多核SVM对传感器动态建模的研究

Study on Sensor's Modeling Using Multiple Kernel Support Vector Machine

针对传感器的动态特性,提出了一种基于多核最小二乘支持向量机对传感器进行动态建模的方法.通过不同核函数的线性加权组合构造新的等价核,由于构造的等价核函数兼具了全局核函数和局部核函数的优点,从而降低了建模精度对核函数及其参数的依赖性.在理论上详细介绍了多核最小二乘支持向量机回归参数和模型输出值的求解方法.通过仿真实验验证了该方法的可行性.将该方法与标准的最小二乘支持向量机方法进行比较,证明了该方法在一定噪声存在的情况下,具有良好的抗噪性和较高的建模精度.

According to the dynamic characteristic of the sensor,a kind of sensor dynamic modeling method is proposed based on MK-LSSVM（multiple kernel least squares support vector machine）. A new equivalent kernel is built by linear weighted combination of multiple kernels to reduce the dependence of modeling accuracy on kernel function and parameters, for the structure of the new kernel has the advantages of both the global kernel function and the local kernel function. The solution of regression parameters and MK- LSSVM output are given in theory. Simulations show that the proposed method is feasible. It is compared with ISSVM （least .squares support vector machine）method, and the result shows that it gets better accuracy in model- ing and has stronger anti-noi capability in the presence of noise.