Beijing University of Posts and Telecommunications–GeoTemporalNet

VAST 2008 Challenge
Mini Challenge 3:  Cell Phone Calls 

Authors and Affiliations:

Qi Ye, University of Posts and Telecommunications, jack_hill@263.net

Tian Zhu, University of Posts and Telecommunications, zhutian0403@163.com

Deyong hu, University of Posts and Telecommunications, deron.hu@yahoo.com.cn

Student team: YES
      
 faculty advisor:

Bin Wu, University of Posts and Telecommunications, wubin@bupt.eud.cn

Bai Wang, University of Posts and Telecommunications, wangbai@bupt.edu.cn

Tool(s):

In order to combine the analysis of geographic information and link relationships, we developed a tool called GeoTemporalNet based on the framework called JSNVA. GeoTemporalNet is composed of two major parts, one part is used to show the geographical information in the grid on the panel and the other part is used to show the structure of different networks. JSNVA (Java Straight-line drawing Network Visual Analysis framework) is a software framework JSNVA for network visual analysis in different applications. JSNVA has a clear architecture and supports a more systematic way of implementing different straight-line graph drawing algorithms which show different networks on different views. JSNVA can be used as a front-end for visualization and a back-end for analysis in applications, and it can be customized for different applications. The paper of more details about JSNVA will be published in ADMA2008. Based on JSNVA, the data in GeoTemporalNet’s applications are stored in a graph structure as nodes and edges, and if there are relations between two nodes there will be an edge. The information in nodes and edges can be stored and achieved by different keys, such as time, position, names et al.

In this mini challenge, we just use the graph data structure to record the relationships of different people as shown in the video. Both the time information and position information of each people are stored in these nodes. In this challenge, we mainly use the force directed layout algorithm to show the social network. By using this tool users can not only get the temporal information of all people’s positions in the building dynamically but also the evacuation trace of each person statically. 


Phone-1: What is the Catalano/Vidro social network, as reflected in the cell phone call data, at the end of the time period  

PhoneNodes.txt

    PhoneLinks.txt

Figure 1 the Catalano/Vidro social network

In this challenge we try to find the social network by the egocentric networks of certain persons whose IDs are 200 and 300, as shown in figure 1. We find that there are two groups of people who coordinate high-level in the social network and communicate with a lot of people. One group is the people who communicate with 200, and their IDs are 1, 2, 3 and 5. The other group is the people who have the equal positions in this social network with the people in the first group. The IDs of these people in the second group are 309, 392, 360 and 306. All the people in the second group communicate with the person whose ID is 300. At last we use the egocentric networks of 200 and 300 to detect the social network in this island. We get the network by detecting the network whose roots are 200 and 300 in 3 jumps. We regard that the hierarchy of social network can be divided into 3 levels: the first level is composed of the persons whose IDs are 200, 300 and 0; the second level is composed of the persons in these two groups whose IDs are 1, 2, 3, 5, 309, 392, 360 and 306. The people in the last level are those who communicate with these two groups. The persons whose IDs are 200 and 300 are in green; the persons who communicate with 200 are in red; the persons who communicate with 300 are in dark blue; others are in light blue.

 

Phone-2  Characterize the changes in the Catalano/Vidro social structure over the ten day period.

Detailed Answer:

We believe that the person whose ID is 200 is Ferdinando Catalano and the person whose ID is 5 must be his brother Estaban Catalano. As the person whose ID is 200 have communicated with 6 persons and the person whose ID is 5 have communicated with him most frequently. They have communicated for 14 times, 15183 seconds during these 10 days. The person whose ID is 3 the second one who communicated with Ferdinando Catalano most frequently, while they had communicated for 8 times, 8520 seconds. So we believe that the person whose ID is 5 must be Estaban Catalano. We use the page-rank algorithm to find the people’s centrality in the social network, and the size of vertices in these figures show their importance. As David Vidro coordinates high-level Paraiso activities and communications, we find that the person whose ID is 0 communicated with all the important people who communicated with 200. He also communicated with the persons in the second group whose IDs are 306 and 309, as shown in figure 2. We also guess that the persons whose IDs are 1 and 2 may be Juan Vidro and Jorge Vidro. The one whose ID is 0 should be David Vidro.

Figure 2 the neighbors of 0

By visualizing the calling behaviors of the people in the video, we find out that the structure of the social network changed greatly on the 8th day. The people who communicated with 1, 2, 3 and 5 changed to communicated with 309, 392, 360 and 306. In order to find out all the persons who share the common neighbors, we set each person’s neighbors as a set and use Jaccard Coefficient to get the similarity of any two nodes’ neighbor set. The coefficient of 1 and 54 nodes is 0.75; the coefficient of 5 and 306 nodes is also 0.68; the coefficient of 3 and 360 nodes is also 0.6; the coefficient of 2 and 397 nodes is 0.59.

Figue 3 (A ) shows the social network of the first 7 days, and figure 3 (B ) shows the social network of the last 3 days. The green nodes are the ones in both social networks. Figure 4 shows the social network by using combing the common nodes in figure 3 (A) and (B) in the third level. We find out these persons who communicate with 200 seldom communicate with others. While the people whose IDs are 309, 392, 360 and 306 take the place of these people in the social network.

Figure 3 the temporal social network (A) the first 7 days (B) the last 3 days

Figure 4 the social network combined by the common nodes in figure 3 (A) and (B)