MODIFICATION OF CROSSOVER OPERATOR ON GA APPLICATION FOR TSP

Genetic Algorithm (GA) has been widely used in many fields of optimization; one of them is Traveling Salesman Problem (TSP). GA in the TSP is primarily used in cases involving a lot of vertices, which is not possible to enumerate the shortest route. One of stages in GA is crossover operation to generate offspring’s chromosome based on parent’s. Example of some crossover operators in GA for TSP are Partially Mapped Crossover (PMX), Order Crossover (OX), Cycle Crossover (CX), and some others. However on constructing the route, they are not considering length of the route to maximize its fitness. The use of random numbers on constructing the route likely produces offspring (a new route) that is not better than its parent. Sequence of nodes in the route affects the length of the route. To minimize uncertainty, then the crossover operation should consider a method to arrange the chromosomes. This article studied incorporating two methods into crossover stage, in order to ensure the offspring has good fitness. Methods to be combined with algorithms are commonly used in the route searching; those are Nearest Neighbor algorithm, and Sequential Insertion. Operators used are CSI (Crossover combined with Sequential Insertion) and CNN (Crossover combined with Nearest Neighbor), named after the method used. Those operators are compared with PMX operator on test using benchmark data from TSPLIB on some independent executions. The tests showed that CSI are better than two other and length of its route was relatively equal to optimal length recorded. Keywords: Genetic Algorithm, Traveling Salesman Problem, Crossover operato