Available online at www.sciencedirect.com
ScienceDirect Procedia Computer Science 55 (2015) 802 – 808
Information Technology and Quantitative Management (ITQM 2015)
Establishment of China information technology outsourcing early warning index based on SVR Siqi Yia,b,Yong Shia,b,c*, Yibing Chena,b b
a School of Management, Chinese Academy of Science, Beijing, 100190, China Research Center on Fictitious Economy & Data Science, Chinese Academy of Sciences, Beijing 100190, China c University of Nebraska, College of Information Science and Technology, Omaha, NE 68182 USA.
Abstract Information technology outsourcing in China has developed fast, it plays a more and more important role in economic development of China. Economic analysis and early warning system of information technology outsourcing, which reflect the status of ITO, can promote the healthy development of the industry. This paper constructed the indicator system by the method of time difference relevance and peak-valley. The weight vector of each indicator is attained by using support vector regression. It also calculated the comprehensive early warning index and established the early warning index system. At last, we used a group of signal lamps to reflect the status at every time. Based on the reality of ITO in China, this paper found that the development speed of ITO is slowing in recent months, the government should take out some positive measures. © 2015 by Elsevier B.V. This is an openB.V. access article under the CC BY-NC-ND license 2015Published The Authors. Published by Elsevier (http://creativecommons.org/licenses/by-nc-nd/4.0/). Selection and/or peer-review under responsibility of the organizers of ITQM 2015 Peer-review under responsibility of the Organizing Committee of ITQM 2015
Keywords: information technology outsourcing, early warning, leading index, coincidence index, SVR
1. Introduction Outsourcing is that the enterprises or organizations transfer their own work to outside suppliers [1]. Information technology outsourcing (ITO) is one of outsourcing types. The Gartner Group holds that information technology outsourcing includes the combination of product support and professional services, providing IT infrastructure and enterprise application services to customs. It also found that the total contract value of publicly announced IT outsourcing deals exceeded $150 billion from 2003 through 2006. In 2008 alone the top 20 worldwide ITO contracts were worth nearly US $20 billion [2, 3]. With the fast growth of information technology and economy in China, the IT outsourcing is acting a more and more important role in our nation economy system.
* Corresponding author. Tel.: 13501093251 E-mail address:
[email protected].
1877-0509 © 2015 Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Peer-review under responsibility of the Organizing Committee of ITQM 2015 doi:10.1016/j.procs.2015.07.156
803
Siqi Yi et al. / Procedia Computer Science 55 (2015) 802 – 808
According to the ministry of commerce, the total ITO contracts value is increased by 28.7% to US $27.08 billion in the first half year of 2014. Economic analysis and early warning is an effective method for supervising the whole economic development [4, 5]. It quantifies a series of indicators, which can reflect the activity process and status of information technology. Niemira uses three main economic indicators to construct the chemical industry leading and consistency index [6]. Ronny Nilsson surveys the OECD's leading indicator system, using leading indicators and other tools to prepare for economic forecasting [7]. Coincidence and leading index are constructed for the economy of euro area by Ataman Ozyildirim [8]. Eric Mayott establish the early warning system of real estate Industry in Hamline [9]. This study is focused on constructing ITO early warning system, we first use economic analysis to select leading and coincidence index, than determine the weight of each index by applying SVR method. Finally, we utilize mathematics method to format the comprehensive index. 2. Support Vector Regression Support vector machine (SVM) is a new method of machine learning based on statistical learning theory. As for structural risk minimization principle, SVM has a better solution to small sample [10]. By introducing a loss function, the SVR problems also converted into solving the dual problem. First we should define a loss function as following: Lε ሺfሺxሻ,yሻ= ቊ
