An explainable and risk-sensitive machine learning model for loan approval in digital banking | |
| Author | Nguyen Thanh Quang |
| Call Number | AIT PJPR PMDS no.25-05 |
| Subject(s) | Bank loans--Data processing Risk management--Data processing Internet banking Machine learning |
| Note | A project report submitted in partial fulfillment of the requirements for the Degree of Master of Science (Professional) in Data Science and Artificial Intelligence Applications |
| Publisher | Asian Institute of Technology |
| Abstract | In the era of rapid digital banking transformation, integrating machine learning into credit decision systems has become increasingly important to improve efficiency, accuracy, transparency and risk management. This study designs, develops and evaluates an explainable machine learning model to support loan approval decisions in a digital banking context. The research pursues two main objectives. First, it aims to construct a predictive model that combines financial and contextual loan attributes in order to enhance loan default prediction accuracy. Second, it applies explainable artificial intelligence techniques, specifically SHAP Shapley Additive Explanations, to improve transparency, interpretability and user trust in model outputs.The study employs a publicly available Small Business Administration loan dataset containing 899,164 records and 27 variables. Comprehensive data preprocessing and exploratory data analysis were conducted to remove irrelevant and post default information, handle missing values and identify key predictive variables. A Random Forest classifier was selected for its robustness with structured data. On an independent test set, the model achieved strong predictive performance with accuracy of about 0.92, ROC AUC of about 0.955 and balanced precision and recall for both Paid in Full and Default classes. SHAP analysis identified Term and ApprovalFY as the most influential predictors, followed by loan size variables, urban or rural location, franchise status and selected sector and state indicators. The findings demonstrate that combining predictive modeling with SHAP based explainability can produce models that are both accurate and transparent for banking decision makers. The study advances the integration of explainable machine learning in credit risk modeling and provides a practical framework for explainable loan approval that can be adapted to digital banking operations in a responsible that is consistent with regulatory expectations. |
| Year | 2025 |
| Type | Project |
| School | School of Engineering and Technology |
| Department | Department of Information and Communications Technologies (DICT) |
| Academic Program/FoS | Professional Master in Data Science and Artificial Intelligence Applications (PMDS) |
| Chairperson(s) | Chantri Polprasert; |
| Examination Committee(s) | Chaklam Silpasuwanchai;Vatcharaporn Esichaikul; |
| Degree | Master of Science (Professional) - Asian Institute of Technology, 2025 |