University of California, Davis

University of California, Davis - MobiVis

VAST 2008 Challenge

Mini Challenge 3: Cell Phone Calls

Authors and Affiliations:

Carlos D. Correa
Tarik Crnovrsanin
Christopher Muelder
Ryan Armstrong
James Shearer
Kwan-Liu Ma, ma@cs.ucdavis.edu [PRIMARY contact]
University of California, Davis

Tool:

MobiVis is a visual analytics system to visualize social and spatial information in one heterogeneous network, developed in our group. The main contributors to the tool are Zeqian Shen and Tarik Crnovrsaning. MobiVis displays social networks as connected graphs, where nodes represent a certain entity or attribute and links between them represent a given association. For example, in the simplest case, MobiVis displays people as nodes, where a connection between people represents a conversation between the two. To handle large data sets, MobiVis allows us to filter semantically and temporally. The former is obtained via an ontology graph, which allows us to group people based on certain attributes of the data (e.g., cell tower). The latter is provided with the help of a time chart, which allows us to visualize a social network at any interval in time.
More information:
PDF | MobiVis: A Visualization System for Exploring Mobile Data. Zeqian Shen, Kwan-Liu Ma, In Proceedings of IEEE Pacific Visualization Symposium, IEEE VGTC, March, 2008, pp. 175-182.

Two Page Summary

PDF

Video:

WMV (70MB)

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

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

Detailed Answer:

In an attempt to characterize the entire network, we loaded the entire data set into MobiVis, which links people (blue nodes) if they had a phone conversation. Unfortunately, the tight connectivity of the resulting network made it impossible to find interesting patterns. Following the lead that person 200 is likely to be Ferdinando Catalano, we filtered the data to visualize only its closest nodes. Figure 1 shows the social network of person 200.

Figure 1. Overview of the social network of Ferdinando Catalano (id 200).

This reflects the general social structure over, at least, the first seven days. We can further characterize this network by looking at the links between the immediate neighbors of person 200. Persons 5, 200, 97 and 137 seem to form a clique, whereas persons 1,2 and 3 form another. Looking at the amount of communication between those, which is depicted as the thickness of the edges, we discovered that 200 and 5 talk a lot among themselves. The color coding of the edges helps visualize the symmetry of the calls. For example, a warm color (orange) in the middle indicates a symmetric connection (both parties call each other frequently), whereas a biased orange color indicates more calls in the direction of the bias. We then characterized the network as being the connection of the two families: the Catalanos, represented in persons 200 (Ferdinando Catalano), 5 (which we believe is Estaban Catalano, since its tight connection to 200), 97 and 137. And the Vidros, represented in persons 1,2 and 3.

We can further characterize the substructure of the Vidros as hierarchical. Although it was not evident at first, person 1 always calls persons 2 and 3, which led us to believe that he has a role of coordinator. We validated this with another capability of MobiVis, which allows us to display people in the social network according to some semantic filtering criteria. In Figure 2(a), we display the people called by 1 and people who called person 1 . Those people who called person 1 are connected to an orange node, while people who where called by person 1 are connected to a red node. We can see that person 1 had a bi-directional communication with Ferdinando Catalano, but only in one direction with 2,3 and 5. Figure 2(b) shows the same analysis for person 5. We noticed an inverse behavior: 1, 2 and 3 always call 5, but not vice versa. Furthermore, it helped us characterized the social structure better. The high symmetry of communication between 200 and 5 validates our claim about their identities being of Ferdinando and Estaban Catalano, respectively. Person 1, however, seems to coordinate the efforts of 2,3 and 5, which suggests that he can be associated to David Vidro.

(a) Distribution of calls made to and made by person 1 during entire week. (b) Distribution of calls made to and made by person 5 during entire week.
Figure 2. Association rules in MobiVis helps us visualize the distribution of calls made-by and made-to any given person. In our case, it helps us validate our claim that person 5 is Estaban Catalano, due to the amount of calls and symmetry of his communication with Ferdinando. Once 5 is ruled-out, person 1 fits the description of David Vidro, who coordinates high level activities. Here, we see how it always starts communications with the other members of the movement (i.e., ids. 2,3 and 5).

By looking at the time chart, we notice that all these cell phones where relatively quiet on and after day 8. After that day, the network seems to break, and no communication is found between the Vidros and the Catalanos. To obtain a clear answer, we looked at the time chart. MobiVis allows us to perform temporal filtering, where we can select any interval during the week and visualize the network of the people making calls during that time. We attempted to see the network only during the last three days. Since the network was too complex, we performed structural abstraction by discarding those calls between low degree nodes. Figure 3(a) shows results of two different abstractions, depending on the amount of calls that were not explicitly linked. On the left, it is difficult to see a clear structure, but helps us detect frequent callers (and callees), such as nodes 306, 309, 360 and 397. With a better abstraction (Fig.3(a) right), we found that they were all connected to identifier 300. The time chart confirmed that this cell phone was only used during the last three days (Fig.3(b)).

