute Error (RAE): This takes the total absolute error and normalizes it by dividing by average of the actual values. Lower values are better. v. Root Relative Squared Error (RRSE): Instead of total absolute error as in RAE, it takes total squared error and divides by the average of the actual values. Finally, the square root of the result is taken. Cluster’s 8, 9 and 10 which contains the subscribers that are likely to churn now becomes the focus of the next stage where the actual churners are determined. Table 1 presents the classification algorithm results. Table 1: Classification Algorithm Result M5P performed better than both DecisionStump and Reptree. Hence M5P algorithm was used in building the predictive model on each cluster. The most significant features in building the predictive model for clusters 8, 9 and 10 are presented in a Table 2. Table 2: Determinant Features for Clusters 8, 9 and 10 Cluster Determinant Features Number 8 ∆FOUs, ∆SOIs, MOU_Final, FOU_Initial, FOU_Final and SOU_Initial 9 ∆SOIs, ∆MOUs, FOU_Initial, FOU_Final 10 ∆MOUs, ∆SOIs, MOU_Final, FOU_Initial and SOU_Initial

3.2.3 M5P Result on Cluster 9 The M5 pruned model tree (using smoothed linear models) generated the following decision tree: ISOIs -0.527 : | FOU_Initial 26 : | | FOU_Final 24.5 : | | | FOU_Initial 37.5 : | | | | FOU_Initial 77 : LM11 (5/22.962%) Cluster 9 had a total of 196 instances. A total of 108 subscribers that fall under the rules LM1:LM4 were classified as churners while the remaining 88 subscribers under the rules LM5:LM11 are classified as non-churners. 3.2.5 M5P Result on Cluster 10 The M5 pruned model tree (using smoothed linear models) generated the following decision tree: ISOIs -0.48 : LM2 (20/47.175%) A total of 75 instances were in cluster 10. Seventy (75) subscribers in cluster 10 that fall under the rule LM1 were classified as churners while the remaining 20 subscribers under the rules LM2 were classified as non-churners. 4. CONCLUSION Inability to distinguish the churner from non-churner has been the problem of telecommunications service provider. There are two alternatives; either to send incentives to all customers (both churners and non-churners), which will be tantamount to a waste of money or to focus on acquisition program (that is, acquisition of new customers) which is more costly than retention effort. Since both alternatives have their negative implication on the finance of the company, distinguishing churner from non-churner is the best approach. Ability of the telecom service to distinguish between churner and non-churner is the central idea and the achievement of this research.

3.2.1 M5P Result on Cluster 8 The M5 pruned model tree (using smoothed linear models) generated just one rule which classified all 44 subscribers in cluster 8 as churner.

33


Table 3: Factor for Differentiations

Cluster 8

M5P Cluster 9

Cluster 10

Cluster 8

RepTree Cluster 9

0.8739

0.9463

0.9283

0.9645

0

0.7278

0.608

0.1702

0.0476

0.0811

0.0866

0.1241

0.1417

0.1478

0.2632

0.2351

0.0613

0.1261

0.1288

0.1897

0.2171

0.386

88.84%

83.10%

77.30%

38.33%

35.57%

39.33%

99.96%

62.14%

67.13%

84.85%

83.28%

48.61%

32.32%

39.91%

26.64%

100.06%

68.69%

79.81%

Cluster 8

Decision Stump Cluster 9 Cluster 10

Perfo rman ce Meas ure CC

0.5292

0.5536

MAE

0.1103

0.1896

RMS E RAE

0.1609

RRSE

Cluster 10

This work has been able to identify the subscribers that are likely to churn in the nearest future in one of the Nigerian Telecommunications Service providers. Specifically, the churn probability of subscribers in clusters 8, 9 and 10 is very high and hence serious retention campaign should commence otherwise those subscribers will be lost to other telecommunications service provider. The work was further able to identify the specific churners from clusters 8, 9 and 10. In-order to improve the interpretation of the results demographic data could also be added in further research works. REFERENCES [1]

[2] [3] [4] [5]

[6]

[7]

[8]

[9]

H. Kim and C. Yoon, Determinants of subscriber churn and customer loyalty in the Korean mobile telephony market. Telecommunications Policy vol. 28, 2004, pp. 751-765. R. Groth, Data mining: building competitive advantage. Santa Clara, CA: Prentice Hall, 1999. SAS Institute, Best Practice in Churn Prediction. A SAS Institute White Paper, 2000. P. Kotler and L. Keller, Marketing management. 12th ed. New Jersey: Pearson Prentice Hall, 2006. A. E. Eiben, T.J. Euverman, E. Kowalczyk and F. Slisser, Modeling customer retention with statistical techniques, rough data models, and genetics programming. Fuzzy sets, rough sets and decision making processes. Eds. A. Skowron & S. K. Pal. Berlin: Springer, 1998. C. Berne, J. M. Mugica and M. J. Yague, The effect of variety seeking on customer retention in services. Journal of Retailing and Consumer Services, vol. 8, 2001, pp. 335-345. J. Burez and D. Van Den Poel, Separating financial from commercial customer churn: a modeling step towards resolving the conflict between the sales and credit department. Expert Systems with Applications. Retrieved Nov. 15th, 2010, from http://www.feb.ugent.be/nl/Ondz/wp/Papers/wp_11_ 717.pdf, 2008. A. Eshghi, D. Haughton and H. Topi, Determinants of customer loyalty in the wireless telecommunications industry. Telecommunications Policy, vol. 31, 2007, pp. 93-106, S. Lee, and K. Siau, A review of data mining techniques. Industrial Management and Data Systems, vol. 101(1), 2001, pp. 41-46.

