基于机器学习的银行客户流失预测

It is undoubtedly the biggest bright spot of computer technology in the application field to find valuable knowledge from large and poorly readable data, understanding customer loyalty is an important part of any business, and predicting potential churn customers in advance can help companies enable early intervention processes to effectively prevent and control. In areas such as churn prediction, fraud identification, disease detection, and accident monitoring, imbalanced data occur from time to time, which may prevent most algorithms from learning correctly. This paper proposes a heterogeneous integration algorithm G_R_L_D based on Stacking technology, which is applied to the customer churn prediction model constructed by unbalanced data, the accuracy is 87% and the precision is79%,which is better than its base classifiers and most other algorithms, which helps to combat data imbalance.In addition, SMOTE-ENN sampling is used to effectively deal with the unbalanced characteristics of bank customer churn data, improve the recall while maintaining high precision, further improve the accuracyof the G_R_L_D and its base classifiers by more than 3%.