Visualization of People Attraction from Mobile Phone

Download PDF
0 downloads 0 Views 671KB Size Report
Comment
suhad.behadili@etu.univ-lehavre.fr ... Baghdad University, Computer. Science ... transportation routes, concentrations of people, urban planning, etc. The ability ...
Visualization of People Attraction from Mobile Phone Trace Database: A Case study on Armada 2008 in French City of Rouen Behadili Suhad Faisal

Cyrille Bertelle

Loay Edwar George

Normandie Univ, LITIS; FR CNRS 3638, ISCN, ULH, Le Havre, France [email protected]

Normandie Univ, LITIS; FR CNRS 3638, ISCN, ULH, Le Havre, France [email protected]

Baghdad University, Computer Science Department, Baghdad, Iraq [email protected]

ABSTRACT

The mobile phone is widespread all over the world. This technology is one of the most widespread with more than five billion subscriptions making people describe this interaction system as Wireless Intelligence. Mobile phone networks become the focus of attention of researchers, organizations and governments due to its penetration in all life fields. Analyzing mobile phone traces allows describing human mobility with accuracy as never done before. The main objective in this contribution is to represent the people density in specific regions at specific duration of time according to raw data (mobile phone traces). This type of spatio-temporal data named CDR (Call Data Records), which have properties of the time and spatial indications for the elaborated environment. City life understandings help urban planners, decision makers, and scientists of different fields to resolve their questions about human mobility. Such studies are using a very cheap, most spread tool that is the mobile phone. Mobile phone traces analysis gives conceptual views about human density, connections and mobility patterns. In this study, the mobile phone traces concern an ephemeral event called Armada, where important densities of people are observed during 12 days in the French city of Rouen. To better understand how people attracted by this event, city area during these days of this ephemeral event, is used. Armada mobile phone database is analyzed using a computing platform integrating various applications for huge database management, visualization and analysis, in order to explore the urban pulse generated by this event. As result, city pulsation and life patterns are explored and visualized for specified regions.

Keywords: mobility, density, mobile phone, CDR, urban, GIS.

1. INTRODUCTION Armada is the name of famous marine festival occurring every 4 to 5 years periodically, for ten days period [6, 7, 8, 9, 10, 11, 12]. Large sailing ships, private yachts, naval ships and even military frigate schooners, are gathering from twenty different countries, in one of the world free spectacle in Rouen city (capital city of Upper Normandy in France). This case study is observed the fifth edition of this

event for 4th -15th July 2008, however its activities starting from 10:00 to 20:00, every day, with miscellaneous activities. The last day of this event coincides 14th July (French national day). Attracting 8 millions of visitors with Rouen citizens ( is estimated to 460,000 in 2008).

2. Mobile Networks Data Mobile networks provide huge databases that are containing valuable spatiotemporal information at level of citywide or even nationwide. This type of information is relevant not only for the telecommunication operator, but also for broader set of applications with social connotations like commuting patterns, transportation routes, concentrations of people, urban planning, etc. The ability to efficiently query CDR databases in search of spatio-temporal patterns is the key to analyze human mobility during this event [1, 2, 3, 4, 5]. However, this kind of analysis have two limitations, first one is the information given by mobile phone traces is incomplete to describe all the individual mobility because a lot of people are moving in the city without using mobile phone. But the importance of mobile phone usage today allows us to produce an acceptable approximation of mobility. Second one is the mobile phone traces are only caught by antenna, which are located at some specific places in the city. To describe the human mobility, it is needed to interpolate individual position and to simulate people behavior from these traces in order to reconstruct human trajectories. The manipulation of this database needs several phases to get required analysis results, obtain precise expected knowledge: indication of the city pulsation, population density and people mobility [13, 14, 15]. Assuming that, each mobile represents a human (occurrence/mobility). 3. Research Methodology The essential material used in this study is a database generated by given mobile phone operator (Orange Company, the historical main operator in France), in form of CDRs (Call Record Details). This data covers the period of July 4-15, 2008, includes 50,982,274 events, 615,711 mobile users, during 273 hours. The mobile phone Database is composed of data records contain: mobile IDs (alias), towers IDs and positions which are geo-referenced by 2D-coordinates (x, y), number of cells on each tower and Cells IDs, mobile activities type (call in/out, SMS in/out, mobile hand over, abnormal call halt, normal call end), date and time of the mobile phone activity recorded. The analysis and representation of people concentration of the considered region, is to obtain the most active (regions/time) in the city. Specific platform has been developed to integrate and analyze the initial raw database; PHP server and SQL query language are used. Data visualization is done using graphs, representing regions densities corresponding to human activities. ArcGIS platform is used for geographic representation, and to extract the aggregated data of five classified sectors corresponding to specific spatial sub-areas. The combination of these tools allows lining up the spatio-temporal data on the city map. Divide the study area into 5 sub-areas to summarize the spatial patterns, as Voronoi polygons like in [16]; each polygon is associated to antenna and depicts the area under main influence of this antenna; hence each Voronoi cell centered on one antenna corresponds to all the closest possible locations of individual detected by this

antenna (antenna coverage area). Aggregated adjacent polygons formulate five sectors, as described in Figure 1: (i) The Center sector is around Seine river banks,

