Mobile Banking Customer Profitability Prediction using Machine Learning Techniques

Commercial Bank of Ethiopia collects and stores massive amounts of customer data. Currently, the company's traditional methods for understanding mobile banking consumers' behavior are incredibly inadequate. Identifying customers which are more likely potential to a service offering is an important issue. The purpose of this study is to design a machine learning-driven classification model to predict the profitability levels of the commercial bank of Ethiopia (CBE) mobile banking customers. To prepare the data for the analysis, the organization Provide us 12,000 rows of customer data to conduct this research. During the preprocessing that involved cleaning, transformation, encoding and normalization feature selection was carried out through a random forest algorithm to obtain the nine most significance attributes. Two supervised machine learning algorithm logistic regression (LR) and K-Nearest Neighbors (KNN), were then implemented. The models were trained on the 80% of the data and tested on the remaining 20%. Accuracy, precision, recall, F1-score, confusion matrices and ROC-AUC were the performance metrics used in assessing the performance of the model. Logistic regression performed slightly better compared to KNN, with accuracy at 98%. The study's overall findings were positive, which supports the potential use of machine learning to the bank industry, particularly, in customer prediction in Commercial Bank of Ethiopia.