核选择是支持向量机研究中的核心问题之一,不同的核函数将产生不同的分类效果。研究了核参数和误差惩罚参数对支持向量机推广能力的影响,然后根据局部核函数与全局核函数的各自优点,提出了一种新的自适应组合核函数,并将该核函数应用于支持向量机中。最后,利用该自适应核进行不同领域数据的实验,实验结果表明由该核函数建立的支持向量机具有更好的预测能力。 Core selection is one of the core problems in SVM research. Different kernel functions will produce different classification results. Based on the respective advantages of local kernel function and global kernel function, a new adaptive combinatorial kernel function is proposed, and the kernel function is applied to the support Vector machine. Finally, the adaptive kernel is used to test the data in different fields. The experimental results show that the SVM built by the kernel has better prediction ability.