Analysis of Condition for the Cooperation Achievement on Arbitrary Networks (Extended Abstract)

We analyze relations between evolution of cooperation on networks and network structures. Most researchers have used simple networks, such as scale-free networks and random networks. We performed decision tree analysis on various networks. The decision tree is constructed from cooperation degree and five network feature. Here, these five network features are selected from a lot of network features. We estimate their cooperation degree by SVR. As a result, estimation is successful because estimation values are correlated with measurement values strongly in the correlation coefficient of 0.8. Therefore, it is said that cooperation degree can be explained by five network features. The decision tree is constructed with cooperation degree and five network features, and then analyzed. We found that the conditions for achieving cooperation are having few hub nodes, being not very disassortative, and having a short average shortest-path length.