Research on Multi-label Propagation Algorithm for Community Detection Based on Weibo

Abstract

Community structure in social network has an overlapping property from where people can find a lot of valuable information by community detection. It will help users to make accurate and valuable judgments as well as decisions. However, the increasing amount of information has led to difficulties for traditional technology to meet the requirement of big data processing. In order to effectively process data in large-scale network, a multi-label propagation algorithm for community detection has been implemented by using BSP parallel model. This algorithm combines the idea of LPA and COPRA. It finds out communities with active Weibo users, and then uses these communities as labels to perform label propagation. In the end, it classifies the remaining users’ communities. The algorithm solves LPA’s problems of instability and low accuracy. Its effectiveness and efficiency were tested and verified by using data from Weibo network.