Speeding up recommender systems with meta-prototypes

Abstract

Recommender Systems use Information Filtering techniques to manage user preferences and provide the user with options, which will present greater possibility to satisfy them. Among these techniques, Content Based Filtering recommend new items by comparing them with a user profile, usually expressed as a set of items given by the user. This comparison is often performed using the k-NN method, which presents efficiency problems as the user profile grows. This paper presents an approach where each user profile is modeled by a meta-prototype and the comparison between an item and a profile is based on a suitable matching function. We show experimentally that our approach clearly outperforms the k-NN method while they presenting equal or even better prediction accuracy. The meta-prototype approach performs slightly worse than kd-tree speed up method but it exhibits a significant gain in prediction accuracy.