Figure 3(a). Social network on the last three days. We attempted to filter out nodes by degree to obtain a clear view of the network. Only after filtering out the low degree nodes (right), we found an interesting pattern around node 300, which seemed very similar to that of node 300.

Figure 3(b). After having found node 300, we also corroborated its appearance after day 7, which indicated that both the networks of 300 and 200 did not overlap much.

We immediately suspected a close relationship between these identifiers and those connected to 200. To visualize this more effectively, we then expanded our initial filtering to include those calls related to identifiers 300,306,309,360 and 397. Figure 4 shows the MobiVis layout for the entire week. To our surprise, our force-directed layout automatically uncovered an important relationship between them. The identifiers reflect, to a high extent, the same social structure that we have observed until day 7. Because these nodes share a lot of the same contacts, it leads us to believe that the tagged overlapping nodes (or in a close vicinity) correspond to the same person, i.e., identifiers 200 and 300 correspond to Ferdinando Catalano, 5 and 306 to Estaban Catalano, 1 and 309 to David Vidro, and the pairs 2/397 and 3/360 correspond to either Jorge or Juan Vidro. So, we can say that the two networks corresponding to both the Catalano and Vidro families are still present throughout the week, although they have different sets of cell phones.

Figure 4. Social structure for identifiers 200 and 300 for the entire week.

To validate our findings, we also used call graphs, which show every call in the entire data set over time. Figure 5(top) shows a timeline, where every node spans a line and calls are depicted as line segments between them. We color-coded the actors of interest, i.e., the associates of identifiers 200 as red lines and 300 as blue lines. The calls from 200 and 300 are coded as green and yellow lines at the top of the plot. We found that the pattern is maintained throughout the week, which suggests that the two connected networks we found belong to the same people. Figure 5(bottom) shows a different layout for two days. We purposely placed the associates of id. 300 mirroring those of id. 200 and we found a similar pattern and degrees for those nodes, which confirmed our initial findings.

Figure 5. Detailed visualization using call graphs. Top: Call graphs represented as a timeline for days 7 and 8 (top half) and the entire week (bottom half). We make the timelines of 200 and 300 coincide, as well as for the timelines of their associates. In this way, we find that some of the communication patterns are maintained. Bottom: Call graphs on a radial layout. By placing the newly-found identifiers mirroring our previous network, we find that they mimic the same patterns, and have similar degrees (denoted as the size of the nodes).

Geographic changes:
Although the Catalano/Vidro network seems to maintain its social structure, we found that it changes spatially. MobiVis allows us to associate nodes with the cell towers they used to make the calls. Figure 6(left) shows that the Catalanos and Vidros operate mostly on the northern and central islands. After day 7, we followed their second cell phones (associates of identifier 300), and found that they are more mobile (they connect to more towers). See Figure 6(right). Furthermore, they seem to also travel to the southern islands, seen here as towers 20,21 and 22.

Figure 6. Geographical dispersion of the Catalano/Vidro social structure. We can see that, during the first seven days (left), using one set of cell phones, their organization extends to the central and northern islands. However, after switching cell phones (right), we can see that their coverage is wider, they are more mobile (spanning more towers), and now they actually spread towards the southern islands (seen here as the towers 20, 21 and 22 in the bottom right corner).

In conclusion, we see that the social network of Ferdinando Catalano is maintained throughout the week, although their geographical extents change. He and his brother are at the center of communication between the Catalanos and the Vidros. The ability to perform structural and semantic abstractions, as provided by MobiVis, was crucial to understand these relationships.


(a) Distribution of calls made to and made by person 1 during entire week.	(b) Distribution of calls made to and made by person 5 during entire week.
Figure 2. Association rules in MobiVis helps us visualize the distribution of calls made-by and made-to any given person. In our case, it helps us validate our claim that person 5 is Estaban Catalano, due to the amount of calls and symmetry of his communication with Ferdinando. Once 5 is ruled-out, person 1 fits the description of David Vidro, who coordinates high level activities. Here, we see how it always starts communications with the other members of the movement (i.e., ids. 2,3 and 5).


Figure 3(a). Social network on the last three days. We attempted to filter out nodes by degree to obtain a clear view of the network. Only after filtering out the low degree nodes (right), we found an interesting pattern around node 300, which seemed very similar to that of node 300.

Figure 3(b). After having found node 300, we also corroborated its appearance after day 7, which indicated that both the networks of 300 and 200 did not overlap much.



Figure 5. Detailed visualization using call graphs. Top: Call graphs represented as a timeline for days 7 and 8 (top half) and the entire week (bottom half). We make the timelines of 200 and 300 coincide, as well as for the timelines of their associates. In this way, we find that some of the communication patterns are maintained. Bottom: Call graphs on a radial layout. By placing the newly-found identifiers mirroring our previous network, we find that they mimic the same patterns, and have similar degrees (denoted as the size of the nodes).