Etudiant : EL-MAKHLOUFY SALAHEDDINE

Filière : LF Sciences Mathématiques et Informatique

Encadrant : Pr. NFAOUI EL HABIB

Annèe : 2025

Résumé : Sequential recommendation systems are designed to predict the next item a user might interact with based on their previous actions. This project implements the BERT4Rec model, which uses advanced transformer-based techniques to understand the sequence and context of user item interactions. By applying this model to the MovieLens dataset, we aim to provide accurate and personalized recommendations by effectively capturing the temporal and contextual relationships in user behavior. The BERT4Rec model leverages a bidirectional self-attention mechanism, enabling it to analyze and learn from user interaction sequences comprehensively. This advanced modeling approach helps improve recommendation accuracy, particularly in scenarios where understanding the order of interactions is critical. To ensure practical application, we have deployed the model into a web-based application, making it accessible and user-friendly. This integration allows users to interact with the recommendation system seamlessly and receive recommendations in real-time. By combining state-of-the-art sequential modeling techniques with an intuitive deployment strategy, this project enhances the usability and efficiency of recommendation systems.