RISK ASSESSMENT OF MORTGAGE LOANS USING MACHINE LEARNING TECHNIQUES

ABSTRACT This project investigates the application of machine learning techniques in assessing mortgage loan risk to enhance decision-making processes within financial institutions. The research evaluates two different models which are K-nearest model (KNN) and Support vector machines (SVM) using online mortgage data. Key metrics such as accuracy, precision, recall, and F1-Score are employed to measure the models' performance in predicting loan defaults and identifying risky loans. Findings reveal that Support vector machine (SVM) emerges as the most effective model, achieving an accuracy of 93% and demonstrating robustness in capturing complex relationships among loan variables. Logistic regression and SVM also exhibit competitive performance but show limitations in handling nonlinear relationships inherent in mortgage data. K-nearest neighbors (KNN), while interpret-able, show lower accuracy and precision, highlighting challenges in capturing nuanced risk factors. The study concludes with recommendations for financial institutions to adopt advanced machine learning models for mortgage loan risk assessment, policymakers to support responsible AI adoption in credit risk management, and future research to explore enhancements in model interpret-ability and integration of alternative data sources. Addressing these recommendations can significantly enhance the accuracy and efficiency of mortgage loan risk assessment, thereby improving portfolio management and reducing financial risks associated with mortgage lending operations.