Comparison of Artificial Neural Network and SPSS Model in Predicting Customers Churn of Iran’s Insurance Industry

This paper has three aims. The first aim is Prediction of Iran’s insurance industry customers churn using data mining techniques and SPSS software. In this way, on the one place, in data mining part, multi-layer perceptron ANN with 8 neurons in hidden layer has applied and the best performance of this network appears in epoch 10. Plus, the structural model of network is added. On the other place, Regression test has used in order to prediction customer churn by SPSS. As a result, the performance of predicting regression and neural network model are compared. Second, the difference between target and output values is presented based on the Root Mean Square Error (RMSE) and Mean Square Error (MSE) codes in Matlab. Indeed, MSE have less value rather than RMSE. Finally, in order to prevent the waste of financial and human resources, the K-Means method has used for clustering customers into two groups of churn and non-churn.