Research On Group Discovery And Behavior Analysis In Mobile Communication Network

With the rapid development of social media networks and information technology, themobile communication network has become the important way that people communicate witheach other and access to information in real social activities. Group relations and behavior in thenetwork space is able to reflect the user’s real social relations and behavior. However, the currentgroup discovery methods mainly exploit the communication relations and attribute informationto mine the group structure of the network, which cann’t be able to discovery user’s real socialrelations in the network. In view of this, the research of this paper is to solve the above problem.Based on the analysis of the existing group discovery methods, a specific group discoverymethod is proposed. Then the method of group associated user mining and group abnormalbehavior mining are proposed to analysis the position trajectory behavior and communicationbehavior based on the specific group.The main achievements and this dissertation can be outlined as follows:1. A specific group discovery method based on mobile communication network is proposed.Considering the aggregation of communication relation and location information, reflecting thespecific group relation, the measurement criterion of communication relation and locationrelation are established. And the improved split clustering algorithm is used to find the group ofclose social relations in real social life. Results show that the utilization of the user’scommunication characteristics and location characteristics be able to reflect the close grouprelations in the social life.2. A depth-relation mining method of abnormal users based on position trajectory associatedis proposed. Based on the specific group, the measurement criterion of abnormal positiontrajectory is established, which is used to find position trajectory abnormal users in the group. Bymeans of the communication relation and position trajectory relation, the associated user miningalgorithm of the specific group is proposed. The method achieves the position associated usermining of the specific group, which reduces the scope of associated user and has a goodaccuracy.3. An abnormal behavior mining method of the specific group based on fuzzy c-meanclustering is proposed. The method firstly classifys the group into several subgroups according tocommunication behavior character. Then the abnormal communication behavior miningalgorithm is proposed by defining the group abnormity gene of subgroups and establishing theabnormal communication behavior criterion. Finally, the abnormal specific subgroup is foundedand the abnormity of the whole group can be measured. the method reduces the false alarm and missing alarm of current abnormal behavior analysis algorithms, improving the accuracy ofgroup abnormal behavior metrics.