Optimizing Knapsack Problem with Improved SHLO Variations

The Simple Human Learning Optimization (SHLO) algorithm, drawing inspiration from human learning mechanisms, is a robust metaheuristic. This study introduces three tailored variations of the SHLO algorithm for optimizing the 0/1 Knapsack Problem. While these variants utilize the same SHLO operators for learning, their distinctiveness lies in how they generate new solutions, specifically in the selection of learning operators and bits for updating. To assess their efficacy, comprehensive tests were conducted using four benchmark datasets for the 0/1 Knapsack Problem. The results, encompassing 42 instances from three datasets, reveal that both SHLO and its proposed variations yield optimal solutions for small instances of the problem. Notably, for datasets 2 and 3, the performance of SHLO variations 2 and 3 outpaces that of the Harmony Search Algorithm and the Flower Pollination Algorithm. In particular, Variation 3 demonstrates superior performance compared to SHLO and variations 1 and 2 concerning optimal solution quality, success rate, convergence speed, and execution time. This makes Variation 3 notably more efficient than other approaches for both small and large instances of the 0/1 Knapsack Problem. Impressively, Variation 3 exhibits a remarkable 14x speed improvement over SHLO for large datasets.