Research On Customer Churn Prediction Based On Semi-supervised Learning And Its Application In Logistics Enterprises

With the increasing competition of logistics industry,the quality of services provided by the logistics enterprises is becoming higher and higher,the differences are becoming smaller and smaller,and the customers have more choices.Therefore,the customer churn is a problem that every logistics enterprise has to face.What's more,due to the fact that the customer churn will bring huge losses to logistics enterprises,it has received a widespread attention in both industry and academia.At present,there are a lot of customer churn prediction methods which can be divided into traditional classification methods and imbalance data classification methods.Although these two methods have achieved good prediction results,they require a large number of labeled customer data to learn.It is expensive to obtain such a large number of labeled customer data in the logistics enterprise customer churn prediction.Therefore,it is urgent to solve the problem of how to use the labeled customer data and the unlabeled customer data to predict customer churn.Semi-supervised learning method is an effective way to learn from labeled data and unlabeled data.Therefore,this research proposed a new customer churn prediction method based on Co-training method to solve the above problem.Firstly,this research analyzed the research status and the existing problems of the research in the customer churn prediction and semi-supervised learning.Secondly,this research analyzed the basic theory of the customer churn prediction,the customer churn prediction in the logistics enterprises and the data mining,including the overview of CRM,the concept and types of the customer churn,the process of the customer churn prediction,the characteristics of CRM in logistics enterprises,the main causes of customer churn in logistics enterprises,two types of errors in the customer churn prediction,the overview of data mining and two types of classification methods in data mining,and so on.Then,based on the above analysis,this research applied the Co-training method to the customer churn prediction,and constructed a customer churn prediction model to solve the problems of imbalance data and semi-supervised learning.Finally,this research developed a prototype system of the customer churn prediction for logistics enterprises.The effectiveness and practicability of the customer churn prediction model were verified by the practical application of the model.The experimental results indicated that the model proposed in this research achieved better results in the practical application.On the one hand,this research applied Co-training method to the customer churn prediction and constructed a customer churn prediction model.The theoretical research of customer churn prediction was enriched and improved.On the other hand,this research applied the customer churn prediction model to logistics enterprises,and developed a prototype system of the customer churn prediction for logistics enterprises.It provided an effective method to predict the customer churn for logistics enterprises.