Research On Mobile Customer Defection Combination Prediction Model

With the expansion of the telecommunication industry,the user market tends to be saturated,and the direction of user growth of the three operators turns to each other's stock users.The implementation of the policy that change operators freely and the acceleration of 5G construction have forced operators to attach importance to and increase the investment retained by users.The size of the stock users is related to the actual interests of enterprises,and it is urgent to reduce customer churn.Previous studies have mostly focused on the construction of a single prediction model.With the wide application of machine learning technology,this thesis intends to explore the method of combining prediction model construction based on ensemble learning.Based on the idea of ensemble learning,this thesis creatively constructs a customer defection combination forecasting model based on Stacking method.With the help of the real business data from a certain mobile company,the business data are cleaned,expanded and filtered in turn.Histogram test and correlation coefficient test are used to select 12 most relevant characteristic attributes,by dividing data into training set and test set,oversampling method is used to process the unbalanced data of training set.Combination model is used to learn the data of training set.Meantime,the data of test set is predicted.Finally,the accuracy,coverage rate,F1 Score and ROC curve of the combined model are obtained by calculating the corresponding evaluation indexes.Compared to the related indexes of the single model,the thesis proves that the combination model established in this thesis has good performance in customer churn prediction.After identifying the target defection customers with the help of the model,it is necessary to carry out customer retention marketing in time in order to maintain the stock of users.Combining with the development trend of telecommunication market,this thesis designs the optimization of retention marketing process,and puts forward the optimization scheme of retention marking process,combining with the research conclusion of the article,the retention marketing strategy is putting forward,so as to improve the marketing efficiency and reduce marketing costs.