Credit Scoring Model Development: Taking Advantage of Credit History

This paper introduces a novel credit scoring model that leverages extensive credit history data and advanced ma-chine learning techniques to enhance the accuracy of assessing creditworthiness. Our approach diverges from traditional credit scoring methods by integrating a logistic regression model that captures intricate patterns in borrower data, enabling more precise predictions of loan repayment probabilities. The results demonstrate that our model effectively predicts borrower defaults with a high level of accuracy, achieving an overall accuracy rate of 90.45% and a strong area under the ROC curve (ROC-AUC) score of 0.94. The model exhibits a precision of 85% and recall of 83% for identifying defaulters, reflecting a balanced ability to detect actual defaults while minimizing false positives. The specificity of the model for non-defaulters is 94%, indicating a robust performance in correctly identifying borrowers who will repay their loans. This study underscores the potential of machine learning to refine decision-making processes in financial lending by offering a quantitative tool that enhances informed credit decisions, thereby promoting financial inclusion while effectively managing risks.