CredWise combines credit risk predictions explanations and policy insights
CredWise: A Controlled Agentic Decision-Intelligence Framework for Explainable and Auditable Credit-Risk Assessment
Computation and Language
Summary
Banks need to know not just whether a loan applicant is risky but also why they are risky and how that fits with other rules and evidence. The authors created CredWise, a tool that brings together computer predictions, explanations, and policy documents into one system. It uses a machine learning model trained on real loan data and shows stable, clear explanations for decisions. CredWise also uses smart search and controlled workflows to help human reviewers make better decisions. This system supports but does not replace human judgment.
What this means in practice
- •For bank credit risk teams: Support loan officers by combining risk scores with clear reasons and related policy evidence in a single workflow for better review decisions.
- •For financial compliance teams: Provide accurate retrieval of relevant policy sections linked to credit decisions to aid auditing and regulatory compliance checks.
Authors
Aakash Kumar Tiwari
Abstract
Credit-risk prediction is important in banking, but a prediction alone does not explain why an applicant is risky or how it should be combined with other evidence. This paper presents CredWise, a decision-support framework that integrates credit-risk prediction, probability calibration, explainable artificial intelligence, policy retrieval, SQL analytics, and controlled agent-based workflows. An XGBoost model is trained on Lending Club data (1,345,310 loans, 18 features) using a temporal split: 2007--2016 for training, 2017 for validation, and 2018 for testing. On the 2018 test set, the calibrated model achieved a ROC-AUC of 0.7109, PR-AUC of 0.2993, F1-score of 0.3714, and accuracy of 65.44\%. Calibration reduced the Brier score from 0.2157 to 0.1273 and the expected calibration error from 0.2862 to 0.0585. SHAP explanations were temporally stable, with a Spearman correlation of 0.9959 between 2017 and 2018 feature rankings. On 28 labeled queries covering nine policy sections, FAISS achieved the best Hit@1 (0.929) and MRR (0.964), while all three retrieval methods reached Hit@5 = 1.0. Agent routing achieved 95.6\% accuracy (43 of 45 cases), and the SQL benchmark scored 1.0 on exact-match, execution-success, and result-match across six cases. These results show that CredWise can combine predictions, explanations, policy evidence, and structured analytics in one controlled workflow. It is an academic research prototype, and final decisions remain with a human reviewer.