Research on Interpretable Loan Approval Identification Using Multi-Dimensional Features
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Loan approval is a pivotal component of financial risk control. Current research often relies on black-box prediction models and lacks in-depth exploration of multidimensional features and interpretability, resulting in shortcomings in model interpretability and robustness. To address this, this paper proposes an Interpretable Loan Approval Identification Model based on Multidimensional Features (ILA-MDF). The ILA-MDF model is constructed using the CatBoost algorithm and is compared with benchmark models such as Random Forest in experimental evaluations. The results indicate that the proposed model performs optimally across six metrics—Accuracy, Precision, Recall, F1-score, AUC, and MCC—achieving values of 93.33%, 87.19%, 82.03%, 84.53%, 97.77%, and 80.34%, respectively. Furthermore, the SHAP framework is introduced to analyze the key factors influencing model decisions. Interpretability analysis reveals that the probability of loan approval increases significantly when the loan interest rate exceeds 14% or when the previous loan default indicator is 0. The ILA-MDF model proposed in this paper demonstrates significant advantages in both predictive performance and decision interpretability, providing a reference for formulating loan approval and risk prevention strategies.
Fields of Study & Subjects
Financial Distress and Bankruptcy Prediction
Explainable Artificial Intelligence (XAI)
Microfinance and Financial Inclusion
Publication details
- Journal / source
- International Journal of Engineering Research and Science
- Volume
- 12
- Issue
- 3
- Pages
- 30–44
- Publication date
- 30 March 2026
- Language
- EN