支持向量回归与常用近似模型的非线性预测性能比较研究

为降低计算成本和提高优化效率,工程实践中广泛应用近似模型拟合或预测非线性系统响应是研究的前沿与热点。引入支持向量回归方法,通过典型数值案例对比分析其与多项式响应面、kriging和径向基函数的非线性预测性能。利用箱线图直观的证明支持向量回归的非线性预测性能明显优于多项式响应面、kriging和径向基函数,且支持向量回归的预测精度对DOE的依赖性最弱,体现出良好的稳健性能,进一步验证了支持向量回归适用于非线性系统响应的近似建模。

Comparative Study of Nonlinear Prediction Capacity of Support Vector Regression and Typical Surrogate Models

Metamodel is usually employed to approximate the nonlinear response of structural performance to reduce the computational cost and improve the optimization efficiency in practice.Support vector regression （SVR） is introduced and the comparative study of approximation accuracy with polynomial response surface,kriging,radial basis function,and SVR is performed on typical test functions using statistic methods.Through illustration of boxplots on prediction error metrics,it is demonstrated that the performance of SVR is better than others.