Data mining i banksektoren

Abstract

This study shows how data mining can be used in the banking sector to reduce churn among mortgage customers. A churned customer is defined as a customer who have terminated their mortgage agreement. Our contribution to reduce customer churn is divided into two key actions: predicting customers who will churn and key insights on those who churn. In a competitive environment, a key to success is keeping your profitable customers. By applying machine learning on data from a major bank in Norway, we have shown that it is possible to predict customers who churn with a precision of 77%. After experimenting with several models we found that XGBoost turned out to be the best fit for this problem. The customers who churn are younger and have been a customer for a shorter period of time compared to those who do not churn. In addition, they are also less wealthy and use the bank’s services less in contrast to the customers who do not churn. By combining predictions with insight we believe that customers in risk of terminating their agreement can be identified at an early stage and retained with the proper measures, which in return will increase the bank’s profitability.