A COLLABORATIVE FILTERING BASED RECOMMENDER SYSTEM ALLEVIATING THE MOST COMMON CHALLENGES

Abstract

Recommender Systems (RS) is a subclass of information filtering system that seeks to predict the "rating" or "preference" that a user would give to an item. These are the Systems for recommending items (e.g. books, movies, web pages, newsgroup messages) to users based on examples of their preferences. With increasing information availability online, it has become extremely challenging to build accurate Recommender Systems. But at the same time, they are extremely important for businesses all around the globe. Hence it becomes vital to study these systems and improve their accuracy. Recommender Systems are generally classified into Content based and Collaborative filtering(CF) based RS. Content based RS focus on the properties of the items while CF based RS estimates user/item similarity. Here, we discuss RS’ with special emphasis on CF based techniques which are further classified as user/item based CF Algorithms. The Issues and Challenges faced by CF based RS makes it an active area of research with many new variants and improvements being proposed in the past years. This work proposes a recommender system that alleviates three challenges simultaneously i.e. shilling attacks, sparsity problem and cold start problem that are most common in this domain. As a result, the proposed system proves to be more effective as compared to the conventional algorithm. Practically, this system can be utilized to produce quality recommendations to its users, as a result of which, the profitability can be maximized. The dataset used for doing the analysis and building the system is the popular MovieLens dataset. The work also presents a study of the Collaborative Filtering Improvements along with analysis of the important aspect of each improvement. This study can be utilized by any researcher working in this direction to have a complete idea of the past and the latest improvements in this area.