Abstract Résumé
This thesis presents EMiGRe, a novel framework that explains why certain items are missing from recommendation lists through two main approaches: Remove Mode (identifying actions preventing recommendations) and Add Mode (suggesting new interactions). Using multiple heuristics and tested on real-world datasets, EMiGRe effectively provides Why-Not Explanations, improving transparency and user trust in recommendation systems while maintaining computational efficiency. The framework is demonstrated through an interactive application that showcases its practical implementation and benefits.
Try EMiGRe Demo Essayez la démo EMiGRe
Explore our interactive graph-based recommender system. Understand why certain items are not recommended and discover new connections in your data.
Thesis Supervision Encadrement de thèse
Dimitris Kotzinos
Thesis Director
Aikaterini Tzompanaki
Thesis Supervisor
Publications Publications
-
Attolou, H.-M. (2021). Why-Not Explanations for Recommenders. In Actes de la conférence BDA, vol. 99.
Access Paper -
Attolou, H.-M., Tzompanaki, K., Stefanidis, K., & Kotzinos, D. (2024). Why-Not Explainable Graph Recommender. In Proceedings of the IEEE 40th International Conference on Data Engineering (ICDE 2024).
Access Paper -
Attolou, H.-M., Tzompanaki, K., Stefanidis, K., & Kotzinos, D. (2024). EMiGRe: Unveiling Why Your Recommendations are Not What You Expect. In Proceedings of the International Conference on Web Engineering (ICWE 2024).
Access Paper