基于分段核函数的支持向量机及其应用

Support vector machine based on segmented kernel function and its application

支持向量机兼顾训练误差和推广性能,已受到机器学习领域的高度重视,而核函数的性能是支持向量机研究中的关键问题。研究了几种常见核函数对支持向量机推广性能的影响,并利用全局核函数和局部核函数的性质,提出了一种新的分段核函数的支持向量机。数据集上的仿真结果表明,该核函数对应的支持向量机泛化能力优于传统核函数对应的支持向量机,具有较好的预测性能。

Considering both the training error and the generalization ability of support vector machine,it has been highly valued in the field of machine learning.The performance of kernel function is an important content in the study of support vector machine.The effect of several common kernel functions on generalization ability support vector machine is investigated.A new support vector machine based on piecewise kernel function is proposed according to properties of global and local kernel functions.The simulation results on data set show that the generalization ability of the support vector machine corresponding to kernel function is better than that of traditional kernel functions.It has a better predictive performance.