[10] [11] [12]

[13]

34

A. Berson, S. Smith, and K. Thearling, Customer retention. Building data mining applications for CRM. New York: McGraw-Hill, Chapter 12, 2000. D. Hand, H. Mannila, and P. Smyth, Principles of data mining, Cambridge: MIT Press, MA, 2001. C. Wei and I. Chiu, Turning telecommunication call details to churn prediction: a data mining approach. Expert Systems with Applications vol. 23, 2002, pp. 103-112. T. J. Ali, Predicting customer churn in telecommunication service providers. Retrieved Nov. 20th , 2010, from http://LTU-PB-EX-09052se.pdf (application/pdf object), 2009.



Lecture Attendance System Using Radio Frequency Identification and Facial Recognition Olaniyi, O.M. Department of Electronic and Electrical Engineering [email protected]

Adewumi D.O. & Sanda O.W. [email protected] * Department of Computer Science and Technology Bells University of Technology, Ota , Ogun-state, Nigeria.

Shoewu .O. Department of Electronic and Computer Engineering Lagos State University, Epe, Nigeria. [email protected],

Reference Format:

Olaniyi, O.M, Adewumi D.O, Shoewu O. & Sanda O.W (2012). Lecture Attendance System Using Radio Frequency Identification and Facial Recognition. Computing, Information Systems & Development Informatics Journal. Vol 3, No.2. pp 35-42

35


Lecture Attendance System Using Radio Frequency Identification and Facial Recognition Olaniyi, O.M, Adewumi D.O, Shoewu O. & Sanda O.W

ABSTRACT We propose a nexus of wireless biometric solution to the problem of lecture attendance records in an academic environment. The conventional method of taking attendance records on paper particularly in an environment with lower student/lecturer ratio is not only laborious but robs on the precious time that could be used for an effective learning. We demonstrated the efficacy of our proposed method against conventional methods as being capable of eliminating time wastage. . Keywords: RFID, Facial Recognition, Lecture, Attendance, Tags, Short range reader.

1. INTRODUCTION The monitoring of attendance in conventional learning environment consists of a number of requirements. The availability of both the learner and the learned usually for a period of not less than seventy percent of entire lecture period and proper record keeping of the learner during the lecture period by the tutor. In most developing countries, lecture attendance is usually noted using paper sheets, file system, surprise quizzes, and roll call of names and/or student identification number etc. These methods have made it so inadequate for the academic department to regularly update and effectively assess the true record of students in a learning environment [16,14]. The current lecture attendance monitoring system in academic environment in developing countries embraces the use of paper based method for taking and usually for computing student’s percentage of attendance [14]. This method of attendance monitoring is time consuming and laborious because the valuable lecture time that could otherwise have been used for lectures is dedicated to student attendance taking. This inadequacy in the process of attendance monitoring leads to wrong compilation of student’s that were in the class for the entire duration of the course. Biometric systems have been widely used for the purpose of automatic recognition of objects based on some specific physiological and behavioural features [10].Many biometric systems can be applied for a specific system but the key structure of a biometric system is always the same. In biometric facial recognition, record of the spatial geometry of distinguishing feature of the face is recorded. Because a person’s face can be captured from some distance away, the technology has been used to identify card counters or other undesirables in shoplifting and monitoring of criminals and terrorists in some countries with the history of terrorism. Biometric Face recognition is one of the few biometric methods with the merit of both high accuracy and low intrusiveness. It has the accuracy of a physiological approach without being intrusive. The technology has drawn the attention of researchers in fields from security, psychology, and image processing up to computer vision [6][7].

Accordingly, there have been proliferations of Radio Frequency Identification (RFID) systems in a number of applications. Successes have been recorded in diverse areas as Healthcare Monitoring [17], Library [15], Home and Business Security Systems [4] and Construction [9] to name a few in Literatures. Radio Frequency Identification (RFID) systems facilitate automatic and identification and tracking of remote components. Research in this field involves improving tags, readers and adapting tags to multiple substrates and function under extreme conditions of temperature, humidity and application of the latest technology to achieve various objectives such as improving traceability, efficiencies, and real-time monitoring system behavior especially in critical health care condition [11][1]. This work seeks to combine value added advantages attributed to these two electronic identity systems: RFID and Facial Recognition in exploring a cutting edge wireless biometric solution to the students’ academic attendance monitoring problem in developing countries. 2.0 REVIEW OF RELATED WORKS A number of related works exists in literature in the application of RFID and Facial Recognition to different areas of attendance monitoring problem. In [12] authors proposed student tracking using RFID. It involves the use of the student identification card to obtain student attendance. The author tried to solve the problem of manual computation of attendance but his work does not eliminate the risk of student impersonation. Consequently, authors in [1] proposed an RFID matrix card based auto identity system to the manual problem of monitoring student in boarding schools. Upon initial study of the three Boarding school in Malaysia, current process of maintaining students records in and out was not only tedious, misinformation always happen as students tend to provide inaccurate information. The fusion of passive RFID Tags, Wireless local area networking and database management system development helps to ease the monitoring of the availability of boarding students as system RFID reader monitors and recorded student identity through their unique and pre-assigned RFID tag.