ȁfሺxሻ-yȁ-ε if ȁfሺxሻ-yȁ≥ε 0
otherwise
(1)
Where ɂ is a positive number, which reflect the distance between ሺሻ and the true value y . The objective of SVR is to find a function f(x) that matches all input data with an error at the most ε and at the same time as be flat as possible[11]. The case of linear function f(x) has been described in formulation (2). ሺሻ ൌ ɘ ή ǡ א୬ ǡ א
(2)
Let the inputs to be୧ ,୧ א୬ response variables ୧ אൌ , i = 1...n. The training samples are T = {ሺଵ ǡ ଵ ሻǡ ڮሺ୪ ǡ ୪ ሻ}. By minimizing the following objective function, we can get the values of parameters w and b. ଵ
ԡԡଶ σ୪୧ୀଵሺɌ୧ Ɍכ୧ ሻሺ͵ሻ ଶ
Ǥ Ǥ୧ െ ሺɘ ή ୧ ሻ െ ɂ Ɍ୧ ሺͶሻ ሺɘ ή ୧ ሻ െ ୧ ɂ Ɍכ୧
(5)
Ɍ୧ ǡ Ɍכ୧ Ͳǡ ൌ ͳǡʹ ڮǡ ሺሻ WhereɌ୧ Ɍכ୧ is used as slack variable, C is a penalty parameter. The dual problem is: ଵ
σ୪୧ǡ୨ୀଵሺ୧ െ כ୧ ሻ൫୨ െ כ୨ ൯൫୧ ή ୨ ൯ ɂ σ୪୧ୀଵሺ ୧ כ୧ ሻ െ σ୪୧ୀଵ ୧ ሺכ୧ െ ୧ ሻሺሻ ଶ
Ǥ Ǥ σ୪୧ୀଵሺ୧ െ כ୧ ሻ ൌ Ͳሺͺሻ Ͳ ୧ ǡ כ୧ ǡ ൌ ͳǡʹ ڮǡ ሺͻሻ Where K is kernel function. By solving this problem, we get the optimal solutionߙത כൌ ሺߙതଵ ǡ ߙതଵ כǡ ߙ ڮത ǡ ߙത כሻ் , the weight vector߱ ഥ ൌ σୀଵሺߙത כെ ߙത ሻݔ . Thus, we can finally get the decision function: ሺሻ ൌ σ୪୧ୀଵሺכ୧ െ ത ୧ ሻሺ୧ ή ሻ തሺͳͲሻ
804
Siqi Yi et al. / Procedia Computer Science 55 (2015) 802 – 808
3. ITO early warning index based on SVR To compile information technology outsourcing early warning index, we first need to determine the reference indicator. The execution amount of contracts is an important indicator for service outsourcing industry, which reflect benchmark fluctuation cycle of this industry. So we select the execution amount of information technology outsourcing contracts as reference indicator. The data is provided by the service outsourcing research center. Information technology outsourcing includes software research and development outsourcing, information technology research and development outsourcing, information system operation and maintenance outsourcing. According to the research of other authors, information technology outsourcing is determined by many factors, such as software industry [12-16]. Finally, we choose business revenue and profit amount of software and electronic information industry as our first indictors. Some macroeconomics indictors, such as fixed asset investment, are also be selected. The total number of first indicators is 54. By the method of time difference relevance and peak-valley, we select 10 indicators as our final index. They are showed in table 1. Table 1.The final indicators number
indicator
abbr
number
indicator
abbr
1
fixed asset investment
FAI
6
money supply M2
M2
2
export of electronic information product
EIP
7
purchase manager index
PMI
3
export of software
EOS
8
export of software outsourcing
ESO
4
import and export
IAE
9
foreign direct investment
FDI
5
fixed asset investment of tertiary industry
FAIT
10
central finance revenue
CFR
Based on support vector regression model, the input variable can be presented as ୧୬ୢ୧ୡୟ୲୭୰ୱǡ୧ ൌ ሺ୍ǡ୧ Ǥ ୍ǡ୧ Ǥ ୗǡ୧ Ǥ ୍ǡ୧ Ǥ ୍ǡ୧ Ǥ ଶǡ୧ Ǥ ୍ǡ୧ Ǥ ୗǡ୧ Ǥ ୈ୍ǡ୧ Ǥ ୈǡ୧ ሻ. The output variable is execution amount of information technology outsourcing contracts ୧ . The loss function is: ȁሺ୧୬ୢ୧ୡୟ୲୭୰ୱ ሻ െ ȁ െ ɂȁሺ୧୬ୢ୧ୡୟ୲୭୰ୱ ሻ െ ȁ ɂ ሺͳͳሻ க ሺሺ୧୬ୢ୧ୡୟ୲୭୰ୱ ሻǡ ሻ ൌ ቄ Ͳ The decision function is of the following form: ሺ୧୬ୢ୧ୡୟ୲୭୰ୱ ሻ ൌ ሺɘ ή ୧୬ୢ୧ୡୟ୲୭୰ୱ ሻ ǡ אଵ ǡ אሺͳʹሻ In order to get the weight vector of each indicator, we choose linear nuclear as kernel function form of the model. By solving the dual problem (7), we get the weight vector: ߱ ഥ ൌ σୀଵሺߙത כെ ߙത ሻ୧୬ୢ୧ୡୟ୲୭୰ୱǡ୧ , it can be presented as: ߱ ഥ ൌ ሺ߱ிூ ǡ ߱ாூ ǡ ߱ாைௌ ǡ ߱ூா ǡ ߱ிூ் ǡ ߱ெଶ ǡ ߱ெூ ǡ ߱ாௌை ǡ ߱ிூ ǡ ߱ிோ ሻ் . We obtain the weight of each indicator by normalize it: ݓ ൌ ߱ Ȁ σୀଵ ߱ ሺͳ͵ሻ We use mean square error (MSE) to measure model fitting effect. The smaller the mean square error is, the better model fitting effect is. ଵ
ൌ σ௧ୀଵሺݕ௧ െ ݕ௧ כሻଶ ሺͳͶሻ
805
Siqi Yi et al. / Procedia Computer Science 55 (2015) 802 – 808
