Research And Implementation Of Time-Varying Mobile Community Mining Algorithm

With the rapid development of the information technology, the social network becomes everywhere. From biology to the internet, from large power grid to relationships and so on. Data mining which we extract effective, credible, novel information from large data becomes today’s hot topic. The application of data mining effectively in social network analysis becomes a new direction in data mining. Many studies and experiments show that a lot of network structure are heterogeneous, namely, social networks are composed of many different properties nodes connecting together. Our goal is to find the social network subgrap which consist of with the same properties, consistent type, i.e. find out the community structure.This thesis mainly uses community mining technology in mobile social network. We firstly collect mobile telephone records from mobile site by programming means, and then according to the records, we put forward a kind of weight relationship measurement method. Because mobile calls are a direct end-to-end communication mode, therefore, these that they communicate directly and closely can be the community members. So using cliques as the structure basis of communities. We firstly get mobile community structures through using clique percolation method with weight (CPMw). Then for better observing the evolvement of the mobile community structures, we put forward a community updating algorithm based on appending and cutting edges (ACUM). We also do some improvement of the algorithm to improve the efficiency by using parallel execution.According to the research of experiment, it shows a better result that we adopt CPMw algorithm in finding community structures. Then we put forward ACUM algorithm which made a good effect in community update, and also adopt parallel community update that improved algorithm performance and got high accuracy.