Parameter Optimization for Support Vector Regression Based on Genetic Algorithm with Simplex Crossover Operator

摘要

The traditional parameter optimization methods of support vector regression mainly employ empirical way or grid search method, which have shortcomings of being guided by human experience and time-consuming. To avoid the randomness of crossover operator of ordinary genetic algorithms, in this paper a new parameter selection method based on novel genetic algorithm, combing the simplex crossover operator with great crossover probability search operator and used in the optimization of SVR parameters, is proposed, where simplex cross-search operation is used to determine the direction of cross-process, so as to enhance the optimizing speed and improve accuracy. Simulation experiments show that this improved genetic algorithm proposed in this paper has a significant improvement in convergence speed and prediction accuracy than those by using traditional genetic algorithm, and SVR predictive results rise markedly than those using the parameters gotten by traditional GA. All these have proved the effectiveness and feasibility of the proposed algorithm. The residuals deal with references, appendix, acknowledges, etc.