Methods for speeding up recommender system computations using a graph database

Abstract

Recommender systems are commonly used for Internet-based activities to assist users in making decisions on what items to select. One very common use of recommender systems is in electronic commerce purchases. The need for recommender systems in electronic commerce is due to the vast amounts of items to choose from. Due to this vast amount of items, generation of recommendations for recommender systems is a computationally intensive activity. This paper reports on studies that were conducted to investigate methods for speeding up the computations for generating recommendations when the data that is used to generate recommendations is stored in a graph database. The proposed methods involve the pre-computation and storage of values that are used in the generation of recommendations. This leads to a speed-up of the computations for generating recommendations.