Analysis Of Users’ Communication Behavior Based On Social Network

With the vigorous development of mobile communication technology, people’s social channels extending from the traditional social networking sites to the mobile field, formed by the mobile device for mobile social network has mobility, portability, instantaneous and customization features vector. Users can obtain the needed information via mobile devices, such as through the mobile phone search for information, listen to music, browse the videos, send message, write micro-blog and so on. Increasing mobile users generate massive user activity data, which provides a valuable resource for academic researchers to carry out various analysis and application research work.Based on this, this paper takes the actual mobile phone communication network as the research object, extracts and analyze user behavior characteristics as its social attributes, geographical attributes, and combined with the characteristics of the existing multi-dimensional data of behavioral characteristics to optimize of existing behavioral characteristics of individual mining techniques. The main research contents and innovations of the paper in the following aspects:(1) Analysis and recognition on multidimensional communication characteristic of mobile users.Based on the existing mobile phone communication data, we analyze using rules of the mobile communication network, explore their calling, text messages and surfing habits, and extracting characteristic values based on user behavior rules of different identities, finally, using the ensemble learning method of hybrid structure for classification experiments on network user identity, to identify the identity of the user.(2) Analysis and research on node importance in social network.We analyze several models for calculating the node weights in communication network of mobile phones, and propose a new evaluation model to measure how important a node is based on its degree and clustering coefficient. Then we propose a new parameter called neighborhood efficiency along with node degree to adapt our model and improve the accuracy for node weights evaluation. We demonstrate the effectiveness of the proposed two methods via a real social network. Compared to traditional methods, the two methods can more accurately evaluate the importance of the node. Experiments also demonstrate that the method based on the degree and neighborhood efficiency has a higher accuracy.(3) Spatial Proximity and social characteristic analysis.We first propose a new measurement called geographic community for evaluating the impact of spatial adjacent to the evolution of the community structure in mobile social network.By a real dataset containing thousands of mobile phone users in a provincial capital of China, we mine the correlation between geographic community and common social properties of users. We first propose two new measurement matrixes which provide a bridge between spatial proximity and the social properties of individuals in mobile social networks. Combing these matrixes and the known adjacency matrix in social network, the joint nonnegative matrix factorization (JNMF) method is used to mine network structure evolution. Finally, a collaborative matrix is obtained to analyze the existing community structure and potential links in mobile social network.The research helps to reveal the mobile users’ behavior rule, and helps the mobile operators to provide better service, gain more profits, and also helps promote the development from human dynamics model research and business intelligence research to quantitative and digital direction. The paper has the important theoretical and practical application value.