Improving recommendation quality and performance of genetic-based recommender system

Abstract

The recommender system came to help the user in finding the required item in a short time by filtering the available choices. This paper addresses the problem of recommending items to users by presenting new three genetic-based recommender system (GARS+, GARS++ and HGARS). HGARS is a combination of GARS+ with GARS++. It is an enhanced version of the genetic-based recommender system that works without the being a hybrid model. In the proposed algorithms, the genetic algorithm is used to find the optimal similarity function. This function depends on a liner combination of values and weights. We experimentally prove that HGARS improves the accuracy by 16.1%, the recommendation quality by 17.2% and the performance by 40%.