Statistical analysis of different recommender systems: An application perspective

Abstract

Applications of recommender systems vary from simple friend recommendations based on similar interests, to complex product recommendation based on buying and social patterns. The vast majority of recommender system models developed over the years have enabled researchers to utilize them in a very application specific manner. For instance, threshold-based recommender models are used for IoT based applications, while pattern recognition-based recommender models are used for analyzing social and ecommerce buying patterns for the users. But, this vast variety of recommender models have also created a dilemma amongst researchers, as to which specific recommender should be used for their specific application? In order to resolve this dilemma, in this paper we have reviewed various state-of-the-art recommender models and suggested applications for which they are most suited. These suggestions are based on the models’ applicability, the quality of service provided by the model, and the accuracy of recommendation. This paper also recommends certain improvements that can be made in specific models in order to make them better in terms of application-specific applicability.