36


Also, authors in [8] reviewed the use of RFID in an integrated-circuit(IC) packaging house to resolve inventory transaction issues. This study suggests that RFID contributes significant improvements to the water receiving process and the inventory transaction process that reduce labour cost and man-made errors. In [5] author proposed the use of finger print to solve attendance monitoring problem. The fingerprint technique verification was achieved using extraction of the fingerprint of students. The proposed system was successful in monitoring attendance but the proposal of [5] lacks the inclusion of a report generation and audit trail system. Similar attendance monitoring solution was developed in [3] to manage the context of the student for the classroom lecture attendance using the Personal Computer of each student. Authors in [11] proposed design and prototype implementation of a secure and portable embedded reader system for reading biometric data from an Electronic passport(E-passport) using Electronic Product Code (EPC) RFID tags. The passport holder is authenticated online by using GSM network. Secure communication through Advance Encryption Standard (AES) encryption technique between server and the proposed e-passport reader helps to provide comprehensive system to create, manage and monitor identity data online.

3.0 MATERIALS AND METHOD 3.1 System overview The system was developed for Lecture Attendance Management Scenario of Bells University of Technology, Ota, Nigeria for each lecture period. The system manages the student lecture attendance using a Windows Application system and the developed RFID and Face Recognition based attendance model. The application system contains a module known as the administrator module. The function of the administrator module is to handle the entire administrator task: Adding, editing and deleting classes, subject and college/department. Only the administrator can view, add and delete data in the attendance system. Figure 1.0 shows the general block diagram of the system. The developed model consists of an RFID Reader incorporated with a µRFID Reader board, RS232 to USB converter cable

TAGGED STUDENT

In [14], authors proposed a simplified and cost effective model of embedded computer based solution to the manual method of managing student lecture attendance problem in higher institutions in developing countries. The developed system is capable speeding up the process of taking students lecture attendance and allows for error free and faster verification process of authenticating student lecture attendance policy required for writing examination in a campus environment but could not provide absolute solution to the problem of impersonation by erring students. In [2] Artificial Neural Networks and Facial Recognition were used to develop a security door system where authorization of facial appearance of privilege users in the database is the only guarantee for entrance. In the system, the personal computer processes the face recognized by the system digital camera and compares data with privileged users in the database. The system control program either sends a signal to open the electromechanical door upon facial existence or deny entry. In this paper, we proposed a nexus wireless biometric solution to the problem of lecture attendance problem in academic environment. The current process of taking student particularly in an environment with lower student/lecturer ratio is not only laborious but robs of the precious time that could be used for an effective learning. The amalgamation of these technologies to student attendance monitoring problem as demonstrated in this study is capable of eliminating time wasted during classical/manual collection of attendance, provide solution to the problem of impersonation liable to similar solution as proposed in [1, 14,5,12] and an avenue for proper academic monitoring of students performance by University administrators.

Facial Recognition

Radio

READER

PROGRAMMED PC

USB to RS232 Communication

IP-BASED CAMERA USB to USB

Fig 1: System Block diagram

3.2 Design Considerations The proposed attendance management system in this work consists of the following considerations: Hardware Design Considerations Considering RFID systems shown in Figure 2.0, electronic tags communicate with the reader through radio waves. RFID Tags can be one of three types: active, semi active or passive. Because these tags do not supply their own power, communication with them needs to be short and usually does not transmit much data usually just an ID code. The range for transmission is from 10mm to 5 meters. There are four different kinds of tags in use, categorized by their radio frequency: low frequency (between 125 to 134 KHz), high frequency (13.56MHz), UHF (868 to 956 MHz), and microwave (2.45 GHz).The tag has a unique set of numbers which makes every card unique, in each case a reader must scan the tag for the data it contains and then the information is sent to the database.

37


Fig 2: Basic RFID System In the foregoing we shall described the hardware consideration: The Electronic Tag: The study exploited popular wide range of EM4100 transponder available for the Micro RFID Reader (µRW) as the low frequency Electronic RFID tag. The tag electronic is mapped with student information (Name, Matriculation number, level and Department) available in the system database. For the lecture attendance management scenario of Bells University of Technology, Ota, Nigeria considered in this study, the RFID tag for four students and untagged card is shown in figure 3:

Fig 3.0: Electronic Tag RFID READER –µRW For this study, the µRW RFID Reader was chosen for cost reason. It was designed to read from EM4100 transponder used as electronic access card at frequency of 134 kHz. In operation the reader continually scan for EM4100 transponder pre-defined at 134 kHz to respond to C# program commands via the UART Receive line (Rx) serially connected through the RS232 to USB converter to the USB port of the PC. The overall circuit of the RFID subsystem is shown in Figure 4: VCC 5V

U3 LM2931AZ-5

POWER IN

LINE VOLTAGE

VREG

COMMON

LED1

R1

C1 10µF

1kΩ

U1

URFID RS 232 CIRCUIT U2

CTS

R2

RTS

PC

1kΩ

TXD

BC337

RXD

Fig 4:. Overall Circuit diagram of the RFID system

38

BUZZER 135 Hz


POWER SUPPLY Most TTL (Transformer-Transfer Logic) digital circuit uses 5V to operate. A 5V source needs to be regulated to power for the µRW RFID Reader circuit. Through 9V to 24V DC unregulated supply, this part was developed by using LM2931AZ5 as the voltage regulator. Facial Recognition/Comparison Due to cost reasons, this stage face capture and comparison session was accomplished through a simple web camera. Once a reader badges his card in for attendance the web camera automatically takes a picture of the person holding the tag and compares it with the enrolled in the system database during initial registration of the student.

