Research On Group Users' Service Behavior Based On Data Of Cellular Networks

With the rapid development of wireless communication technology,the traffic volume of mobile communication network is increasing. The development of communication network is restricted by the frequency resources and energy consumption. In cellular networks, the users' service behavior significantly affects the energy efficiency of the network. In order to establish highly efficient network, the characteristics of the users'service behavior must be fully understood. However, there are too many users in cellular networks, and the difference between users makes the research value on the individual behavior is limited. In real networks, users are often aggregated as a group which presents a similar service traffic pattern, namely, the group users behavior. In order to establish an efficient network, service behavior of user groups should be modeled, which will be beneficial to the construction and improvement of the existing network,bring an energy efficient development of the whole network architecture.Based on the analysis of massive data collected in cellular wireless network,this thesis studies the group users' service behavior in detail. The main research contents and innovations are as fellows.First of all, the four typical data types in cellular networks and their three main characteristics are given in this thesis. The characteristics can be summarized as "spatial-temporal distribution" , "data aggregation properties" and "social correlations". This thesis also systematically expounds the research value of massive data collected in cellular networks.Secondly, this thesis makes a detailed modeling of the group users'service traffic distribution, and results show that the traffic is subject to Log-normal distribution in the spatial domain and sine wave superposition model in the time domain. By analyzing the data of multiple cities, a variety of networks and different wireless scenarios, results show that compared with other kinds of distribution, the fitting error of Log-normal distribution is minimum in spatial domain. As for time domain, the fitting accuracy of sine wave can reach more than 0.9.Then, this thesis puts forward the spatial social model of users' service.Based on the social model, the evaluation of cellular network traffic and the recognition of typical network scenarios are realized. The social network analysis (SNA) is used to study cellular network data, with the base station as node and relationship of traffic changes as tie to construct base station social network (BSSN). Results show that, based on BSSN,the traffic flow of other base stations can be evaluated by only 8% of the whole base stations, and the typical traffic scenarios can be accurately identified.Finally, this thesis proposes a method to extract user groups service behavior pattern. Based on user-level data, cluster analysis is used to divide users into different groups, and then the service behavior of each group is studied. The results show that seven typical groups can be identified in the dataset.In this thesis, the service behavior of group users is studied from multiple dimensions based on the massive real data obtained in cellular networks. The analysis results can provide guidance for network optimization, improve network performance, enhance the user experience and refine the operation of the wireless network.