Uncovering Significance of Feature-Selection in Recommender System Models

Abstract

Recently, recommendation systems have shown significant improvement in terms of return on investment (ROI). The fundamental idea of a recommender system is to assist user to shortlist item(s) based on his/her choices and preferences. Recommender system model takes in to consideration set of item features to generate relevant recommendations. Recommender system is applicable on items from multiple business domains. For each item, different feature-set may turn out to be useful to obtain accurate recommendation results. This paper discusses the compatibility of textual and visual movie features through various scenarios. We analyze recommendation results obtained by using distinct movie features. This study suggests possible extensions and also helps in taking decision with respect to the selection of appropriate item features.