Collaborative Filtering with Hybrid Clustering Integrated Method to Address New-Item Cold-Start Problem

Abstract

Recommender Systems (RSs) are a valuable and practical tool to cope with information overload, as they help users to find interesting products in a large space of possible options. The Collaborative Filtering (CF) approach is probably the most used technique in RSs field due to several advantages as the ease of implementation, accuracy and diversity of recommendations. Despite being much favored over Content-Based (CB) techniques, it suffers from a major problem related to the lack of sufficient data for new-item cold-start problem, which affects recommendations’ quality. This paper is focused on resolving issues related to item-side in order to produce effective recommendations. To overcome the above problem, we use a powerful content clustering based on Hybrid Features Selection Method (HFSM), to get the maximum profit from the content. Then, it will be combined side by side to CF under a hybrid RS to improve its performance and handle new-item issue. We evaluate the proposed algorithm experimentally either in no cold-start situation or in a simulation of a new-item cold-start scenario. The conducted experiments show the ability of our hybrid recommender to deliver more accurate predictions for any item and its outperformance on the classical CF approach, which doesn’t work as usual especially in cold-start situations.