Modèle Hybride IA–FinTech pour l'Évaluation et la Gestion des Risques : Application à Maison Souki-Pay

Abstract :
This article proposes a hybrid risk-assessment model combining statistical methods, supervised machine learning and generative artificial intelligence to enhance the detection, interpretation and management of anomalies in digital payment systems. The model addresses challenges faced by emerging FinTech platforms operating across multi-jurisdictional environments where regulatory expectations and behavioural patterns are highly variable.

The proposed architecture operates through a five-layer decision pipeline integrating (1) statistical scoring for baseline anomaly detection, (2) predictive classification using supervised ML, and (3) generative AI for interpretability, scenario simulation and regulatory justification. Applied to the case of Maison Souki-Pay, a digital payment solution connecting Europe and Morocco, the model demonstrates measurable improvements in robustness, reduction of false positives, transparency, and strengthened AML/CFT compliance. Expected outcomes—supported by recent academic benchmarks—include increases in recall and precision, significant reductions in operational noise, and enhanced decision explainability aligned with modern regulatory requirements.

This study contributes theoretically by integrating explainability, predictive modelling and compliant mechanisms within a unified risk framework; and practically by providing a transferrable model applicable to FinTech platforms in emerging markets.

Keywords : FinTech, AI hybrid, AML, risk management, digital payments, anomalies, Maison Souki-Pay.