Application Of Customer Churn Forecasting Based On PCA-NB Algorithm

In recent years,in the fierce market competition,enterprise's marketing strategy has changed their attention from the product to the the customer.especially under the premise of the development of a new customer costs rising,maintain existing customers,predicted in advance the old customer churn intention,and take the appropriate measures to prevent the loss becomes more significant.Therefore,how to accurately and effectively predict a customer's churn intention then develop scientific and reasonable customer retention strategies,for enterprises to reduce the losses and increase profits,has become the focus o of many enterprises.Naive Bayes classifier is one of the most simple and effective classification models in the practical application of classification model.But the conditional independence assumption makes it ignore the correlation between the attribute variables,and this affects the classification accuracy.In order to solve this problem,we use principal component analysis before the naive Bayesian classification modeling,established o the naive Bayesian classification model based on principal component analysis.It meets the naive Bayes assumption of conditional independence,give full play to the advantage of the naive Bayes classifier,with non principal component analysis of naive Bayesian classification model are compared to show that the PCA-NB algorithm effectively improve the rate of classification accuracy.Finally applied the principal component analysis treated naive Bayes algorithm to the telecommunication enterprise customer loss prediction problem,in order to improve classification accuracy of simply using the naive Bayesian classification model for customer churn prediction,and according to the different types of customers take different measures to retain customer and prevent the unnecessary loss of enterprise,saving the development cost of new customers for the enterprise.