Performance Evaluation of a Genetic Algorithm Based on Parameter Configuration in the 0-1 Knapsack Problem

Experimentation with different configurations of the genetic algorithm revealed that a low mutation probability, combined with an appropriate tournament size, significantly improves the algorithm s performance. This is reflected in the reduction of the average GAP and the maximization of the achieved objective value, thereby establishing quantifiable benchmark parameters for efficient implementations of the 0-1 knapsack problem. The novelty of the study lies in providing concrete guidelines for practical applications, identifying consistent relationships between parameter configuration and algorithm behavior. While genetic algorithms are effective in exploring large search spaces in NP-complete problems, this work demonstrates that they require precise calibration to avoid premature convergence. It was found that the mutation probability and tournament size directly influence the algorithm s exploratory and exploitative dynamics. In general, combinations with low mutation and intermediate tournament sizes offered a good balance between solution quality and stability, whereas extreme values (mutation of 0.20 or tournament size k = 2/k = 4) increased performance variability, especially in larger-scale problems. These findings highlight the algorithm s sensitivity to these parameters and underscore the importance of careful tuning to preserve its effectiveness in complex scenarios.