Machine Learning Based Suspicion of Customer Detention in Banking with Diverse Solver Neighbors and Kernels

In the current moving technological business sector, the amount spent for attaching the new customer is highly expensive and time consuming process than adopting some methods to hold and retain the existing customers. So the business sector is in need to make a research on with holding the existing customers by using the current technology. The methods to make the retention of the existing customers with high reliablility are a challenging task. With this view, we focus on predicting the customer churn for the banking application. This paper uses the customer churn bank modeling data set extracted from UCI Machine Learning Repository. The anaconda Navigator IDE along with Spyder is used for implementing the Python code. Our contribution is folded is folded in three ways. First, the data preprocessing is done and the relationship between the attributes are identified. Second, the data set is reduced with the principal component analysis to form the 2 component feature reduced dataset. Third, the raw dataset and 2 component PCA reduced dataset is fitted to various solvers of logistic regression classifiers and the performance is analyzed with the confusion matrix. Fourth, the raw dataset and 2 component PCA reduced dataset is fitted to various neighboring algorithms of K-Nearest Neighbors classifiers and the performance is analyzed with the confusion matrix. Fifth, the raw dataset and 2 component PCA reduced dataset is fitted to various kernels of Support Vector Machine classifiers and the performance is analyzed with the confusion matrix. The implementation is carried out with python code using Anaconda Navigator. Experimental results shows that, the rbf kernel of Support vector machine classifier is effective with the accuracy of 85.8% before applying PCA and accuracy of 80.9% after applying PCA compared to other classifiers.