An Efficiency Boost for Genetic Algorithms: Initializing the GA with the Iterative Approximate Method for Optimizing the Traveling Salesman Problem—Experimental Insights

Abstract

The genetic algorithm (GA) is a well-known metaheuristic approach for dealing with complex problems with a wide search space. In genetic algorithms (GAs), the quality of individuals in the initial population is important in determining the final optimal solution. The classic GA using the random population seeding technique is effective and straightforward, but the generated population may contain individuals with low fitness, delaying convergence to the best solution. On the other side, heuristic population seeding strategies provide the advantages of producing individuals with high fitness and encouraging rapid convergence to the optimal solution. Using background information on the problem being solved, researchers have developed several population seeding approaches. In this paper, to enhance the genetic algorithm efficiency, we propose a new method for the initial population seeding based on a greedy approach. The proposed method starts by adding four extreme cities to the route, creating a loop, and then adding each city to the route through a greedy strategy that calculates the cost of adding every city to different locations along the route. This method identifies the best position to place the city as well as the best city to add. Employing local constant permutations improves the resultant route even more. Together with the suggested approach, we examine the GA’s effectiveness while employing conventional population seeding methods like nearest-neighbor, regression-based, and random seeding. Utilizing some of the well-known Traveling Salesman Problem (TSP) examples from the TSPLIB, the standard library for TSPs, tests were conducted. In terms of the error rate, average convergence, and time, the experimental results demonstrate that the GA that employs the suggested population seeding technique performs better than other GAs that use conventional population seeding strategies.