Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{196434,
        author = {Gayatri M. Chutake and Aryan S. Bute and Sanika S. Chaudhary and Adit D. Khair and Neha A. Kandalkar},
        title = {Machine Learning Analysis of Borrower Risk Progression in Financial Lending Systems},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {11},
        pages = {4056-4063},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=196434},
        abstract = {Credit risk assessment is one of the major steps in the loan approval process of banks and financial institutions, as incorrect credit approval decisions would result in loss in financial transactions and increase in non-performing assets. Classic methods for credit scoring is based on predefined rules or linear statistical methods, which limit their ability in modeling complex nonlinear applicant behavior found in real-world financial data. This work proposes a machine learning-based credit risk prediction framework that improves detection of potential loan defaulters.
Credit risk evaluation has been one of the most crucial steps within the lending process undertaken by banks and other financial institutions since inappropriate credit risk evaluation could lead to loss of finances as well as increased non-performing assets within the financial agencies. The conventional credit scoring models have been known to apply the use of either rule-based techniques or linear models. The constraint with the former shall be viewed from the standpoint that the ability of the conventional credit scoring models would be impaired by the non-linear behaviors depicted by the credit borrowers.
The proposed framework embeds a wide range of data pre-processing tasks, such as handling missing values, processing categorical variables, and addressing the problem of class imbalance using the Synthetic Minority Over-sampling Technique. Several classification algorithms-for example, Logistic Regression, Random Forest, and XGBoost-are tuned and tested for performance by using appropriate measures to handle imbalanced data. Optuna is also used for hyperparameter tuning.
The result of the experiment with a publicly available credit dataset shows that the tuned XGBoost model performs better compared to other models for correct classification, which is able to achieve 97.4% accuracy in terms of the F1-score. The above mention outcome shows that the developed approach does have the efficacy to strike a balance between the default detection and the incorrect loans being approved. The present study underlines the feasibility of using machine learning approaches in efficiently performing credit risk analysis.},
        keywords = {Credit Risk Prediction, Loan Default Detection, XGBoost, Optuna Optimization, SMOTE, Machine Learning},
        month = {April},
        }