Abstract: Standard genetic algorithm in solving the traveling salesman problem (TSP) is not efficient since it is easy to fall into local optimal solution. To improve the efficiency of genetic algorithm, this paper presents an improved genetic algorithm. First, according to the diversity of individuals and the population distribution, the method to determine the cut-off algebraic of genetic algorithm is proposed. Second, by adding initialization information and improving cross-operator, the accuracy and convergence of the genetic algorithm could be improved.