Domain Abstraction of Highly Correlated Pairs to Recommend in the Long Tail

Abstract

Among difficulties encountered by modern shopping recommenders is the long tail shape of sold items also related to cold-start issues. Various  approaches  including   content-based recommendations attempt to overcome this problem that has serious impact on the accuracy of recommendations especially when new products are continuously added to the catalogue. This paper investigates the use of an  algorithm to search for highly correlated pairs between abstractions of items.  The  advantage of this approach is evaluated on the basis of real data showing better results compared to an approach onlybased on the concrete pairs of items. Using rigorous protocols such as Given-n, experimental results show significant improvement in both the recommendation accuracy and the recommendation of products in the long tail.