Font Size: a A A

Aviation Customer Value Assessment And Churn Prediction Model Based On Data Mining Analysis

Posted on:2023-02-15Degree:MasterType:Thesis
Country:ChinaCandidate:G W ZhangFull Text:PDF
GTID:2568306833987139Subject:Applied Statistics
Abstract/Summary:PDF Full Text Request
In recent years,with the progress of the times and the development of technology,the number of airlines has increased rapidly,which has also led to fierce competition within the aviation industry,of which the most critical issue falls on customer resources.Since the total number of customer resources is limited,if the airline cannot well grasp its own resources and the introduction of new passenger sources,it will face the risk of being eliminated by competitors in the same industry.In order to deal with this problem,airlines must be able to effectively and accurately identify the value of customers,and timely control the situation of customer churn.This paper takes the behavior data of an airline customer as the research object,and explores and analyzes the value classification of customers and the loss of customers.The main research contents are as follows:First,the research on the classification of customer value,this paper selects the indicators from the existing value and potential value of customers,and then uses the improved spectral clustering algorithm to classify customers into four categories,and then based on the RFM expansion model and subjective and objective comparative analysis The four categories of customers are defined as: high value customer,important to maintain customers,important to retain customers and low value customers.Secondly,for the research on customer churn prediction,this paper first uses multiple regression to define three types of customer churn: non-churn,quasi-churn and churn customers,and then uses five models based on decision trees(CART decision tree,random forest,GBDT,XGBoost,Light GBM)to predict and analyze customer churn,and compare models based on various evaluation indicators.In contrast,XGBoost has the best prediction effect and each indicator value is above 0.9Finally,a comprehensive analysis of customer classification and churn is conducted.On the one hand,the churn of each type of valuable customers is statistically analyzed,and it is found that the churn rate of high-value customers is low,and the corresponding lowvalue customer churn rate is relatively high;The churn factors of customers with similar value are explored,and it is found that the factors affecting the churn of each type of customers are not completely consistentThe purpose of this paper is to provide certain theoretical suggestions for airlines to formulate marketing strategies through the above research and analysis,so that they can use more reasonable and effective strategies to grasp customers and allocate limited resources to different types of customers,to maximize corporate profits.
Keywords/Search Tags:customer value classification, churn prediction, improved spectral clustering, decision trees, ensemble learning
PDF Full Text Request
Related items