Overlapped Community Detection using Extended Node Similarity by Local Expansion

Abstract

The study of real networks like social network has been increasingly interested in community research area. With this study, overlapping community detection plays an important role in studying hidden structure of those networks. There are many overlapping community detection algorithms in recent years. In detecting overlapped structure by local expansion strategy, seeds or core nodes are important because communities are formed on chosen seeds. As a result, inappropriate seeds produce low accuracy of community structure. This paper proposes extended jaccard similarity to find appropriate seed. Firstly, identifies seed using extended jaccard similarity fitness. Then, local communities are detected by extending seed according to quality value. The experimental result obtains improved accuracy and performance of algorithm are compared to other local optimization algorithms.