Abstract:
Financial institutions like banks are suffering from challenges of credit risk; one of the reasons is lack of effective credit risk evaluation method. Hibret bank is one of competent private commercial bank in Ethiopia, However, the bank does not have enough computer-based systems to support the credit risk evaluation process whereas it is made by human experts which are fully manual. Credit analysts may commit some mistakes due to fatigue, stress, or other negative factors. The absolute purpose of this study is to develop a credit risk assessment model using machine learning approach. To achieve the aim of the study, a total of 27042 credit records were collected from Hibret Bank centralized database. The collected data were preprocessed by handling missing values, outlier detection, feature selection, normalization, and data balancing. Five machine learning algorithms previously stated as a benchmark by different researchers were selected to develop a credit risk assessment model. Then, a total of five experiments were conducted. In each experiment the selected algorithms hyperparameters were optimized by grid search cv algorithm. In order to select the best-performed model from each algorithm several train test splits were applied. Finally, to recommend the best model, the selected models were compared with five performance measurement metrics. The outcome of the comparison showed that the model developed by XGBoost and Random Forest algorithms generated promising results than other algorithms model in terms of Accuracy, Precision, Recall, AUC, and F1-score. A model developed by Random Forest, and XGBoost algorithms generated approximately 98% accuracy and AUC. They also produced the same 98% macro average of Precision, Recall, and F1-score. Then, through proposed machine learning models’ a prototype was developed to show the functionality and reliability of the study. After that user acceptance test was performed to gather the users’ feedback about the prototype. Finally, the study concluded that, Hibret Bank and other banks can aid their credit approval process and minimize defaulter risk by including the proposed credit risk assessment models in their system.
