Improving Home Loan Predictions: A Fusion of PCA, Decision Tree and Random Forest Approaches

Loan default happens when a borrower receives funds from banks but fails to refund the loan. Debt is commonly utilized by individuals to acquire assets like homes and vehicles that would otherwise be unaffordable. As global economies become more interconnected and interdependent, the demand for capital has significantly increased. Loans can offer financial benefits when used wisely, but they also pose notable challenges. The last decade saw a surge in retail, small and medium-sized enterprises (SME), and commercial borrowers, with rising defaults impacting financial institutions. In this proposed research, the authors aim to develop a predictive model for identifying potential loan defaulters in the consumer lending sector. Dataset from Kaggle i.e. loan prediction based on customer behavior in .CSV format has been used which contains valuable data on previous client behavior, including demographic characteristics of each customer and a target variable indicating loan default or non-default. Principal Component Analysis (PCA) is used in our work which identifies principal components and helps in dropping features of less importance. Leveraging this data, our objective is to forecast the credit risk of new consumers and distinguish between higher-risk and lower-risk individuals. The result shows that 89% accuracy has been achieved using Random Forest (RF) classifier which is bit higher than the accuracy of Decision Tree (DT) classifier which is 88%. This analysis would assist financial institutions in making informed decisions when acquiring new customers, enabling them to effectively manage risk and optimize lending strategies.