A further enhancement of improved-grey wolf optimization (IGWO) algorithm applied in music recommender systems

ABSTRACT: This study further enhances IGWO by addressing three key limitations: initialization bias, redundant iterations, and performance degradation in high-dimensional spaces. To tackle initialization bias, the Mersenne Twister (MT) Algorithm was integrated to randomly initialize wolves positions within the bounds, leading to a 23.11% to 38.64% improvement in optimization efficiency. To reduce redundant iterations, an Adaptive Counter Threshold was implemented effectively minimizing unnecessary computations by 66.58% to 81.43% while preserving the algorithm’s ability to reach optimal solutions. Additionally, parallelizing the algorithm’s Dimension Learnong-Based Hunting (DLH) process improved performance in high-dimensional scenarios, achieving an efficiency boost of 78.60% to 82.06%. Furthermore, on a heuristic approach, the algorithm was applied in the context of implicit music recommender systems wherein the results confirmed that the modifications did not alter accuracy, as reflected by metrics like Precision@K, MAP@K, NDCG@K, and AUC@K, all reporting an almost 0% change, ensuring that improvements focused solely on efficiency without degrading the solution quality. Moreover, execution time was decreased by 65.51% due to parallelization and adaptive counter threshold, which is an expected trade-off for handling high dimensional search spaces and redundant iterations were reduced by +67. Surprisingly, the results not only confirm the robustness of the approach but also establish a strong foundation for further advancements in optimization algorithms. By addressing key inefficiencies, this study contributes to the continuous improvement of IGWO, making it a m

Filipiniana