Research On Predictive Analysis Of User State Transition Based On Mobile Communication Data

Mobile communication technology and its related service business are gradually combined with Internet technology and Internet of things technology driven by the computer technology.It's growth of data scale has been promoted.Big data analysis for user data application can traces its behavior and provide it with high qualified and reliable reference for maintenance and marketing.It has been an important research subject to effectively predict the potential users of churn and to reduce user's churn rate at the same time to increase the efficiency for retaining and recall the former users.Based on existing researches,this paper focuses on the prediction of users' state transition in mobile communication data.It mainly includes the following three aspects:Firstly,uer state transitions found with the mobile communication data may be chosen as the start point.Forming features set by the method of retrieving user behavior data within a certain unit time prior to the state transition accordingly.The extracted feature will then be used for the machine learning and form the measurement indicator for the feature simultaneously.Experimental results show that the machine learning model has a good recognition effect on user state transition.Next,to solve the problem in the researches in the past,this thesis presents the method that serializes users' real-time data by taking day as the granularity and on which the analysis of time series data would be applied.Users' data would be sampled and analyzed by using the sliding windows.Then the churn inclination among the users would be predicted in advance by collecting and sorting out their antecedent behavior occurring before their churn.The test experiment was conducted with real mobile communication records of social users and comparing with the traditional modeling which takes month as the granularity as its statistical item,the timeliness and predicting competency of the method presented in this article has been verified both theoretically and by the result of the test.Finally,through the user static data and dynamic data were clustered respectively before conclusion summary,the research on mining similar user behaviors was further refined and deepened witch based on previous work.In the static data clustering,the discrete and continuous features are distinguished.In the dynamic data,the clustering is carried out by the behavior of windows.All clustering works have taken the objective measurement as the basis for the parameters selection.