A Genetic Algorithm with Best Combination Operator for the Traveling Salesman Problem

Abstract

In this research, we propose a genetic algorithm with best combination operator (BC x,y O) for the traveling salesman problem. The idea of best combination operator is to find the best combination of some disjoint sub-solutions (also the reverse of sub-solutions) from some known solutions. We use BC 2,1 O together with a genetic algorithm. The proposed genetic algorithm uses the swap mutation operator and elitism replacement with filtration for faster computational time. We compare the performances of GA (genetic algorithm without BC 2,1 O), IABC 2,1 O (iterative approach of BC 2,1 O), and GABC 2,1 O (ge-netic algorithm with BC 2,1 O). We have tested GA, IABC 2,1 O, and GABC 2,1 O three times and pick the best solution on 50 problems from TSPLIB. From those 50 problems, the average of the accuracy from GA, IABC 2,1 O, and GABC 2,1 O are 65.12%, 94.21%, and 99.82% respectively.