Résumé: The growth of social networks and various web services (e-commerce, e-learning, e-health, etc.) urgently requires recommendation techniques to satisfy users. Providing high-quality recommendations with a minimum of common feedback is a major challenge for a new recommendation algorithm. In this paper, we use more informative modelling with a weighted bipartite network around the item entity. The idea is to exploit the item-user connectivity to extract hidden information. The problem is to predict non-existent links based on existing links by double projection forward and backward. The results are promising with the implementation of real data sets.

Langue: Anglais
Thème Informatique

Mots clés:
Prediction
Ranking
Recommender systems
Neighbourhood
Collaborative