4. Experiment 4.1. Data description We choose the monthly data from July in 2010 to August in 2014 as our sample (the sample number is 55). The data is derived from national bureau of statistics of the people’s republic of China and wind database. All of them are ratio data. We preprocess them by the method of X-12 seasonal adjustment. According to the result of time difference relevance and peak-valley, we cluster the 10 indicators into two classes, leading index, coincidence index. Fixed asset investment, export of electronic information product, export of software, import and export, fixed asset investment of tertiary industry, money supply M2, PMI are leading index. Fixed asset investment and export of software are 2 phase before benchmark index, export of electronic information product, import and export, fixed asset investment of tertiary industry, money supply M2 and PMI are 5 phase before benchmark index. Export of software outsourcing, foreign direct investment, and central finance revenue are coincidence index. Before setting up the SVR model, we lag the seven leading indicators as follows. ܺௗ௧௦ǡ௧ ሺݏሻ ൌ ܺௗ௧௦ǡ௧ି௦ ሺͳͷሻ 4.2. SVR model According to the SVR model we introduce in section 3, we construct the SVR model using libsvm 3.20 software. We pick up 80 percentage of data as the training set, others as the test set. Finally, we run SVR on the data set and get the results as follow. Table 2.The weight of each indicator
SVR
FAI (2)
EIP(5)
EOS(2)
IAE(5)
FAIT (5)
M2(5)
PMI(5)
ESO
FDI
CFR
1.152
0.505
1.096
1.127
0.605
0.707
1.249
1.236
0.974
1.348
The MSE of the SVR model is 64. From the weight vector, we can see that the ten coefficient are all positive, which is consistent with economic meaning. For example, the weight of FAI (2) is positive, which means that there is a positive correlation between the fixed asset investment and execution amount of information technology outsourcing contracts. The export of electronic information product’s weight is minimal, which means this indicator is less important to IT outsourcing than other indicators. 4.3. ITO early warning system According to the indicators we select, we construct the coincidence and leading index by the method of economic analysis. From figure 1, we can see that benchmark index lag leading index about six to nine months; coincidence index is consistent with benchmark index. From July in 2010 to May in 2011, information technology outsourcing industry of China has developed rapidly, the growth rate reached nearly ninety-five percent in May of 2011, which may because ITO industry is in the early stage of development, and the development status of ITO is a little overheating. After May of 2011, the development of ITO is slowing. Industry signal lamps is adopt the way of traffic lights to describe industry development status of some important indicators. The red light means the growth speed is too rapid, the yellow light shows a little fast, the green light represent that the status of industry is normal and stable, the shallow blue indicate the speed is somewhat slow, the blue light means the development of industry is very cold. According to the signal lamps figure, we can observe the state of each indicator in the industry, and thus have a better understanding of the
806
Siqi Yi et al. / Procedia Computer Science 55 (2015) 802 – 808
industry. We use 3ɐ statistical rules for warning limit, The critical value of five status is Ɋ െ ʹɐǃɊ െ ɐǃɊ ɐǃɊ ʹɐ respectively, Ɋ is the average of each indicator, ɐ is the standard deviation. The interval of blue, shallow blue , green, yellow and red is [െλǡ Ɋ െ ʹɐ]ǃ[Ɋ െ ʹɐǃɊ െ ɐ]ǃ[Ɋ െ ɐǃɊ ɐ]ǃ[Ɋ ɐǃɊ ʹɐ]ǃ ሾɊ ʹɐǃ λሿ respectively. We gain the signal lamps of ITO in recent 12 months as figure 2. From September in 2013 to September in 2014, the status of ten indictors is mainly green and shallow blue, which means the development of ITO is normal and stable, but it has a trend of declining. From September in 2013 to December in 2013, only the indicator export of software outsourcing is in the status of shallow blue, since that, the number of indicators which is in the status of shallow blue has increased. benchmark index
coincidence index
leading index
95.00
110
85.00
105
75.00
100
65.00
95
55.00
Fig. 1. ITO coincidence and leading index
Note: indicator refer to FAI, EIP, EOS, IAE, FAIT, M2, PMI, ESO, FDI, and CFR respectively. Fig. 2. ITO signal lamp of early warning in recent 12 months
2014-07
2014-05
2014-03
2014-01
2013-11
2013-09
2013-07
2013-05
2013-03
2013-01
2012-11
2012-09
2012-07
2012-05
2012-03
2012-01
2011-11
2011-09
2011-07
2011-05
80 2011-03
25.00 2011-01
85 2010-11
35.00 2010-09
90
2010-07
45.00
Siqi Yi et al. / Procedia Computer Science 55 (2015) 802 – 808
807