Fig 5: The RFID Student Attendance Monitoring Hardware Prototype

Software Design Consideration In the development cycle of the proposed RFID system, decisions are made on the part of the system to be realized in the hardware design and the parts to be implemented in the software. This software module consists of modules that can be easily decomposed and tested as individual units; this was done to make sure software meets design considerations. The attendance monitoring program was written in Microsoft visual C# programming language in a Visual studio development environment. Figure 6.0 shows the overall flowchart of the system for both RFID and Facial Recognition sub systems.

Y

Facial Compariso n

Valid face?

Take Attendance

N Tag

µRFID Reader

Valid Tag ?

Y

N Fig 6: Overall Flowchart of the student RFID and Facial Recognition based Attendance System

39


This corresponding ASCII format code is then decoded by the programmed PC through the RS232 to USB converter shown in figure 5.0 Since two-level authentication and verification is required for acknowledgement of student attendance for each lecture, equal facial comparison of the student at the entrance with pre- enrolled facial appearance of the student stored in the database by the intelligent IP camera justify the biometric verification of the student and thus acknowledged the student attendance for the lecture automatically by the Programmed PC. This mutual exclusiveness of the wireless radio waves monitoring between the EM4100 RFID tag and µRFID reader and facial comparison of the real time student facial appearance with facial appearance in the database is shown in Figure 7.0.

4. SYSTEM OPERATION/TESTING & DISCUSSION Considering figure 6.0 every student with a pre-programmed EM4100 transponder RFID tag has a privilege to attend lecture through the entrance door, a serial number of tag is associated with the student database entry on the Programmed PC. Each time a student flips his/her card/RFID tag, the µRFID reader responds wirelessly through the pre-defined commands via the UART Receive line of the URFID. The availability of EM4100 transponder RFID tag selected in range of 135kHz makes its serial number to be read, set the LED color from red to green, buzzer to function and associated data transmitted on the UART Tx line in serial ASCII format.

Fig 7: Illustration of the RFID and Facial Recognition Operational Principle The buzzer is activated when a valid RFID tag passes through a radio frequency of the µRFID reader. If the tag and the captured face is similar to the captured face in the system database, then the system register the student as present in the class. Due to cost and flexibility reasons, this RFID attendance system uses passive tags and thus for every class, students needs to swipe their tags close to the reader (about 15mm from the reader). The reader reads the tag and the application reads check-in time and when the student is leaving the same process is repeated and the application reads check-out time. Also the facial recognition is accomplished with a web camera. If an invalid EM4100 RFID tag is used, the program will give a notification that the tag has not been registered to any student and requires a valid tag. The database contains the name of student, Matric number, Address, E-mail, Course duration and Course Information. Figure 8-figure 12 shows the Graphic User Interfaces (GUI’s) of the system application control program developed with Visual C# programming Object Oriented Programming Language:

Fig 8 : Home page

Fig 9 : Course Registration Interface

Fig 7: Student Information Enrollment Interface

40


Fig 12. Attendance Facial Comparison

Fig 10: Attendance Monitoring Interface

Fig 11: Attendance Check-in and Check-out Fig 13: Attendance Report Page (2) Application of an active reader for effective RFID performance.

5.0 CONCLUSION This paper has successfully presented a simplified, low cost wireless biometric solution to the problem of lecture attendance records in an academic environment in developing countries. The prototype implementation of RFID and Facial recognition in attendance taking and the objectives stated on previous section has been achieved. The major strength of the system lies in its portability and high scalability but with less flexibility in programming as compared to the previous design and implementation in [1, 14, 5, 12]. By careful examination, it can be inferred that the proposed system could not only speed up the process of taking attendance, it also solves the problem of impersonation which was encountered in previous solutions.

(3) Browser testing and extended wireless testing must be conducted for possible deployment situations. REFERENCES [1]

[2]

6. FUTURE WORK The developed system is not without exceptions. Hence the following recommendations could be made for improvement in the immediate future: (1) Incorporation of Iris and IP camera for secured Facial Recognition that would further increase the efficiency and security of the system against impersonation in distributed Network of different real time lecture room monitoring respectively.

41

[3]

Abdul Kadir H, Abdul Wahab M and Siti Nurul A(2009),”Boarding Students Monitoring Systems(E-ID) Using Radio Frequency Identification”, Journal of Social Sciences, Volume 5(3),pp 206-211. Arulogun O.T, Omidiora, O., Olaniyi O.M, and Ipadeola A.A. (2008), “Development of Security System Using Facial Recognition”, Pacific Journal of Science and Technology. 9(2):377386. Cheng K, Xiang L Hirota T and Ushijima , K(2005),”Effective Teaching for large classes with Rental PCs by Web System ”,Proceedings of Data Engineering Workshop(DEWS 2005). ID-d3, Japan.


