In this paper, in order to measure the customer churn rate in the Iranian banks, a new approach is introduced. First, using data preparation, the data is normalized. Then, using a k-medoids method, data clustering is formed. To assess the clustering performance, Davis-Bouldin index is applied. To detect the patterns available in the data, different Neural Network methods were exploited, Support Vector Machine (SVM), Radial Basis Function Neural Network, Generalized Regression Neural Network (GRNN) and Multilayer Perceptron (MLP) to name a few. To increase the precision in the model performance measurement, Hybrid of first and second error type cost function is considered. The results suggest that the MLP and SVM models show higher precision and lower cost function.