A Binary Fisherman Search Procedure for the 0/1 Knapsack Problem

The 0/1 Knapsack Problem is a widely studied problem of binary discrete optimization that has applications in a number of diverse, real-world applications. From existing algorithms for solving this problem, the three best meta-heuristics in the state of the art were selected, namely: Modified discrete shuffled frog-leaping, Soccer league competition, and Simplified binary harmony search. These algorithms were compared with a new binary algorithm of fisherman search procedure. In order to perform this comparison, instances of the 0/1 knapsack problem with low and medium dimensionality (100 and 200 items) were used. Medium instances have three levels of complexity (uncorrelated, weakly correlated, and strongly correlated). Used instances were generated previously by other authors (in order to avoid bias). The results were analyzed using three criteria: success rate in reaching the global optima, execution time, and number of fitness function evaluations. The results enable it to be seen that the proposed algorithm is the best meta-heuristic for solving these types of 0/1 knapsack problem.