Precision Marketing System Of Terminal Based On Data Mining

With the increasingly fierce competition of telecom operators,the traditional marketing methods fail to meet the requirements of new situation.Operators will have stronger competitiveness when they accurate grasp the needs of users and accurate market to the users.All the time the telecommunication industry produces large amounts of business data,which implies a lot of information and valuable.So how to mine the useful information from these massive data for precision marketing is an urgent need.This paper mainly through the data mining technology to discover the potential replacement users,and achieve the precision marketing of the terminal to users,the main work is as follows:Choose the useful attribute replacement model from the huge,multi-dimensional business data based on information gain ratio,and integrated into a complete data set.Cleaning and filtering of the data by the related technology of data preprocessing,and processing some abnormal data.Finally construct an ideal data set,which will be used to establish the model of terminal replacement.A sampling strategy is presented in view of the imbalanced data.First of all,introduce the sample distance which is weighted by information gain ratio,and filter out the boundary samples of categories based on the KNN algorithm.Then propose an approximate linear time clustering algorithm,and clustering the remaining big class samples.Use the stratified sampling according to the cluster to make the date set balance.Firstly,users are classified by C4.5 decision tree algorithm.Then combine with the idea of boosting to optimize the C4.5 algorithm by using Random Forest to improve the precision and robustness.Finally,a two stage classifiers is presented to filter the potential target user again and improve the precision.