Figure 1: ArcGIS Output from Armada DB. Each Voronoi cell is centered on each Antenna. The 5 sub-area (described by 5 color zones) are built by grouping Voronoi cells. Black anchors symbols represent places along the Seine river platforms where boats are situated during the Armada.

the place of Armada event, (ii) The Eastern sector to the Center, (iii) the Western sector to the Center, (iv) the Northern sector to the Center, (v) the Southern sector to the Center. Figure 2 describes activity density of the 12 days period, each curve represents one day activities, note that day 9 has lack in its data at period 02:0003:00 Pm producing irregular curve. 4. Data Analysis and Visualization The analysis of daily activities ratios of the 5 main sectors is described in figure 3, by summing all the activities of each day. All sectors have highly activities ratios in day 4, since it is former to Armada event, influence by arrangements is clear. All sectors have decreasing in their activities ratios, in days 6, 12, 13 (off days), means that citizens are in lowest activities in them, however their highly activities ratios in days 7-11 (work days). Anomalous event appear in day 14, where all sectors are in their lowest activities ratios except the west sector, where its activities ratios mark peak values, means that France National Day influences this sector, the ceremonial may be done there. Corresponding to each of the five sectors (explored in figure 1), each has activity densities indications according to the 12 days period: The center sector, as in Figure 4, activities are decreasing over time interval (0:00-06:00), lowest activities ratio at hour 05:00. The activities increased over time interval (06:00-20:00) and decreasing gradually over time interval (20:00-23:00), the peak activities ratio in times (12:00, 17:00-18:00). Regarding activities ratios, day 4 has highest ratio (former to Armada event), the preparations start and visitors coming, followed orderly by days: 11, 8, 10, 7, 5, 15, 13, 6 and14. This sector has very similar patterns, sometimes interleaved; the lowest activities ratios in day 14 (French National Day/vacation). Since it is approximately the center of city in addition to Armada event occurrence hence, it is active almost days.

Figure 2: Activity patterns according to people density during day hours, over the whole Region area

Figure 3: Daily analysis of the 5 sectors along 24 hours with average over the days period of Armada

The east sector as in Figure 5 has decreasing activities ratio over time interval (0:0006:00), whereas lowest activities over time interval (04:00-05:00), then increased over time interval (06:00-20:00) at peaks (09:00, 12:00, 18:00), which are times of start work, lunch, end work. Days 4 &10 have anomalous activities with peak at 09:00; still day 4 has the highest activities. Activities in almost days in descending order: 4, 10, 11, 8, 7, 15, 12, 13, 6 , 14, where most of them are off days, so it could be an entertainment or commercial region. Day 9 has lost data but according to the provided (15:00-23:00), it is close and interleaved with other days, day 14 has lowest activities ratio.

Figure 4: Centre sector activity analysis, according to the average of people density along day hours .

Figure 5: East sector activity analysis, according to average of people density along day hours

The west sector as in Figure 6, marks different patterns with highest activities in day 14, most influenced sector by National France Day, activities increasing over time interval (08:00-15:00), in peak hours (11:00 and 19:00). Almost days activities are decreasing over time interval (00:00-07:00), then increasing over time interval (08:00-19:00). Days activities in descending order: 14, 4, 11, 10, 8, 15, 12, 5, 13 and 6, here activities are approximated except day 14. The north sector as in Figure 7 has similar patterns with highest activities in day 4. All days have similar patterns to previous sectors, where activities are decreasing over time interval (00:00-06:00), the lowest over time interval (04:00-05:00), then increasing over time interval (06:00-19:00), at peak hours (12:00 & 18:00), then decreasing over time interval (20:00-23:00). Most active days here could be ranked according to their activities descending: 11, 10, 8, 9, 15, 12, all are work days except day 12 is off day; whereas 14, 6, and 13 (off days). The south sector as in Figure 8 has similar patterns, where most highly activities ratios in day 4, then decreasing over time interval (00:00-06:00), again they are decreasing over interval time (06:00-20:00) at peak

hours (12:00 & 18:00), which are lunch and after work time. Then activities are decreasing over time interval (20:00-23:00). Most active days in according to their descending order: 4, 11, 10, 8, 7, 15, 5, 12. In spite of days 5 &12 (off days), but they have remarkable activities, whereas lowest ratios are in days 13, 6, 14 (off days).

