Feature ranking for classification in credit scoring: A comparison between interpretable predictive techniques
Pier Giuseppe Giribone et al.
What the paper says
The objective of this study is to conduct a comparative analysis of three different fully explainable Machine Learning models, i.e., Logistic Regression (LR), Decision Tree (CART), and Logic Learning Machine (LLM), applied to a classification task on loan defaults. For each model, feature rankings have been retrieved to assess which characteristics had the greatest impact in predicting defaulted loans. For this study, a large, real-world dataset from a Peer-to-Peer lending platform was utilized. To deepen the analysis, a subdivision of the dataset was implemented, based on three hypothetical severity scenarios to classify the quality of loans. The findings show powerful predictive abilities for all three models, with AUC scores consistently beyond 0.95 in each scenario. The models confirmed distinctive strengths: the LLM showed robust overall accuracy and generated intelligible rules with high-coverage rates; the Decision Tree was the best in Precision by identifying nuanced risk profiles; and Logistic Regression provided the most reliable probability estimations.
Evidence weight
Balanced mode · F 0.40 / M 0.15 / V 0.05 / R 0.40
| F · citation impact | 0.50 × 0.4 = 0.20 |
| M · momentum | 0.50 × 0.15 = 0.07 |
| V · venue signal | 0.50 × 0.05 = 0.03 |
| R · text relevance † | 0.50 × 0.4 = 0.20 |
† Text relevance is estimated at 0.50 on the detail page — for your query’s actual relevance score, open this paper from a search result.