Research And Application Of Predatory Search Genetic Algorithm

Genetic algorithm is an adaptive, global optimization, probabilistic search algorithm, it has simple and universal, not dependent on the specific areas, strong robustness and other characteristics, so it has been used in many areas widely. Some problems were found in many applications of genetic algorithm, the main two issues are premature convergence and weak capacity of local search due to low solution precision and efficiency. Although there are many improvements, most of the improvements are targeted, limited, less universal and the algorithms taking into account two issues is rare. So how to solve these problems and improve the optimization efficiency of genetic algorithm is very necessary for the practical application of it.Informed by accessing to relevant information, premature and weak capacity of local search problems are caused by this contradiction of global search and local search, the predatory search strategy can balance this contradiction well. The predatory search strategy is a strategy and has not specific search method, so, we use the idea of predatory search strategy to improve genetic algorithm. In global search, a genetic strategy based on information entropy is presented, that is, selection pressure, crossover probability and mutation probability are changed adaptively according to the difference between current actual population entropy and current expected population entropy, in order to achieve the purpose of adjusting the population diversity and improving the premature phenomenon; in the local search , the strategies of narrowing rang of parameters arithmetic crossover and Gaussian mutation is used, in order to achieve the objective of improving local search ability .In order to test the improved algorithm, the paper selected four complex and different characteristics mathematical functions. Through testing the four functions, the results show that the method can avoid premature and improve search quality and search efficiency. Then predatory search genetic algorithm is presented to determine the parameters of the support vector machine, and regard the three parameters of support vector machine (width parameterσ, insensitive parameterε, penalty parameterC ) as the decision variables, the optimal objective in reality application are used as objective to optimize by PSGA. Finally, the PSGA-SVM are applied to establish the modeling of the glutamic acid fermentation process, the simulation results show that the method can optimize the effective parameters of support vector machine and glutamic acid fermentation process model has a good prediction.