Machine Learning Approaches for Retail Bank Marketing Practice

Abstract

The retail business is usually one of the most essential bank divisions that directly handles personal customers. As restrictions for COVID-19 are gradually phasing out globally, the market share of retail banking is also rapidly growing, which made improving current bank marketing practices necessary. However, most banks lack an objective and reliable approach to locating target customers. Meanwhile, they also need affordable and efficient ways to process the massive data generated in information systems to predict the purchase intent of their client. This paper introduced two machine learning algorithms to solve the issues mentioned above, which were developed on a dataset derived from the Portuguese bank’s marketing campaign. Firstly, this study trained a Random Forest model to help select features to focus on when predicting client subscription on a term deposit. Secondly, an Artificial Neural Network model was implemented to evaluate the probability for an individual customer to purchase a bank product. The proposed Random Forest model successfully indicates that age and communication with banks are two of the most critical factors that drive customer subscription on term deposits. The proposed Artificial Neural Network also achieves an excellent accuracy of 91.82% in assessing whether a specific client will continue to purchase a specific financing product only at a loss of 18.88%. These results shed light on machine learning’s promising applicability in retail bank marketing, which will vastly improve related costs and efficiency.