OPTIMIZATION-DRIVEN ENHANCEMENTS IN RECOMMENDATION SYSTEMS: A COMPUTATIONAL APPROACH

Abstract

This study investigates the impact of optimization algorithms on enhancing recommendation systems that utilize a dual-score approach—contextual similarity and precise similarity—to improve ranking accuracy and relevance. The research aims to optimize the weighting of these similarity scores using five prominent optimization techniques: Particle Swarm Optimization, Simulated Annealing, Differential Evolution, Bayesian Optimization, and Genetic Algorithms. By fine-tuning these weight parameters, we assess the improvements in recommendation effectiveness and ranking accuracy. The experimental findings reveal that Differential Evolution outperforms other optimizers, yielding a 5.98% improvement in the F1-score and a 23.1% increase in ranking accuracy (measured using a custom Deviation Score metric). This research provides a practical framework for enhancing information retrieval and personalization in recommendation systems, with significant implications for business intelligence, e-commerce, and content platforms. By demonstrating the effectiveness of advanced optimization techniques in refining recommendation mechanisms, this study contributes to both the theoretical and applied domains of information management, and decision support systems.