Hybrid Influential Centrality based Label Propagation Algorithm for Community Detection

In recent times, user activities on web-based social networks has increased enormously irrespective of time and place that generates a variety of datasets which further offers tremendous scope for both mining and knowledge discovery. Due to a large number of Social Networking websitesand interactions among people via these sites, rapid growth in social networks has taken place. Community detection is one of important activity at some point in social network analysis in order to exploit the network properties. It is widely applied in various domains like business, marketing, healthcare, biological networks etc. Out of many existing algorithms available for community detection, Label Propagation Algorithm (LPA) is one of fastest graph based semi-supervised community detection algorithm which has near linear time running complexity. Due to it effectual results and easy implementation with low complexity, it is widely used in various domains.However, total randomization in LPA leads to poor stability and low robustness. Sometimes it leads to ahuge community structure or single community even.The proposed algorithm introduces hybrid influential centrality measure in order to improve the performance of LPA. Proposed algorithm reveals more stable and effective results andexperimental results show its improved performance.