Research on the Influence of Crossover Probability and Mutation Probability in GA-SVM Model
Xiujun Wu · 2019
SVM(Support vector machine), a common seen algorithm in machine learning, is well known for its good performance in classification. However, support vector machine is quite sensitive to the value of kernel function parameters. As the times require, some methods employed for the selection of heuristic parameter emerge, and they include particle swarm optimization, genetic algorithm, et al. These methods are mainly applied to the selection of SVM kernel parameters and thus promote the good performance of the SVM model. The process of GA (genetic algorithm) is composed of selection, crossover, and mutation. The probability of crossover and mutation is usually fixed, and this model with fixed probability fails to adapt to the changes within the population. Therefore, a proper probability model is put forward to optimize this kind of problem through the employment of comparative study. By using the new probability model, the performance of SVM classification optimized by genetic algorithm is improved, and the parameters of precision rate, recall rate and F1 value of the new model are improved as well to some extent.