Evolutionary Algorithm Based On Statistical Learning Method

Evolutionary algorithm is an efficient algorithm for solving global optimization problems. In has attracted much attention in both academic research and engineering applications. In the research and application process, it has been found that evolutionary algorithms may face some challenges, such as large amount of calculation, poor local search ability and blind searching. This paper uses statistical methods to improve the efficiency of evolutionary algorithm. The main contributions include:(1) We propose an evolutionary algorithm based on surrogate model. It constructs an approximate model with smaller amount of computation and less consumption in replace of the original analytical model with high-precision and instructs individual evolution to control the overall spending.(2) We propose an evolutionary algorithm based on orthogonal local search. In accordance with the orthogonal table in the experiment, it selects representative samples to conduct experiment. Besides, the efficiency of evolutionary algorithm has been improved greatly by grouping based on statistical analysis in orthogonal block.(3) We propose an evolutionary algorithm based on mean shift. It instructs evolution by the mean value and population density. Experiments have shown that three methods can be applied to solve different problems, which is beneficial for obtaining optimal solution in a faster and more accurate way.