Tests with two heuristic search strategies in the simulated annealing algorithm for an inventory problem

The problem in the results when implemented metaheuristics techniques to solve instances of the problem of multi-product replenishment is the deterioration of the quality of the solution. It was noted that the implementations have focused on the parameters of the algorithm paying little attention to the strategy to access the neighboring solution. In this paper, we study experimentally an implementation of Simulated Annealing algorithm exploring several combinations of parameters and also, two schemes for obtaining the neighbor solution by comparison with the RAND algorithm. A 2-factorial experimental design was constructed and the study was conducted over 2.000 randomly generated instances, the results show that under the same combinations of parameters, perturbing a variable at a time provides poor results, however, to access the neighbor solution taking into account families of products provides better results and Simulated Annealing algorithm behaves robust against the increase in the size of the problem.