Bank Term Deposit Service Patronage Forecasting using Machine Learning

Abstract

Term deposit is one of the financial services offered by the bank. Effective bank marketing campaign to forecast possible customer to engage in personal term deposit marketing interaction is vital because it’s hard to stand out, considering that all banks offer similar products. Trailing to this, this study proposed the use of machine learning algorithms to develop a bank term deposit patronage forecasting models which have the ability to study the characteristics of customers to identify potential term deposit customers. Random Forest and Xtreme Gradient Boosting algorithms along with Portuguese institution marketing campaign dataset were used to develop bank term deposit service patronage forecasting model. The data balancing algorithm utilized is the Synthetic Minority Over-sampling Technique and Edited Nearest Neighbors (SMOTE-ENN) and feature selection was conducted using Information Gain. The Random Forest model achieved an accuracy of 95%, recall of 92% and f1 scores of 94%. Xtreme Gradient Boosting model achieved an accuracy of 97%, recall of 97% and f1 scores of 97%. The results of the experiment revealed the Xtreme Gradient Boosting emerged as the best model