Machine learning for Customer Behavior Prediction on Banking Product

Abstract

This article aims to study the parameters for customer to make decision in bank product via telemarketing and to apply machine learning for customer behavior model of banking product. The technique in this study is the Cross-Industry Standard Process. There are 6 processes that are business understanding, data understanding, data preparation, model development, evaluation, and deployment. The data consisted of 41,188 customers and 20 parameters. To do the data preparation, feature selection i.e., data transfer, data reduction, data cleaning using Knime a tool. There are 3 algorithms for customer behavior model of banking product that are Decision tree, Random Forest and Tree ensemble. Algorithm performance testing by using accuracy, sensitivity, and specificity for the best algorithm. The results can demonstrate that the top three parameters impacting decision making are the length of time spent talking on the phone with customers, the number of days elapsed since the customer was last contacted from a previous campaign, and the results of previous marketing campaigns. The most accuracy algorithm for customer behavior prediction in bank product is Tree ensemble with accuracy 90.94% comparing with Decision tree and Random Forest tree with accuracy 87.54% and 90.29%, respectively. To do the data visualization using Power BI that is very beneficial for planning and leads to higher sales. On-going research, to apply this methodology with bank product marketing in Thailand.