Investigation on Customer Churn Prediction Using Machine Learning Techniques

Abstract

Churn prediction is generally considered a major use case in banking business. We assumed that the scenario of customers of an international bank determined to quit from the services offered by the bank. The bank decided to investigate this scenario of high rate of customer quitting the services offered by the bank. The dataset used for experimentation contains 10 K records, and we employ it to examine and find the potential customers who are more likely to quit the value additions provided by the bank in near future. The approaches exploited in this paper are supervised classification models using the various state-of-the-art machine learning algorithms; the various classification models have been leant on above-said huge volume of historical banking data so as to make predictions on the upcoming customers in order to recognize the potential churn. The dataset consists of 13 attributes and a class label. We found that comparatively, the best accuracy has been obtained using Naïve Bayes model with 86.29%. The churn prediction techniques could be effective utilized for the applications in telecommunication sector in order to identify the customers who will be changing port to other network soon and also in human resource department to find out the employees who will be leaving the organization in near future, which would enable the organization to plan for hiring of new employees well in advance.