Abstract :
As the churn of customers can result in the decline of business performance, preventing customer churn becomes an important task for e-commerce enterprises. This paper studies customer churn by using shopping data sets of e-commerce customers and machine learning technology. Firstly, a k-medoids clustering algorithm is adopted to cluster customers and divide them into three types. Then we adopt Logistic Regression, Support Vector Machines and Adaboost models to predict the churn of these three types of customers. The results show that the k-medoids algorithm can accurately identify churn and non-churn customers and that the Adaboost model boasts good prediction performance. The research results can guide enterprises to enhance customer relationship management and formulate appropriate marketing strategies and provide a critical theoretical basis for the research of customer churn prediction.
Keyword :
Customers churn, E-commerce industry, K-medoids clustering algorithm, Machine learning, Predicting technology.