The kernel function in the Support Vector Regression (SVR) machine has a great influence on the quality of model. Currently, however, every kernel has its advantages and disadvantages. Based on the fact that the regression accuracy and generalization performance of the SVR models depends on a proper setting of its parameters, the continuous multi-ant colony optimization (MACO) method based on gridding partition is applied in mixture-kernels SVR parameters. The cross-validation error is used as the fitness function of MACO. The optimal values in ant system were reflected by the 5 parameters of SVR. Simulation results show that the optimal selection approach based on MACO-SVR has good robustness and strong global search capability. The method used for the research of modeling in the traffic flow forecast obtains higher accuracy than the models constructed with the Genetic Algorithm.%核函数是支持向量回归机的重要部分,每种核函数都有其优势和不足.本文基于支持向量机回归机模型相关参数的选取原则,给出了一种具有混合核函数的支持向量机,以基于网格搜索的多蚁群算法为基础,给出了此类混合核函数支持向量回归机参数优化的一种新方法.该方法以最小化交叉验证误差为目标,对包括混合比例和各类核函数的参数在内的5个参数进行优化.仿真结果表明,与遗传算法相比,本方法在参数优化方面有良好的性能,建立的预测模型精度较高.
展开▼
