A Content-Based User Model Generation and Optimization Approach for Movie Recommendation

Contents

Personalization has become a powerful approach for constructing more precise and easy to use information search and recommendation systems. The quality of the personalization is heavily dependent on the accuracy of the user models created by the system and it is very important to incorporate content information of the working domain in order to enrich these models. This paper proposes a content based movie recommendation algorithm to make recommendations for the target user through building content based user models from collaborative-based user models and characteristics of the movie domain. Constructed user models are fine-tuned through highly liked, highly not liked, and don't care flags. The user models are presented to the users in terms of the most important features and dimensions in their profile. This makes explicit the users' implicit and unknown preferences of the movie domain. The system is evaluated and the results are presented using decision-support metrics.