Unknown but interesting recommendation using social penetration

Abstract

With the recent rise in popularity of social networks, millions of users have included social network Web sites into their daily lives. Traditional social recommendation systems suggest items with high popularity, familiarity, and similarity to users. Such recommendation processes might encounter two problems: (1) if the recommended item is very popular, the target user may already be familiar with it; (2) the target user may not be interested in items recommended by users familiar to them. To improve upon traditional recommendation systems, we propose a SPUBI algorithm to discover unknown but interesting items for users using social penetration phenomenon. SPUBI considers the popularity of items, familiarity of other users, similarity of users, users interests and categories, and item freshness to obtain a social penetration score, which are used to generate a recommendation list to the target user. Experimental results demonstrate that the proposed SPUBI algorithm can provide a satisfactory recommendation list while discovering unknown but interesting items effectively.