Fine‐tuning a Tabu Search Algorithm with Statistical Tests

Tabu Search is a metaheuristic that has proven to be very effective for solving various types of combinatorial optimization problems. To achieve the best results with a tabu search algorithm, significant benefits can sometimes be gained by determining preferred values for certain search parameters such as tabu tenures, move selection probabilities, the timing and structure of elite solution recovery for intensification, etc. In this paper, we present and implement some new ideas for fine‐tuning a tabu search algorithm using statistical tests. Although the focus of this work is to improve a particular tabu search algorithm developed for solving a telecommunications network design problem, the implications are quite general. The same ideas and procedures can easily be adapted and applied to other tabu search algorithms as well.