Application Of Data Ming In Mobile Telecom--The Implementation Of Jilin Mobile Churn Forecast Model

Nowadays, telecom companies have made great efforts in market under the confrontation of many domestic and international competitors. But the increase of custom losing and the decrease of average custom life cycle are seriously impacting the development of motion telecom. It is a real big problem in front of us that how to furthest reduce the churn ratio in the fierce competitions and diverse market requirements. An available answer is Data Mining.How to implement data mining technology in custom losing analysis system? The primary measure is to find some mathematic models represented the relationship between the sea of data and custom losing. These models can be established by custom attributes, servings attributes and custom consumption data, which from the former information we have gotten. When given the information of custom attributes and servings attributes and custom consumption, the probability of custom losing will be achieved through the model. And the market department could monitor the probability of custom losing by using it. If the probability was so high that approaching the threshold we have initialized, sales promotion must be carried into execution to hold back the losing trend. Therefore, data miming will solve the monitor problem drastically. Data mining is based on precise mathematics theory, which can bring scientifically artificial decision-making in custom-relationship-managing system of telecom companies.The implementation of losing prediction model in JiLin motion telecom is presented in this article. And then the actual application of data mining in losing prediction model is discussed according to this implementation. Therefore we lucubrate how to utilize data mining in our true-life better. In this paper we mainly focus on the establishing procedure of losing prediction model, especially in extraction of the original data, pretreating of data, model training and model validating. And the practical application of data mining make a very kind attempt among telecommunication trade.Clementine tool of SPSS company is employed in our losing prediction model with Oracle9i as the background data warehouse. The analysis of data, extraction of data, model training and model using are all implemented rapidly and efficiently by this tool.However, because of some objective reasons, such as time limited and lack of experience, there are still little flaw in our losing prediction model: Firstly, data are imperfectness. We can not get whole maturity information from the production system as the result of the actual state of operation. Secondly, the model overemphasize some direction targets leading the result of the digging with poor validity and timely. That means the customer perhaps has already leaved when we are getting the results during investigation. Future research of this topic is to improve veracity of the model and to find some other application terrains in motion telecom in order to extending the using area of data mining.