Based on the weight vector we attained from SVR model and the data of ten indicators, we calculate the comprehensive index. The warning limit of comprehensive index has referenced the method used by China Economic Monitoring & Analysis Center National Bureau of Statistics of China†. The centerline of green light area is Nh3, N is the number of indicators. The boundary between green and shallow blue is Nh˄3+2˅/2, the boundary between green and yellow is Nh˄3+4˅/2, the boundary between shallow blue and blue is ˄N h2˅-1, the boundary between red and yellow is ˄Nh4˅+1. We draw the graph of comprehensive index as figure 2. From July in 2010 to July in 2014, the development status of IT outsourcing changed from heat to normal. From July in 2010 to May in 2011, the comprehensive index is in the yellow area, the growth speed of ITO is a little fast. Since then, the comprehensive index decline to green area, which means the development of ITO is normal and stable, but the value of comprehensive index in 2014 is at the bottom of the green area, and it has a trend of reducing. If the government does not take any measures, the development of IT outsourcing is likely to slow.
Fig. 3. ITO early warning index 5. Conclusion The objective of this paper is to construct the indictor system of information technology outsourcing and establish the early warning index of ITO. By the method of time difference relevance and peak-valley, we finally choose 10 indicators. This paper also presented how SVR method can be used in determining the weight of each indicator. Based on the real data of China, we calculated the early warning comprehensive index, we found the development of IT outsourcing in China is slowing, the government should take out some positive measures to promote the development of IT outsourcing. For example, from figure 1, we can see that the indictor fixed asset investment is in the shallow blue area, the government should increase the investment in fixed assets.
† http://www.cemac.org.cn/Ozsbz5.html
808
Siqi Yi et al. / Procedia Computer Science 55 (2015) 802 – 808
Acknowledgements We would like to thank service outsourcing research center for providing the data set. This work has been partially supported by grants from National Natural Science Foundation of China (NO. 71331005, NO71110107026.) References [1] Corbett MF. The outsourcing revolution: why it makes sense and how to do it right. Kaplan Publishing, 2004. [2] Gartner. Gartner on Outsourcing. Gartner, Inc; 2008–2009 [3] DataMonitor, Global IT Services: Industry Profile, Reference Code: 0199–2313,2007 [4]Wong KKF. The relevance of business cycles in forecasting international tourist arrivals. Tourism Management,1997;(18):581586. [5] James Kolari, Michele Caputo, Drew Wagner. Trait Recognition˖An Alternative Approach to Early Warning Systems in Commercial Banking[J].Journal of Business Finance&Accounting.1996;23(9/10). [6]Niemira M P. Developing Industry Leading Economic Indicators. Business Economics, 1982: 5-16. [7] Ronny Nilsso. OECD System of Leading Indicators Methodology and Application. National Development and Reform Commission Wrokshop and Expert Hearing on Business Cycle and Indicators; 2004. [8] Ataman Ozyildirim, Brian Schaitkin, Victoe Zarnowitz. Business Cycles in the Euro Area Defined with Coincident Economic Indicator and Predicted with Leading Economic Indictors. Journal of Forecasting, 2010;29: 6-28. [9] Myott E. Early warning system feasibility in the Hamline Midway area. Neighborhood Planning for Community Revitalization, 1999. [10]Vapnik VN. The Nature of Statistical Learning Theory. New York: Springer; 2000 [11]Yu X, Qi Z, Zhao Y. Support Vector Regression for Newspaper/Magazine Sales Forecasting. Procedia Computer Science, 2013;17: 1055-1062. [12]Arora A, Gambardella A. The globalization of the software industry: perspectives and opportunities for developed and developing countries. National Bureau of Economic Research; 2004. [13]Koh C, Ang S, Straub D W. IT outsourcing success: a psychological contract perspective. Information Systems Research, 2004; 15(4): 356-373. [14] Loh L, Venkatraman N. Determinants of information technology outsourcing: a cross-sectional analysis. Journal of management information systems 1992: 7-24. [15]Dossani R, Kenney M. Offshoring: Determinants of the location and value of services. Asia Pacific Research Center, Stanford University; 2004. [16]Lacity M C, Willcocks L P. An empirical investigation of information technology sourcing practices: lessons from experience. MIS quarterly, 1998: 363-408.
