A New Hybrid-Enhanced Recommender System for Mitigating Cold Start Issues
Nouhaila Idrissi, Ahmed Zellou, Oumaima Hourrane, Zohra Bakkoury, El Habib Benlahmar · 2019
With the significant expanding flows of data circulating on the Internet, users are overwhelmed by an era of information explosion. Hence, Recommender Systems have become essential tools in helping each user find items that meet her needs and interests and filter out uninteresting ones. However, recommender systems suffer from a significant impediment that limits their performance which is known as the cold-start problem. This hurdle refers to the situation where the system is lacking valuable data to infer accurate neighbors for new users or items. By leveraging demographic, semantic, and collaborative recommendation techniques we propose, in this paper, a new hybrid recommender system to tackle cold-start challenges. A comprehensive set of experiments on the MovieLens dataset, using a broadly applied performance metric, verify that our proposed system outperforms state-of-the-art methods, in terms of recommendation accuracy, and cold-start mitigation.