基于TS-GA的LS-SVM参数优选

LS-SVM Parameter Optimized Selection Based on TS-GA

将禁忌搜索和遗传算法相结合,提出一种改进的最小二乘支持向量机(LS-SVM)参数优选方法。利用自适应遗传算法进行全局搜索,使用禁忌搜索进行局部寻优,由此提高求解速度和解的精度。采用某冶炼厂净化工段的现场数据建立模型进行仿真实验,结果表明,该方法能使LS-SVM模型具有较好的泛化能力,模型精度满足工艺要求。

This paper proposes an improved Least Squares Support Vector Machine（LS-SVM） parameter optimized selection method by combining Tabu Search（TS） and Genetic Algorithm（GA）. Self-adaptive GA is used to search the global space and TS is used for searching the local area, so that the efficiency and precision of the solution are improved. A prediction model based on the method is established and simulated by the field data from the purification process in a smelt factory. Simulation results show that the method makes LS-SVM model have good generalization performance and high precision which can satisfy the technology requirement.