Description
Date depot: 15 octobre 2020
Titre: An EXplainable Artificial Intelligence Credit Rating System
Directrice de thèse:
Maria TROCAN (LISITE)
Encadrante :
Natalia DIAZ RODRIGUEZ (U2IS (EDX))
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé: Assessing credit risk is the key task for a risk underwriter. Reducing theirworkload allows them to focus on assessing the most critical cases. Artificialintelligence methods are applied to automate this task. However, most AItechniques are labelled as black-box models due to their lack of explainability. When assessing a company it is crucial to explain why a certaincompany has received a rating or not. This is the main reason why bankingand insurance companies do not use rating algorithms based on the latestAI models. In this thesis we will first develop a framework based on machine learning, deep learning and natural language processing to emulate theability of a credit risk analyst. The ultimate goal of this thesis is to betterunderstand the important factors key to draw a given decision in order tohave more informed decision support systems in complex situations wherefinancial data is not available.
Résumé dans une autre langue: .
Doctorant.e: El Qadi El Haouari Ayoub