Artificial Intelligence and Machine Learning Models for Credit Risk Prediction in Morocco

Abstract

This study investigates the application of artificial intelligence and machine learning models for credit risk prediction using a real-world dataset collected from a Moroccan credit institution. The data reflect clients' demographic, socio-economic, and financial characteristics, as well as behavioral information related to credit history and interactions with the institution. Six supervised learning models Logistic Regression, Random Forest, Support Vector Machine, Decision Tree, k-Nearest Neighbors, and Naïve Bayes were trained and evaluated using key performance metrics such as accuracy, recall, F1-score, AUC, and average precision. Results indicate that Random Forest outperformed all other models, demonstrating strong discriminative power and robustness to class imbalance, while Logistic Regression provided consistent and interpretable baseline performance. These findings highlight the effectiveness of ensemble and margin-based methods in credit scoring applications and emphasize the importance of feature importance analysis for transparent and informed decision-making in financial risk assessment.