Optimization of Recommender Systems Using Particle Swarms

Abstract

Background: Recommender systems are one of the most widely used technologies by electronic businesses and internet applications as part of their strategies to improve customer experiences and boost sales. Recommender systems aim to suggest content based on its characteristics and on user preferences. The best recommender systems are able to deliver recommendations in the shortest possible time and with the least possible number of errors, which is challenging when working with large volumes of data. Method: This article presents a novel technique to optimize recommender systems using particle swarm algorithms. The objective of the selected genetic algorithm is to find the best hyperparameters that minimize the difference between the expected values and those obtained by the recommender system. Results: The algorithm demonstrates viability given the results obtained, highlighting its simple implementation and the minimal and easily attainable computational resources necessary for its execution. Conclusions: It was possible to develop an algorithm using the most convenient properties of particle swarms in order to optimize recommender systems, thus achieving the ideal behavior for its implementation in the proposed scenario.