Improving Louvain Algorithm for Community Detection

Abstract

Community is one of the important characteristics of reality network, which can effectively reflect the inner information of network and the relation among nodes. For the division of the community there already had many effective algorithms, the Louvain algorithm based modularity is a more popular community discovery algorithm because it can divide network into different hierarchical community structure quickly and efficiently. But, with the increasing size of network, the Louvain algorithm still has a serious problem that has relatively high time complexity in handling massive data. Faced with this situation, in this paper we ensure the merit of the Louvain algorithm and combine with the LPA algorithm which has advantage of effectiveness, proposing an improved algorithm integrating the Louvain algorithm with the LPA algorithm. Through later experiments, the improved algorithm can obviously decrease time complexity, reduce execution time, and ensure the result accuracy compared to original Louvain algorithm.