Research On Predicting Customer 's Potential Transfer Client Based On Data Mining

Massive data has been collected with the development of database technologies. In order to break through the situation of data rich but information poor, data mining is becoming increasingly important for many companies. Due to the popularity of mobile phone usage, people tend to depend on mobile phones and replace mobile phones frequently. Therefore, the study on predicting the replacement of mobile phone is significant.Under this background, this paper concentrates on the methods for predicting the replacement of mobile phone based on data mining technology. Firstly, we apply Hadoop to analyze the behavior data which rely on the project of precision marketing requirement of phones terminal of a company in X province. According to the CRISP_DM model, high quality data could be obtained by data understanding, data cleaning, and data transformation. Secondly, we build predictive models by determination tree C5.0, neural networks, and logistic regression algorithm, and compare and evaluate these models. The experimental result suggests the determination tree C5.0is more appropriate for prediction of user mobile replacement. Finally, this paper analyzes the model for developing the user market, launching terminal precision marketing, improving business simple recommendation success rate, behavior monitoring and other aspects of the user terminal applications.Confronted with the real-world application, this paper studies on predicting end-user trends based on data mining. This work is important to decision and marketing, and is significant to the relevant study. However, this study also can be improved on some aspects, such as evaluation criteria, business knowledge, and methods on data processing. And these aspects would be the future works.