[4]

Fujikawa M, H. Doi and Tsuijii (2006),”Proposal for a new Home Security System in terms of Friendliness and Prompt Notification”, Proceeding of the IEEE International Conference on Computational Intelligence for Homeland Security and Personal Safety, October 1617,IEEE Explore Press, Alexandria VA,pp 6166.DOI:10.1109/CIHSPS.2006. 313308. [5] Kokumo B (2010),”Lecture Attendance System Using Biometric Fingerprint”, B.Tech Dissertation, Department of Computer Science and Technology, Bells University of Technology, Ota, Nigeria. [6] Kelly, M.D., (1970),“Visual Identification of People by Computer”, Stanford AI Project, Stanford, CA, Technical Report. AI-130. [7] Lin S (2000),”An Introduction to Face Recognition Technology”, Informing Science Journal: Special Issue on Multimedia Informing Technology, Volume 3 No1, pp1-7. [8] Liu C.M and Chen L.S (2009),”Application of RFID Technology for Improving production efficiency in an Integrated-Circuit Packaging House", International Journal of Production Research, Vol. 47, No. 8, pp. 2203-2216. [9] Lu, M., Chen. W. X. Shen, H.C. Lam and J. Liu,(2007),”Positioning and Tracking Construction Vehicles in Highly dense Urban Areas and Building construction Sites. Automatic Construct., 16: 47-656. DOI: 10.1016/j.autcon.2006.11.001 [10] Maltoni D., Maio D, Jain, A. K.and Prabhaker S(2003),“Handbook of Fingerprint Recognition”, Springer, New York, Page 3-20. [11] Mohammed A.B, Ayman A, Karrem M (2009),” Implementation of an Improved Secure System Detection for E-passport by using EPC RFID Tags”, World Academy of Science, Engineering and Technology (WASET) Journal, volume 60, pp114-118. [12] Mahyidin M. (2008), Student Attendance Using RFID System , B.Eng Thesis, Electrical and Electronics Engineering Department, University of Malaysia Pahang, Retrieved online at http://umpir.ump.edu.my/345/1/3275Firdaus.pdf on 21st September,2011 [13] Pala Z and Inanc N (2007),”Smart Parking applications Using RFID Technology”, Proceeeding of the first Annual Eurasia September 5-6,IEEE Explore Express, Istanbul pp1-.DOI:10.1109/RFIDERASIA.2007.4368108

[14]

[15]

[16]

[17]

42

Shoewu O, Olaniyi O.M. and Lawson, A (2011).” Embedded Computer-Based Lecture Attendance Management System”, African Journal of Computing and ICT. Vol 4, No. 3, Pp 27- 36. Singh, J., N. Brar and C. Fong,(2006),” The state of RFID applications in libraries”, Inform. Technol. Libraries, 25: 24-32,Retrived online at: http://cat.inist.fr/?aModele=afficheN&cpsidt=17 860855 on 21st September 2011. Sanda O.W(2011),”Development of an Automated Lecture Attendance Management System Using RFID and Facial Recognition”, B.Tech Dissertation, Department of Information Technology, Bells University of Technology, Ota,Ogun State, West Africa. Wang, S.W., W.H. Chen, C.H. Ong, L. Liu and Y.W. Chuang (2006),” RFID Application in Hospitals: A case study on a demonstration RFID project in a Taiwan hospital”, Proceeding of the 39th Annual Hawaii International Conference on System Sciences, Jan. 04-07, IEEE Xplore Press, USA., pp: 184-184.



Employee’s Conformity to Information Security Policies- The Case of a Nigerian Business Organizations ADEDARA, Olusola Department of Computer Science The Federal Polytechnic Ado-Ekiti Nigeria [email protected] KARATU, Musa Tanimu Computer Department Faculty of Science University of Ibadan Nigeria OLAGUNJU, Abiodun Department of Computer Science University of Ibadan Ibadan, Nigeria [email protected]

Reference Format:

Adedara, O., Karatu, M.T. & Lagunju, A. (2012). Employee’s Conformity to Information Security Policies In Nigerian Business Organisations (The Case of Data Engineering Services PLC). Computing,, Information Systems & Development Informatics Journal. Vol 3, No.2. pp 43-50

43


Employee’s Conformity to Information Security Policies – The Case of a Nigerian Business Organizations Adedara, O.; Karatu, M.T. & Lagunju, A.

ABSTRACT We evaluated employee’s conformity to information security policies using four research questions and three research hypotheses. A survey research methodology was used to evaluate staff of the case study (Skannet) using both questionnaires and interview. Data gathered were analysed using both descriptive and inferential statistics. The findings of this study reveal that majority of Skannet’s employees have adequate knowledge of Skannet’s information security policies. However, the level of employees’ compliance with the ICT policies is very low. Among organizational measures put in place by Skannet to ensure information security compliance include regular training / re-training for staffs on policies, regular survey sent to all staff to test level of awareness and punitive measures by Human Resource Department. The study further revealed that Skannet’s Information Security policy positively affect the attainment of organizational goals such as improvement in quality of services, promotion of information sharing, transparency and accountability among staffs in the organization . Keywords: Policies, security, compliance, training and services. 1. INTRODUCTION The Internet is a collection of networks linked together using a common protocol – global computer network achieved through the interconnection of smaller computer networks around the world. People, computers and information are link together electronically by a common protocol or set of communication rules. It should be evident now why telecommunications, broadcastings and the internet all have to be dealt with not necessarily by a single policy but within a single policy frame work. They all use the same infrastructure to transmit messages (copper cable, optical fiber, satellites) over the radio spectrum and they can all deliver the same content (voice, data text, pictures, videos etc) to the same users. (Kate Wild 2004). Information Security policies generally covers three areas, they are: 1. Telecommunication 2. Broadcasting 3. Internet (Networking Technologies) A corporate policy is usually a documented set of broad guidelines formulated after an analysis of all internal and external factors that affect the firms objectives, operations and plan, formulated by the firm’s board of directors. (Folayan, 2010, Boubakar, 2008). Information security on the other hand is the process of protecting information. It protects its availability, privacy and integrity. Access to stored information on computer databases has increased greatly. More companies store business and individual information on computer than ever before. Much of the information stored is highly confidential and not for public viewing. (Toni & Tsubuira, 2002). Effective information security systems incorporate a range of policies, security products, technologies and procedures. Software applications which provide firewall information security and virus scanners are not enough on their own to protect information. A set of procedures and systems needs to be applied to effectively deter access to information. (Steve, 2008; ISP US, 2004)

