Research On Users' Network Behavior Based On 3G Core Net Data

With the popularity of 3G network,it is more convenient for people to use the Internet,the communication operators have a large number of 3G users,more and more researchers have joined the ranks of 3G network data analyses and researches.At present,the research of 3G network data is mainly the analyses and researches of users' mobile behavior based on the operation and maintenance data,but rarely involves the users' network behavior.In today's Internet era,it is very important to study users' network behavior.By studying network behavior of mobile users,operators can not only keep track of users' network behavior and explore the deeper needs of users,but also can use the results of the study to guide operator's marketing strategy and realize the oriented marketing of users.In the study of users' network behavior,the thesis achieve the recognition of mobile applications based on the acquisition data from 3G core network;then design and implement users' network behavior recognition and prediction system.Through these works the thesis analysis and mining of users' network behavior comprehensively.The main work of this thesis is divided into the following parts:Firstly,the thesis introduces the research purpose of users' Internet behavior as well as the domestic and foreign research results on this topic.By introducing the acquisition and processing of 3G core network data,the thesis confirm the data acquisition interface and get available data set that can be used.Secondly,the thesis realize the identification of mobile applications and makes analysis of user's network behavior preferences.Through the DPI technology,the thesis analyzes current mainstream mobile applications and extracts the features of every application,then builds the mobile application feature library.By matching the collection data and the DPI feature library,it can identify the mobile application of users.After this,the thesis propose a users' network behavior recognition system based on dispersedidentification algorithm.Through this system,the thesis can make the recognition of the preference of mobile users' network behavior,including the preference of the time period and the content.Thirdly,the thesis builds a prediction model by using the decision tree algorithm to predict users' network behavior.By comparing the commonly used decision tree algorithms,this thesis chooses the C4.5 algorithm to build the decision tree.Also,the thesis gives a discretization algorithm to complete the process of data discretization.Then the thesis realizes the pruning and optimization of the prediction model by using the post pruning technology and obtains a relatively reliable prediction model of users' network behavior.