Predicting Financial Credit Risk Using Machine Learning Classification Models

Tan, Zi Yi (2025) Predicting Financial Credit Risk Using Machine Learning Classification Models. Final Year Project (Bachelor), Tunku Abdul Rahman University of Management and Technology.

Text RMM_Tan Zi Yi_Full Text.pdf Restricted to Registered users only Download (3MB)

Abstract

This project develops a credit risk prediction model by using several advanced machine learning algorithms to classify the risk into low and high risk based on historical financial data. It utilizes 6 models such as Logistic Regression, K-Nearest Neighbors (KNN), Random Forest, Naive Bayes, Support Vector Machine (SVM), and XGBoost. The study focuses on improving credit risk assessments and minimizing default risk for financial institutions by integrating traditional methods, such as the FICO score, with machine learning models. It highlights the importance of accurate credit risk prediction in making informed lending decisions and addresses the limitations of relying solely on conventional credit risk assessment techniques. By combining these approaches, the study aims to bridge the gap in current credit risk evaluation methods. The project emphasizes the significance of data cleaning, feature selection, and model evaluation in developing an accurate predictive model. Data preprocessing involved handling missing values using KNN imputation and mode imputation, followed by feature encoding and standardization. The class imbalance issue was tackled using SMOTE and Optuna for hyperparameter tuning, improving model efficiency over traditional methods. SHAP values were used to identify the most influential features, such as "Previous Default" and "Years at Current Job." Among the six models, XGBoost and Random Forest outperformed the others in terms of accuracy and F1-scores. XGBoost was selected as the best model due to its computational efficiency and was retrained with the top 10 features from SHAP values. The model's moderate performance was attributed to an imbalanced dataset and illogical data. The study concludes that deploying the XGBoost model can support decision-making in financial institutions. Keywords: Credit Risk Prediction, Machine Learning Algorithms, XGBoost, FICO Score, SMOTE, Optuna, SHAP Values, Class Imbalance, Model Evaluation.