Based On The Degree Of Interaction Module Weighted Graph Found Online Community

Community detection is a hot research subject in computer science and technology recent years,both the experts and scholars in the sociology and nature science fields are doing their best to research the secret of the technology of community detection. After Euler gave the method of solving the Knigsberg Bridge Problem,the scientists and mathematicians or other researchers have devoted their lives and energies to the work of discovering the secrets of graphs.Since1930s,the research of social network has always been an important research topic.Most algorithms of community detection are based on the graphs which are undirected and unweighted,but some of the networks in the real world are weighted or directed.Taking this situation into consideration,we think that it is significant to do the work of community detection on the weighted and directed graph,and this will also meet the actual requirement of the real world to some degree.We have done these work and have some innovation points below:(1)First,considering the situation that there is rather few studies of the community detection about weighted social network,we take the weights and directions among members in the weighted graph into consideration. We improved the method of measuring the results of community detection,raised a new concept called interaction modularity and explained how to define it and calculate it on the weighted social network.All these work done are based on the theory of interaction degree which is created two years ago in my teacher’s doctoral dissertation.(2)Designing an algorithm of community detection based on the theory of interaction modularity and choice of greedy strategy.We design this algorithm aimed at doing the division and community detection work on the weighted graphs,hoping we will find the maximum of interaction modularity and the division of social network which matches it.We deem that the division that matches the maximum of interaction modularity is an ideal situation close to the real case of the social network.(3)According to the definition and calculating methods of interaction modularity,we do the work of community detection on the weighted graph by making use of the algorithm above to see whether it works or not. We test it on the Karate Club social network and other networks,the results are similar to the real situations and the deviations are acceptable.We find that the maximum of the interaction modularity matches an ideal division of the weighted social network that is close to the real case.The interaction modularity that was raised in this article can effectively measure that whether the result of partition the network is acceptable or not and is similar to the real fact.We test the effectiveness of interaction modularity on the Karate Club network,dolphin social network and the rhesus monkey social network,the results indicate that interaction modularity is valid in measuring the partition of the weighted social network graphs.