Studying the Effect of Data Structures on the Efficiency of Collaborative Filtering Systems

Abstract

Recommender systems is an active research area where the major focus has been on how to improve the quality of generated recommendations, but less attention has been paid on how to do it in an efficient way. This aspect is increasingly important because the information to be considered by recommender systems is growing exponentially. In this paper we study how different data structures affect the performance of these systems. Our results with two public datasets provide relevant insights regarding the optimal data structures in terms of memory and time usages. Specifically, we show that classical data structures like Binary Search Trees and Red-Black Trees can beat more complex and popular alternatives like Hash Tables.