Author(s): Talawar Basavaraj and Talawar AS

Abstract: In the rapidly evolving financial landscape, the precision of credit scoring models has become a cornerstone of risk management and profitability for lending institutions. This study embarks on a pioneering exploration into the predictive capabilities of different classifier models within the realm of credit scoring. By delving into the intricate dynamics of behavioral and collection scoring, we reveal how these models not only assess creditworthiness but also preemptively identify potential defaulters with remarkable accuracy. Our analysis transcends traditional methodologies, integrating advanced machine learning techniques to enhance predictive power and decision-making efficacy. The findings illuminate the nuanced interplay between borrower characteristics and default probabilities, offering unprecedented insights into the optimization of credit risk strategies. This work not only augments the toolkit of financial institutions but also sets a new benchmark in the scientific discourse on credit scoring. Through this endeavor, we aim to elevate the standards of credit risk assessment, ensuring that the allocation of credit is both judicious and equitable. Our unique approach promises to transform how financial institutions navigate the complexities of creditworthiness, ultimately fostering a more robust and resilient financial ecosystem. Based on the analysis, we conclude that, the loan percent income is most significant variable and home ownership, the other is the least significant. Among statistical models, linear discriminant analysis (LDA), logistic regression (LR) and Gaussian Naïve Bayes (GNB), GNB has higher receiver operating characteristic curve (ROC) and area under the curve (AUC). The support vector classifier (SVC) is considered as the best classifier with consistent evaluation metrics across train and test split.

DOI: 10.22271/maths.2025.v10.i8c.2140

Pages: 196-205 | Views: 794 | Downloads: 17

Download Full Article: Click Here