Research And Implementation Of Community Detection For Multi-relational Networks

The development of online social media has promoted the progress of complex network and traditional social network research,providing more plentiful data source than ever before.Nowadays,research based on real social situation has caused wide concern in academic world increasingly.While,community detection for social networks is a basic research content.Currently,most of those algorithms are based on homogeneous networks,which means that nodes of a network model are in single relationship.The reseach has some achievements,but the information implied in a network are usually ignored.This paper analyzed the feature of social network based on models from other scholars,and defined the concept of influence between nodes in a network.Then MRCD algorithms are proposed,which were sutitable for static multi-relational social networks and more closely to real social situation analysis.The algorithm were verified through the theory explanation and experiment as well.In order to adapt to dynamic network,this paper analyzed the changes in networks which includes adding and removing nodes or edges,which results in the growth,cutting,merger,splitting,birth,and death in community changing process.Then criterion for the six changes were provided in the DMRCD algorithm.In the DMRCD algorithm,data were split into several parts according to the recorded timestamp,then the network were analyzed in several phases and a reasonable explanation were given finally.Currently,experimental data suitable for multi-relational network from the Internet is extremely lacking and limited,which is an obstacle to research deeply into community detection of social network.To study more on these problems,an experimental platform were implemented to simulate the real social situation in the recent two years.Hundreds of college students were invited to take part in a variety of interactions online,a mount of data for multi-relational network were collected to verify the proposed algorithms.The experiment result shows that the propagators of the platform have more influence ability in the network,which is consistent with the actual situation.As time goes on,along with the increasing of the network module,its community structure become more obvious,which reveales the validation and rationality of the algorithms.