This website has been developed by research scientists in Public Data Research Center, Advanced Institutes of Convergence Technology (AICT).
Our research agenda is about utilization of public data based on public data mining and visualization.
Through our development result, the general public can gain insights from raw public data.
In particular, we would like to show the voting patterns of UN members over time.
Below is the outline of our work.
We collected UN voting record between 1946 and 2012.
In practice, we used approximately 5,000 bills in http://hdl.handle.net/1902.1/12379.
For each bill, the vote for every UN member (e.g., Korea, Republic of) is recorded in In favor, Against, Abstain, or Absent. From about 5,000 UN General Assembly voting records, the goal of our research is to identify how similar the voting pattern of UN members is.
To achieve this goal, our visualization technique consists of three steps as follows.
Step 1: Computation of similarities.
Let us suppose that two countries such as Korea and Japan voted two bills.
In the first bill, both voted the same (e.g., In favor).
On the other hand, in the second bill, Korea voted as On favor, while Japan as Against.
In those cases, we can simply compute the similarity value (sim) between Korea and Japan as follows.
sim(Korea, Japan) = x / y, where x and y indicate the number of the same votes and the number of votes, respectively.
In this example, sim(Korea, Japan) = 1 / 2 = 0.5.
Please note that Korea and Japan voted the same in the first bill so x is 1.
The similarity value between two countries a and b is sim(a, b) as normalized in between 0 and 1.
If sim(a, b) = 1, it implies that the votes of a and b are totally the same.
Otherwise, the vote of a is not correlated with that of b.
Step 2: Social network formation.
If sim(a, b) > a pre-determined threshold value, a is connected to b in a social network.
Until now, every previous method has applied this process to many different domains.
However, since social networks generated are considerably dense and complex, it is hard to visualize and understand such social networks with insights.
To tackle this problem, we propose a friend-matching algorithm.
After similarity values are estimated in the first step, we filter out undesirable links among nodes as follows.
Assuming four countries a, b, c and d, we suppose that b and c are considered as friends of a because sim(a, b) and sim(a, c) are the most highest scores among the other similarities sim(a, *)s.
In the same way, we also assume that a and d are friends of b by the two highest similarities of sim(b, *)s.
Then, a is connected to b because a and b are friends each other.
In other words, a and b are linked in the social network if a is a friend of b as well as b is a friend of a.
Step 3: Layout algorithm for visualizing a social network.
For our implementation, we modified Fruchterman-Reingold and Anchoring algorithms.
Initially, nodes are plop down randomly on a plane.
Iteratively, nodes push each other apart and edges pull related nodes together.
Finally, final (x, y) coordinate for each node is computed on the plane.
Politiz UN shows three different types of graphs.
The first one is social networks called Network.
The second one is bubble heap graphs named as Bubble.
Finally, we have a look at circular graphs called Circular.
-Network: This shows social networks of voting patterns of UN members over time (i.e., from 1946 until 2012).
If a country is near to another country, the voting pattern of the two countries is similar.
Otherwise, the voting patter of them is different.
In addition, UN voting bills are clustered to six categories with the same topic Middle East, Nuclear, Disarmament, Colonization, Human Rights and Economics.
In each category, we can see the voting patterns of UN members.
-Bubble: In this graph, we show Bubble Heap graphs for UN voting record.
Please note that our proposal named as bubble heap graphs shows better visualization than existing social networks so as to compare a pivot node to all nodes in a social network, where each node stands for a country and the numeric value between two countries (a pivot node as one of them) indicates the similarity value of voting patterns between the two nodes in the social network.
-Circular: The basic concept of circular graphs is similar to that of bubble heap graphs.
In a circular graph, a pivot node (e.g., U.S.) is located in the center of nested circles that seem archery targets.
Some nodes may be in distance from the pivot node over time.
If a node is far away from the pivot node, it indicates that the voting pattern of the node is different from that of the pivot node.
In addition, we also present a proximity based visualization technique for figuring out temporal pattern like what significant events have occurred over time.
Indeed, it is non-trivial to find out temporal patterns including events in existing social networks.