Information security means protecting information and information systems from unauthorized access, use, disclosure, disruption, modification, perusal, inspection, recording or destruction.The terms information security, computer security and information assurance are frequently incorrectly used interchangeably. These fields are interrelated often and share the common goals of protecting the confidentiality, integrity and availability of information; however, there are some subtle differences between them. (Wikipedia 2011). 1.1 Knowledge Gaps Information and communication technology is increasingly penetrating all social and economic activities. It is a high stake game that involves all sectors of society comprising many stakeholders. Information Security policies are often made as a result for concern for issues bothering on system vulnerabilities occasioned by authorized usage. Information Security policy is a set of principles or a broad course of action that guides the behavior of governments, organizations, corporations and individuals (Steve, 2008). Information Security policy covers information and communication technologies network, services, markets and relationships between the different actors involved in these – from the operators of submarine cable systems to the users of telecentres (Microsoft, 2010; OECD, 2012). It may be national or international in scope. Each level may have its own decision making bodies, sometimes taking different or even contradictory decisions about how Information Security will develop within the same territory (David, 2009). In order to give staff members the feelings of autonomy and a sense of belonging, they need to know the rules and allowable usage limits of organization al information systems. These go beyond what time to show up, vacation time and health benefits. Written company policy that covers Information Security policy must be produced and adhered to.

44


This paper identifies and evaluate the manner in which Information Technology policies are defined and the level of usage compliance in a corporate setting using General Data Engineering Services (Skannet) Nigeria as a case study. The paper attempts to ascertain Skannet’s Information Security policies awareness by the employees; examine the extent to which the employees comply with the Information Security polices; identify organizational measures instituted by Skannet to ensure policy compliance by employees; ascertain the impacts of Information Security policies on Skannet’s organizational goals and make recommendations on Skannet’s Information Security policies improvement. 2. EXISTING INFORMATION SECURITY POLICY AT SKANNET General Data Engineering Services being an Internet Service Providing Company has certain rules and procedures governing what users can do on her network (Folayan, 2010). This Information Security Policy is as highlighted below. Access Management and Control To prevent or minimize unauthorized access to computer systems or damage, theft or loss of equipment, the following must be adhered to: 1. Physical Control a. Access to server room and other major ICT facilities should be adequately secured at the doors and windows and only authorized persons can be allowed into the Server room. b. The Contact Centre is required to maintain a Register (access log) where authorized staff logs in any activities carried out in the server room. Logical Control a. Each user (staff and client) on the network must have a Username and Password to access networked facilities. b. Both staff and client username must not exceed eight characters. The password can be as long as desired by the users. c. Each staff upon appointment must fill a staff account creation form. The form shall be administered by the human resource department. d. The newly appointed staff will submit the completed form which must be signed by his supervisor to the officer on duty at the contact centre. e. The duty officer is responsible for creating new staff accounts and activating the account on the network. f. The duty officer should ensure that the entries in the form is correctly filled and signed by the staff before creating such account. g. Each prospect/client will fill an account creation form at sign-up. h. The client sign-up form will be administered by the Marketing Department. i. The contact centre shall activate such account upon verification of the client identity.

Internet Usage It is unacceptable to use General Data Engineering Services networks to: a. View, make, publish or post images, text or materials that are, or might be considered as illegal, paedophilic or defamatory. b. View, make, publish or post images, text or materials that are, or might be considered as, indecent, obscene, pornographic or of a terrorist nature. c. View, make, publish or post images, text or materials that are or might be considered as, discriminatory, offensive, abusive, racist or sexist when the context is a personal attack or might be considered as harassment. d. Send Spams, unwanted and unsolicited emails. Network Control a. The setup of PCs, laptops, printers, etc for network access should be done by the Engineering Department (GDES Workshop). b. Point-to Point-over-Ethernet should be setup for the user's authentication on the network during the installation. c. The company antivirus should be installed on each staff Computer. The License shall be administered by the Engineering Department. Others include Troubleshooting, Repairs, Maintenance and Replacements; Disaster Recovery and Contingencies and Electronic Mail (Email), 3. RESEARCH METHODOLOGY In this study, research design using a descriptive method was used, the population of design and the population of study. Sample size and sampling technique was used to manage the research work. Data collection procedure such as questionnaires, personal interviews, was instruments used for collection of data. Data analysis method used were descriptive statistical method of sample frequency counts and percentages were used to analyze the demography and research questions while inferential statistics of T-test, ANOVA and Chi-square was used to analyze the hypotheses. 3.1 Research Questions Based on the foregoing, the research questions that this paper seeks to address include: a. What is skannet’s employees’ awareness level of Information Security policies? b. To what extent do Skannet’s employees comply with Information Security policies? c. What are the organizational measures put in place by Skannet to ensure Information Security policies compliance by employees? d. What are the impacts of Information Security policies on Skannet’s organizational goal?

45


3.2 Research Hypothesis: Hypothesis 1: There will be no significant difference between Skannet’s employees compliance with Information Security and attainment of organizational goals.

Data Analysis Method Descriptive statistical methods of simple frequency counts and percentages were used to analyse the demography and research questions while inferential statistics of T-test, ANOVA and Chi-square was used to analyze the hypotheses.

