Community Detection Using an Optimized Label Propagation Algorithm

The aim of social network analysis is to search for implicit, previously unknown, and potentially useful information. And community discovery is an important method to obtain these information. Label propagation algorithm can provide an efficient way to discover communities in a large-scale network. In each iteration, the label for each vertex is replaced with the most frequent label from its labels. However, this method will lead to many fragmentary communities. To modify this algorithm, we first select some influential nodes as seeds. And then a node will choose its label based on the summation of influences of its neighbors. Our experiment show that the communities discovered by the optimized algorithm is improved.