Research On The Loss Of Customers Based On Logistic Regression And Decision Tree Algorithm

In the increasingly severe market environment,more and more enterprises regard old customers as the most important resources of the company,and actively adopt effective marketing policies to maximize the retention of old customers to avoid large losses.This paper mainly studies the customer's churn prediction problem,and regards the customer churn analysis as a two-category problem,that is,the churn state can be divided into two states: churn and non-loss.Secondly,the common classification algorithms are introduced in detail: Logistic regression algorithm and decision tree algorithm,and these two algorithms are applied to the empirical analysis of the data of a member of airlines.In the empirical analysis stage,the ROSE algorithm is first used to process the unbalanced data set,and then the variables used to construct the model are determined.Then the logistic regression algorithm and the decision tree algorithm are used to establish the customer churn prediction model,and the ideal results are obtained.Based on the logistic regression model,the proportion of customer churn is calculated to be 0.245,while the objective loss ratio of customers is 0.270.The decision tree model created has a decision rate of 2.84% on the training sample set and 5.18% on the test sample set,and the classification accuracy is higher.Therefore,the research results of this paper have certain reference value for airlines.