支持向量机是以结构风险最小化的原则建立费用估算模型,在非线性小样本数据预测中能够到达较好的效果,但支持向量机的核函数选择还没有统一的理论。当样本数据特征不同时选择相同的核函数并不合理,且单核、混核以及混核权重的确定还没有深入研究。提出考虑参数影响的最优核函数支持向量机的设备费用预测方法,该方法以预测精度为优化目标,分析支持向量机基本参数及核参数对核函数选择的影响,确定最优的混合核函数。实例分析表明考虑支持向量机参数影响的最优核函数与其他核函数比较有更高的预测精度。 SVM is a cost estimation model based on the principle of minimizing the structural risk, which can achieve good results in the prediction of nonlinear small sample data. However, there is no unified theory to choose kernel function of SVM. It is not reasonable to choose the same kernel function when the characteristics of the sample data are different, and the determination of single kernel, mixed kernel and mixed kernel weight has not been further studied. This paper proposes a prediction method for equipment cost of optimal kernel function SVM considering parameter influence. The method takes prediction accuracy as optimization objective, analyzes the influence of support vector machine basic parameters and kernel parameters on kernel function selection, and determines optimal mixed kernel function . The case study shows that the optimal kernel function considering the influence of SVM parameters has higher prediction accuracy than other kernel functions.