Study of Imperialist Competitive Algorithm Hybrid Data Mining Techniques for Traveling Salesman Problems

Traveling salesman problem (TSP) is a classic combinational optimization problem (COPs) aiming at visiting all customers with the minimum distance. As in previous researches, blocks are defined as the highly fit schemata of a high quality solution in COPs. Therefore, if it is possible to obtain the high-quality blocks by analyzing the found feasible solutions, it could be greatly helpful for getting a high-quality solution. In this paper, a novel imperial competitive algorithm (ICA) hybrid data mining techniques (EKICA) is proposed. Entropy evaluation and K-means are the data mining techniques. Since entropy is used to measure the disorder of random variables, the high-quality link can be obtained between related cities to build the blocks by entropy evaluation. Next, ICA provides multiple self-evolutionary countries clustered by K-means to enhance the diversity of population. Through a series of experiments, the comparison result shows that the proposed algorithm can quickly provide blocks with high performance and combine solutions with high quality.