A Genetic Based Recommender System

Abstract

In this paper we propose a genetic based recommender system. Our goal is to predict relevant items (films) from a huge space of data based on users ratings. We use MovieLens data set (which contains data on ratings made by users about films) to validate our approach. Our idea consists on finding for an active user, the most similar group of users using genetic algorithms. After that, we recommend items (films) based on prediction of the appreciation that the active user can give for each one. Several metrics have been used to prove the efficiency of our approach such Mean Absolute Error, precision and recall.