基于加权精度的ε-SVR组合参数优化

针对支持向量机参数的选取还没有一套完整的理论支撑,提出以加权精度来评价某一组参数的预测效果。通过循环交叉验证和全局变步长的方法,对最优参数进行搜索。考虑参数间的相互影响,研究参数的组合形式对精度的影响,确定参数的最优组合形式。实例分析表明,参数的最优组合能够提高支持向量机对设备费用的预测精度。

Combined parameter optimization for ε-SVR based on weighted accuracy

Aiming at the lack of integrity theories for choosing the parameters of the support vector regression machine (SVR), the combination accuracy is proposed to evaluate the estimated effect. The methods of circulation crisscross verification and variable step length are used to search the optimal parameters. The interaction of the parameters is considered. This paper researches the influence of the combined form of parameters on the estimated accuracy, and assures the optimized combined form of the parameters. The result indicates the optimized combined form of the parameters can improve the expenses estimated accuracy.