Figure 6: West sector activity analysis, according to average of people density along day hour

Figure 7: North sector activity analysis, according to average of people density along day hours

5. Conclusions The spatio-temporal datasets that are acquired from mobile phone data have a huge size for millions subscribers and it holds many events either anomalous or normal events. The normal events give a good reflection to the daily life in the studied region like citizens’ habits and social activities. The anomalous events data are very useful in detecting the disasters or catastrophes immediately (in real time) with the spatial and temporal indications, so the evacuation, emergency reactions

Figure 8: South sector activity analysis, according to average of people density along day hours

could be taken as soon as possible to any alarm indications, which are acquired from the CDRs data. All CDR details are reported periodically in seconds; therefore it is very accurate real time data for any cellular phone user activity. The major limitation is the mass size of these data (CDRs): it needs sampling and accurate manipulations to preserve their accuracy and to avoid any misused or misunderstanding to its reality. The life patterns of urban areas are extracted by classifying activities density, according to different time intervals over specified region and sectors. This analysis is an attempt to study the city pulse, with regard to the people density alteration in specific regions over different time intervals. The mobile events density point to people density, since each mobile event represents human activity. However, that doesn’t mean absolute indication to pure people activities, because the human could be existed in a place out of coverage area, so this probability make him exist actually, but not necessarily appeared on the mobile net data. Nevertheless, call activities reflect people density in somehow. In our case study, CDRs don’t give us indication to path mobility. In future, the drift of human activity in different places and times would be calculated using simulation platform. The study region sectors density analysis is done along the daily hours, so it had been realized that all sectors have very asymptotic activities patterns. However, the north is coming in the moderate rank, which is close to the calm and stable sectors (west, east) in activities and patterns, rather than active sectors (north, south). Acknowledgments Behadili Suhad Faisal and Cyrille Bertelle activities, described in this paper, are supported by the region Haute Normandie and European Union by the ERDF RISC. Behadili Suhad Faisal is also supported by a PhD grant from the Iraqi Higher Education Ministry and implemented by Campus France.

6.-References 1. Enrique Frias Martinez, Vanessa Frias Martinez, “Method for an automatic identification of urban dense areas from cell phones records,” United States Patent 8,577,380, Nov. 5, 2013. 2. Juli´an Candia, Marta C Gonz´alez, PuWang, Timothy Schoenharl, Greg Madey, Albert-L´aszl´o Barab´asi1, “Uncovering individual and collective human dynamics from mobile phone records,” J. Phys. A: Math. Theor. 41 224015, 2008. 3. Albert-Laszlo Barabasi, “From network science to human dynamics,” in: The first of CEE's DES4: Distinguished Engineering and Science Speaker Seminar Series, Nov 3, 2011. 4. Thomas Couronné, Zbigniew Smoreda, Ana-Maria Olteanu, “Individual mobility and Communicate on patterns, Analysis of Mobile Phone Datasets and Networks,” in: NETMOB conference, MIT, Cambridge, MA, Oct. 1011, 2011. 5. Zbigniew Smoreda, Ana-Maria Olteanu-Raimond, Thomas Couronné, “Spatiotemporal data from mobile phones for personal mobility assessment,” Transport Survey Methods: Best Practice for Decision Making, Emerald, 2013, pp 745-767. 6. http://www.armada.org/en/armada/historique/2008-l-armada-rouen-2008. (Accessed: 15 Apr. 2014). 7. http://rouen-armada-2013.soft112.com/. (Accessed: 15 Apr. 2014). 8. http://www.thefreelibrary.com/NORMAN+ARMADA%3B+TOM+Bodden+ at+the+Tall+Ships+festival+in+Rouen.-a0182116661. (Accessed: 15 Apr. 2014). 9. http://www.armada.org/en/armada/historique/naissance-d-une-idee. (Accessed: 15 Apr. 2014). 10. http://www.invs.sante.fr/Publications-et-outils/Rapports-etsyntheses/Environnement-et-sante/2012/Surveillances-sanitaire-etenvironnementale-de-l-Armada-de-Rouen-edition-2008. (Accessed: 15 Apr. 2014). 11. http://www.meretmarine.com/fr/content/le-port-de-rouen-se-prepareaccueillir-larmada. (Accessed: 15 Apr. 2014). 12. http://www.drakkaronline.com/article62430.html. (Accessed: 15 Apr. 2014). 13. Elissalde B., Lucchini F., Freire-Diaz S., “Mesurer la ville éphémère. Données Urbaines,” 6, Pumain D., Mattei M-F. (eds), Anthropos, Economica, 2011, pp 105-122. 14. Lucchini F., Elisssald B, Freiré-Diaz S., “Caractériser l’attractivité des quartiers urbains par les données de la téléphonie mobile. Information Géographique,” 2013, 1, pp 44-62. 15. Lucchini F., Bernard Elissalde, Sylviano-Freire-Diaz S, “From the Event to the Ephemeral City,” mesm 2011 conference, Amman, Jordan, Nov. 14-16, 2011, pp 88-95. 16. Ryan Bravo, “Research technical paper,” summer research , August 17, 2006.