A Machine Learning Approach to Identify Customer Attrition for a Long Time Business Planning

Customer attrition is one of the most important studies to be made by any company. A company nowadays entirely relies on the customer. If a customer leaves the service of a particular company, it indicates some lack in the company's services. On the other hand, for a long-term business policy, a company or an organization needs to focus on those customers who are less likely to decline the company’s services in the future. The phenomenon of a customer leaving or reducing the services they had been taking is called customer attrition. For a long-term business, it’s essential to analyze the customer attrition rate and its reasons. In this study, a machine learning approach to customer attrition has been proposed, and an introduction of supervised learning in this area has been analyzed. Studies were made by applying different classifiers to predict customer attrition. The results show that the random forest model can predict customer attrition 96% accurately, better than any other model considered in this study. Not only in accuracy, but the random forest also performed pretty well in other statistics like sensitivity, specificity, and area under the curve. These results confirmed that though data was not balanced (massive difference in both classes for response variable), the random forest model has not picked the standard results only and made the prediction very effective.