Abstract: Modified decimal MIMIC algorithm is a kind of discrete estimation of distribution algorithm, which is based on binary MIMIC algorithm and convenient to solve traveling salesman problem. Considering the drawbacks of the MIMIC algorithm while solving larger-scale TSP, this paper improved the encoding mode and probability model, proposed new individual strategy, introduced greedy algorithm at the initial phase of the probability matrix, and adopted crossover operator, mutation operator, etc. during the process of evolution, employed dynamic adjusted method to determine the population size. These modifications gurantec the population diversity even in small population and for larger-scale TSP. Experiment results show that problem scale, solution quality and speed of the optimization are improved significantly.

Key words: MIMIC algorithm, TSP, Estimation of distribution algorithm, Probability matrix