Credit risk assessment plays a vital role in ensuring the financial stability and sustainability of lending institutions. Traditional credit scoring methods, primarily based on statistical models such as logistic regression, often fail to capture complex, non-linear relationships inherent in borrower data. This limitation results in reduced predictive performance, especially in dynamic financial environments. To address these challenges, this paper presents CreditShield AI, an explainable machine learning-based loan default risk prediction system designed to enhance accuracy, transparency, and reliability in credit decision-making.The proposed system leverages the Give Me Some Credit dataset, comprising over 150,000 borrower records with multiple financial and behavioral attributes. A structured data pipeline is developed, including data preprocessing, missing value imputation, outlier detection, feature scaling, and exploratory data analysis. Robust statistical techniques such as median imputation and percentile-based winsorization are applied to ensure data quality and consistency. Furthermore, the system adopts Robust Scaler normalization to mitigate the impact of extreme values.A key contribution of this work is the emphasis on explainable AI, ensuring that model predictions can be interpreted in compliance with financial regulatory standards. The system is designed to integrate advanced machine learning models and interpretability techniques such as SHAP and LIME in future stages. The proposed framework not only improves prediction capability but also promotes fairness, transparency, and responsible AI practices in financial risk management..
Mounika et al. (Thu,) studied this question.