Content-based Effective Link Communities Detection

As the abstract of surfing on the Internet of people, social networks become the object of studying social regulation, intelligence architecture, information spread and information retrieval. Communities are groups of related nodes that correspond to functional subunits. Therefore, community detection is the hotspot and difficulty of social networks study.Recently, communities in networks are often overlapping, hierarchical organization and large scale. Node communities detection cluster the vertices. Link communities is a set of closely interrelated links where each vertex contains some links. Therefore, link communities can explain overlapping intuitively. However, existing methods are complex. So, sometimes they are inefficient on large scale networks. We proposed a method, named LCDCC, for link communities detection on clustering coefficient in linear time complexity.Like many other existing methods, LCDCC relies entirely on network topology. But great noise leads to declined accuracy. In social information networks, vertices have content. Combining content and link can increase the accuracy of community detection. Based on LCDCC, We proposed a way of combining the link and content on weighted local clustering coefficient to detect link communities, called content-based effective link communities detection, CELCD. It can detect not only traditional communities but also general communities in linear time. Experimental results on real-world and synthetic networks show that comparing with other algorithms, both LCDCC and CELCD can find meaningful overlapping communities quickly arid efficiently.