Analysis Of Telecom Customer Churn Based On Data Mining

With the rapid development of data mining technology,data mining has been widely used in many fields.The application of data mining technology is widely used in telecommunication industries,especially in the telecommunication customer churn analysis.The main research works and innovations of this include the following aspects:(1)Considering the difference between cost and benefit of algorithm,a new index of evaluating the customer churn prediction model is proposed regarding specific scenes,i.e.,Maximum Profit for Subdivision Customer churn(MPSC).Compared with the traditional classification algorithm index,the target of new index which guides for the company to choose the customer churn classification algorithm is not to maximize the accuracy of the customer churn prediction model,but to maximize the actual benefits of the enterprise.Moreover,considering the different important degree of different value customers,the customer has been subdivided and it is used in the new index.(2)An improved decision forest classification algorithm i.e.,Serial Decision Forest(SDF)is proposed,the core of which is based on the new index to guide the direction of its construction.Relative traditional combination algorithms such as Random,Forest Bagging,Adboost,the establishment process of the new algorithm is not only taking the precision of the algorithm as a guide,but also considering the cost and benefit of the algorithm.It eventually leads the accuracy of the new algorithm is not the best,but the new index performance of this algorithm is the best.(3)Pointing to the customer churn problem,this thesis puts forward a new framework of customer churn analysis system.The framework of the system is closely related to the evaluation index,the prediction model and the application of the model.Considering from the overall point of the customer churn problem and in order to maximize the economic benefits of the enterprise,the framework of the system is built to constantly optimize the evaluation indicators and forecasting models.Thus,it is to be better applying the prediction model and be good solved the problem of customer churn.In addition,the churn customers of the output of prediction model will be subdivided to be better guiding the enterprise to make better retention strategies.