| Because of the development of the airline industry,customer churn has become a central issue for every company.In order to tackle this problem,using data mining technology to establish customer churn predictive model for any airline company has its theoretical and practical significance.As data mining has been widely applied in customer churn prediction in telecommunication,finance and traditional industries,this methodology is becoming more and more developed.As analysis in customer churn in domestic airline market is still in its early stage,and many data mining techniques have not been implemented in predictive model,the above-mentioned analysis should be brought into focus.This article is based on the two main factors in customer relationship management: customer value and customer profit generating ability,with consideration of the features of airline customer behavior,to study the data mining application in airline customer churn.This predictive model,an improvement over the traditional RFM model,investigates an airline customer spending data,utilizes Python and SPPS data mining software to preprocess and analyze the data with SVM,Random Forest and LightGBM,and compares the three algorithms with GridSearchCV to determine the best models by evaluating different variables and AUC value.It draws the conclusion that with an AUC value of 0.92 and the accuracy of 0.93,the LightGBM has a better predicting outcome than the other two models.Combining the three customer churn predictive models using Voting Classifier,it shows that the values of precision,recall,F1-score and accuracy for the combination model are all 0.02 greater than those of any single model.At the end of the paper,the results will be applied to customer churn evaluation and lost customers will be divided into four categories by using K-means clustering algorithm so that specific retention strategy can be developed for different groups of customers. |