Hypothesis 2: There will be no significant relationship between Skannet’s employees department and employees’ policy compliance

4. DATA PRESENTATION AND ANALYSIS

Hypothesis 3: There will be no significant difference between Skannet’s employees Information Security policies awareness and compliance.

This section is divided into two; section A is on the demography and section B presents data to validate the hypothesis raised in the study. The data are first presented in tables before they are presented in essay form; inferences are made from the data.

3.3 Research Design The research design adopted for this study was survey descriptive research method. This method was adopted because it enabled the researcher to collect data on the concerns of this study from the population at the locations of the population and to describe in a systematic manner the state or condition of the objects of research in this study.

SECTION A Table 1: Distribution of Respondents By Departments Departments I.T Customer Care Sales Finance Transmission Transmission Planning Switch Site Acquisition Site Maintenance Radio Frequency HR Fleet Management Total

3.3.1 Population Of The Study The population of this study is made up of this staff and management of General Data Engineering Services in the South-western Zone and East of Nigeria; this include Oyo, Ogun, Kwara, Ekiti and Enugu State. A total number of ninety-six Skannet staff is in these states. 3.3.2 Sample Size And Sampling Technique The entire population of this study, a census, was used in this study because it was manageable by the researchers. 3.3.3 Research Instruments A self-designed questionnaire was used to collect data from this study. The instrument was divided into two sections A and B. In section A, the demography of the respondents such as age, years of experience, sex, marital status, educational background and religion was solicited. In section B, questions meant to solicit information to answer all research questions and hypothesis were asked. The instrument contained both open-ended and close end questions. Also, an interview guide was used to interview six Heads of Departments at Skannet. 3.3.4 Validation Of Instrument In order to ensure both face and content validity of the instrument, the questionnaire was submitted to the researcher’s supervisor and two other scholars for constructive criticisms. It was after their corrections were effected that the researcher went to the field to administer questionnaire. 3.3.5 Data Collection Procedure The researcher distributed copies of the questionnaire among the staff of Skannet in the South-Western Zone division of the organization. Along with two other research assistants, the researcher made sure that the respondent were given adequate time to fill the questionnaire and to ensure high return rate, the instrument was handed over to solicit the assistance of the authority figures in the offices.

Frequency 13 7 8 3 3 8 3 8 6 7 13 6 88

Percentage (%) 14.8 8 9.1 3.4 3.4 9.1 3.4 9.1 6.8 8 14.8 6.8 100

The above table shows various departments at General Data Engineering Services. The organisation has twelve (12) departments. It is evident from the data that 13 (14.8%) respondents are from Information Technology department, 7 (8%) respondents are from Customer Care, 8 (9.1%) are from Sales department, 3 (3.4%) respondents are from Finance departments, 6 (6.8%) respondents are from Transmission, 8 (9.1%) respondents are from Transmission Planning department, 3 (3.4%) respondents are from Switch department, 8 (9.1%) respondents are from Site Acquisition department, 6 (6.8%) respondents are fro Site Maintenance department, 7 (8%) respondents are from Radio Frequency Department, 13 (14.8%) respondent are from HR department while 6 (6.8%) respondents are from Fleet Management. The data implies that the respondents cut across all departments in the organization; they will make their contributions in this study to be perspectives. SECTION B Research Question 1: What are the employees’ awareness levels of Skannet’s Information Security policies? Organizational employees are to know about policies before such could be obeyed. This research question is to ascertain employee’s awareness of Information Security policies.

46


Research Question 2: To what extent do Skannet employees comply with Information Security policies?

Table 2: Skannet’s Employee Awareness of Information Security policies Question Is there any policy on Information Security at Skannet?

Options

Frequency

(%)

Yes

71

80.7

Question

19.3 100

Do you comply with policies on Information Security

No Total

17 88

Table 2 shows the data on Skannet’s employees Information Security awareness. It is obvious from the data that majority of the respondents are 71 (80.7%) affirm that there are policies on Information Security at Skannet while 17 (19.3%) respondents decline. These data is an indication that many of the employees are aware or rather they know about the policies guiding the use of Information Security at Skannet. It could therefore be inferred that Skannet Management does intimate the employee with the policies. However, these data do not indicate the degree of the employees’ awareness. Therefore the next question is asked and answered. Table 3: Degree of Information Security Awareness By Skannet Employees Question

Options

Freq

(%)

How much of the Information Security Policy do you know?

Much

54

61.4

Very Much Not Much Not Very Much Total

21 3

23.9 3.4

10

11.4

88

100

In Table 3 data on the degree of Skannet’s employees Information Security awareness is revealed. It is evident in the data that 54 (61.4%) respondents know “much” of the policy, 21 (23.9%) respondents know “very much” of the policies, 3(3.4%) respondents know “not much” of the policies while 10 (11.4%) respondents know “not very much” of the policies. The employees generally speaking are expected to know details of the Information Security policies because of its importance to daily operation or organisational conduct of the workers. Meanwhile as it could be logically expected, the data reveals that vast majority of the workers are familiar with the details of Information Security policies at Skannet. Meanwhile, interviews conducted for six departmental heads reveals that Information Security policy orientation which acquaints employees with specific details of how Information Security policies are to be applied is a usual orientation course for all staffs that are deployed to the departments. According to some of these interviews, because Skannet’s operations are Information Security based, it is of utmost necessity for the staff to be oriented on how to and how not to use the Information Security

Table 4. Employees Compliance with Information Security Options

Freq

(%)

