This project tackles the problem of customer churns. Based on different features we want to build a powerful predictive model that predicts if a customer for a bank with specific features (income, age) could leave the bank or not Firstly we build such an artificial neural network. Then try to rely on Machine learning classification models like support vector machines with depending on Radial basis function kernel (RBF) or Gaussian kernel to solve the nonlinearity Then we improve the model accuracy by applying cross-validation. Using the Grid Search techniques to detect the best values of hyper parameters that give the highest accuracy