Cross Platform Recommendation System

Abstract

Mammoth amount of data is scattered across the Internet and surfing in this ocean of data is an endless task. To make user interaction smoother, recommendation systems are used to reduce the overload of information. In single domain recommendation systems, although the item space is humongous, users usually rate only few items and hence these systems focus on users having specific interests rather than relying on the wisdom of the majority. Hence, they suffer from a plethora of problems related to cold start, lack of data for new user/items etc. In cross platform recommender system (CPRS), acquired information from other domains is used to diminish such problems, i.e. the cold start problem faced by the target platform can be alleviated by transferring knowledge available in other platforms (known as the source domain or platform). This not only helps to establish relations between the items but also leads to higher user involvement and better suggestions. In this paper we explore the classic framework of item-item based collaborative filtering using which we propose a CPRS that maps the wisdom from our source platform Netflix to our target platform MovieLens. CPRS aims at achieving better accuracy and adds a new dimension in solving the cold start problem.