Yes

61

69.3

No Total

27 88

30.7 100

Table 4. reveals that majority of Skannet’s staff do comply with Information Security policies of Skannet. In the table, 61 (69.3%) respondents affirm that they do comply while 27 (30.7%) respondents claim that they do not. The data indicates that worker’s compliance with the Information Security policies at Skannet is not doubtful. This implies that the staffs are not only aware of the policies but they do also comply. Meanwhile five out of seven head of units interviewed submit that the level of workers compliance with the Information Security policies is low. For example, some of them affirm that the staff secretly and sometimes openly flout the policies when they know that they are not being monitored or supervised. At a point, one of them points out that staffs are not mindful of the policies at all until a reorientation programme was organized in 2010. Table 5: Extent to which Skannet’s Employees Comply with Information Security Policies Question

Options

To what extent do you comply with Information Security policies?

Great Extent Some Extent Little Extent No Extent Total

Freq

(%)

10

11.4

21

23.9

51

57.9

6 88

6.8 100

Table 5 shows that 31 (35.3%) respondents do comply with the Information Security policies to a large extent while 57 (64.7%) respondent do comply to little/no extent. The above finding is an indication that most staffs at Skannet do comply with the Information Security policy to a small extent. Therefore, there appears to be a gap between the knowledge of the policies and the compliance with the same. In essence, there is apparent disregard for the policy. These findings substantiate the submission unit heads earlier presented which indicated that Skannet staffs give little regards to Information Security policy implementation.

47


Research Question 3: What are the organizational measures put in place by Skannet to ensure policy compliance by employees? It is an organizational practice to put measures that will ensure compliance to corporate policy in place. Therefore, this question is to identify the corporate measures put in place by Skannet to ensure Information Security policy compliance. Table 6: Organizational Measures Put in Place by Skannet to Ensure Information Security Policy Compliance Measures Frequency (%) Regular Communication from management to staff on Information 82 93.2 Security policies Regular training/re-training of staffs 61 69.3 on policies Regular surveys sent to all staffs to 68 77.3 test level of awareness Compulsory tests taken on 80.7 Information Security by staffs 71 Punitive measures by Human 68 77.3 Resources Department Table 6 shows data on organizational measures put in place to ensure that staffs comply with the Information Security policies. It is evident from the table that 82 (93.2%) respondents acknowledge that regular communication from the management to the staffs on policies is to ensure that policies are adhered to, 61 (69.3%) respondents affirm that a regular training and retraining is another measure, 68 (77.3%) respondents affirm that a regular survey is usually sent out to test level of staffs awareness to Information Security policies; also, 71 (80.7%) respondents affirm that compulsory tests are taken by staffs on Information Security policies; the last item on the table reveals that Punitive measures are used by Human Resources Department to ensure compliance with Information Security policies. Research Question 4: Table 7: What are the impacts on Information Security policy on the attainment of Skannet’s organizational goal? Question Do you think Information Security policy of Skannet positively affect attainment of organisational goals

Options

Frequency

(%)

Yes

69

78.4

No Total

19 88

21.6 100

In Table 7, majority of the respondents 69 (78.4%) respondents affirm that Information Security policy positively affect the attainment of Skannet’s organizational goal. Human Relation head of Skannet, when interviewed affirmed in consonance with the findings of the staff respondents.

According to him, policies are put in place because both the organization and its workforce will be beneficiaries; therefore, it is self-evident that the policies are there to take care of the work effectiveness in the organization. Table 8: Impacts of Information Security policy on Skannet’s Organizational Goals Impacts Frequency (%) It improves quality of service and products in the organization 59 67

It promotes information sharing, transparency and accountability among staffs in the organization

56

63.6

It provides information and communication facilities and services at reasonable costs

62

70.5

It provides individuals and organization with adequate Information Security knowledge

71

80.7

In table 8, 59(67%) respondents affirm that Information Security policy of Skannet improves quality of services and products in the organization, 56(63.3%) respondents affirm that it promotes information sharing, transparency and accountability among staffs in the organization; it also provides information and communication facilities and services at reasonable costs and lastly, it provides individuals and organizations with adequate Information Security knowledge. The findings in the above data reveal that the policies on Information Security in Skannet have impacts on attainment of to ensure compliance with Information Security policy (Ibrahim, 2010; Davis, 2008. The findings in the above data reveal that there are organizational measures in place at Skannet’s to ensure implementation of Information Security policies. Among measures put in place are regular Information Security policies and preventive measures. Customer care unit head, sales unit head and Human relations, rear to corroborate the findings of the questionnaire; According to the trio, the attention of Skannet is focused on awareness of Information Security policy. According to them when revile round and are frequently reminded they would do, this is the organizational belief of Skannet. However, as it has been earlier established in this study that through there is awareness of the policy but the policy adherence by staff is low, according to one of the interviews the reason behind this is because Skannet as an organization care less about adherence because, according to him, he has never seen anybody punished for flouting the policy.

48


Hypothesis 1: There will be no significant difference between Skannet’s employee’s compliance with Information Security and attainment of organizational goals.

ANOVA test is used to test the above hypothesis. The data is represented below:

Table 9: ANOVA test on Skannet’s employee’s Information Security policies Attainment of Organizational Goal. F

t

Df

Mean Difference

Std. Error Difference

Sig (2-tailed)

Decision

627.291

10.067 -6.648

86

-6296 -6296

0.6254 0.09471

0.000

*Sig

The above table reveals that the test result .000 is less than 0.05 level of significance (P.000