Research On Close Relationship By Using Mobile Track And Communication Information

With the development of mobile communication technology, mobile phone users are getting into the 4G interconnection era from 3G era. Because of the high communication speed and improvements of intelligent terminals, the quality of the daily communication gradually improves. Lots of tech enrich people's interpersonal exchange and make their lives colorful, such as stable network, precise object positioning technology, a wide variety of mobile applications and so on. In general, people's social community will be gradually stabilized over time. Using the social behavior related data between mobile users, we can predict close relationship between them. The close relationship is the basis for researches and applications in many other fields, such as friends recommendation system, discovery of communities with similar interests, criminal gangs prediction, operators'marketing strategy. In this paper, we will mine the close relationship between mobile phone users by using their trajectory and communication information. The main work and innovations are as follows:(1)An algorithm is proposed, which utilizes mobile tracks to find user's co-occurrence patterns. The algorithm emphasizes the regularity. From the living pattern standpoint, we can suggest that there may be a close relationship between users by finding the co-occurrence. Experimental results show that the algorithm can effectively find meaningful community relationships and the changes in the user's lifestyle.(2)A close relationship model is built by using mobile phone users' communication data. Besides the duration and frequency, we extract the attribute of time frame, which clearly depicts the time behavior of users. In addition, the algorithm quantifies the feature of wishes between users, which expresses subjectivity of communication. Based on the timeliness and subjectivity, the proposed model can further divide the close relationship. Results of tests show that the algorithm can effectively demonstrate close relationship between users.(3)A user's co-occurrence pattern parallel algorithm is proposed. To solve the bottleneck problem of data storage and computational efficiency, we use the MapReduce programming model and big data processing platform to implement the parallel algorithm. The experimental results show that this parallel algorithm can improve the efficiency of the co-occurrence pattern discovery. Meanwhile, with the growing amount of data, the running time of the parallel algorithm increases linearly.