Hybrid system for personalized recommendations

Abstract

Recommender systems are an important research area due to the various expansion possibilities that enhance the quality of the recommendations. A possible approach to improve the performance is to combine different recommendation techniques in a hybrid system that benefits from their complementarity and strengths. Our goal is to combine case-based reasoning and collaborative filtering to implement a scalable and domain-independent recommender system. The case-based reasoning engine will represent the core module of the system and will use the records of previous similar experiences to make suggestions or create new items to recommend. The collaborative filtering engine will be mainly used to adapt the recommendations to the preferences of the users and ensure a degree of diversity and novelty in the suggested items. Although the system needs to use the domain knowledge to generate personalized recommendations, it must be designed in a domain-independent way in order to make it adaptable to any application. In this paper, we present the global architecture of our hybrid recommender system and the ontology-based reasoning approach that will allow us to overcome the constraint of domain-independence.