Customer churn model based on complementarity measure and random forest

How to prevent the loss of bank customers, especially the loss of high-quality customers, is a great concern of banks, for which an accurate churn prediction model is of great importance. The accuracy of the integrated classifier is better than that of a single classifier. Random forest is a kind of ensemble learning. Traditional random forest uses all decision trees for voting. Some poor decision trees will reduce the overall performance of random forests. To improve the performance of traditional random forest, the random forest based on complementarity measure is proposed. The decision trees in the forest are pruned using complementarity measure. We use the proposed method to predict bank customer churn. Firstly, affinity propagation clustering (AP clustering) algorithm is used for attribute selection. Then the improved random forest method is used to establish an early warning model of customer churn. Compared with the general churn prediction model, this model has higher accuracy.