Research On Forecasting Analysis Of Shifting Customer Churn Based On Data Mining

The article mainly studies the problem of customer churn prediction based on data mining algorithms in the telecommunications industry.How to more accurately use massive data to build a customer churn prediction model through data mining algorithms is the focus of this topic.That is to say,through the data mining technology in the massive historical data analysis to build the loss prediction model,analyze which users have the highest probability of loss,and provide support for marketing decisions according to the consumption behavior of the lost customers and the customer is other relevant factors.Based on the combination of theoretical research and empirical analysis,this paper constructs a loss prediction model based on Logistic regression algorithm and neural network algorithm based on the customer data of Unicom Majiapu Telephone Bureau.It elaborates on the attributes,data preparation,and Model building and the entire process of model evaluation and application.The paper uses the post-book indicators and graphical indicators to evaluate the model,and proves the feasibility and accuracy of the run-off prediction model using Logistic regression algorithm and neural network algorithm.Using the results of the model,the paper analyzes the probability of current network users flowing in the next month and discusses the attributes of monthly consumption,network duration,and number of calls for users with high probability of loss,summarizing the causes of customer churn and proposing corresponding customer retention.Measures.Finally,the article summarizes the main research work done,and really put forward ideas for future research directions.