Building Prediction Model For Telecom 4G Customers Based On Data Mining Technology

December 2013, China has formally entered the 4G era. At the same time the intense competition in telecommunications operators proceeding. With the wide application of data mining techniques and generated telecom big data, people now pay more and more attention on how to make full use of the existing experience and data mining technology to mining telcom big data. In 4G era, with the demands of the telecommunications customer relationship management (CRM), expanding the market share of 4G customers is having a great practical significance and economic benefits for telecom operators using the existing data warehouse and data mining technology. The research study of 4G customer prediction is based on such a background in order to expand the number of carrier’s 4G customer in this paper.The dataset we used is obtained from a telecom company in the paper. The main goal is established and implemented a high accuracy 4G customer prediction model in this paper. We designed and implemented of telecom 4G customers predictive models based on the principles of data mining and CRISP-DM methodology. First data was pre-processed before establishing the model, including data integration, data cleaning, data convert, data explore and so on. Second, three prediction models which used decision tree, Logistics regression and SVM algorithm were established for 4G customers. Through training and comparing those models, we chose the decision tree model as the best model. In the deployment phase of the model, each customer was given a score by the prediction model, the customers with high score are the potential 4G customers, telcom company could take measure to expand the scale of 4G customers. At last, this paper builded a Hadoop cluster with 9 nodes, and implemented C4.5 decision tree algorithm parallelization, effective solved the problem that one computer can’t handle larger amount of data. verified the Hadoop platform efficiency in handling large data telecommunications to be significantly higher than the stand-alone.In this paper, the theory and practice of data mining projects was combed. We established 4G telecom customers prediction model The results showed that the application of the established model is reasonable in line with the actual needs, could provide valuable information for the decision-makers, the prediction model is useful for telecom operators to expand 4G customers.