A new similarity measure for extraction information from social networks and improve the community detection and recommendation results

Social networks contain many information about their authors and extracting these information is one of the important tasks nowadays. Network´s weight matrix just shows the relationship between adjacent nodes and cannot show more information about network´s structure. On the other hand, clustering and community detection is one of the underlying problems in social networks that network´s weight matrix is the basic entry of all clustering algorithms and if we have some extra relationship between nodes, we can do a better clustering. For this aims we propose a new way that extract some similarity between nodes and produce a square similarity matrix. This matrix help to extract more information about community structure and indirect relationship between nodes, and also using of it instead of weight matrix improve clustering results. We use NMF and SVD as the underlying clustering algorithm and seen that our produced similarity matrix improve the results.