摘要
利用支持向量回归机(SVR)算法,对在ε-insensitive和Quadratic两种不同损失函数下的3种核函数分别进行了研究与讨论。在样本数据学习中,发现归一化后的数据明显优于原始数据。在量化指标λ下对6种组合进行了分析研究,给出了参数的取值范围,对支持向量机在类似工程上的应用具有借鉴价值。
Three kinds of kernel functions were researched based on two different loss functions (ε - insensitive & quadratic). It was found that normalized data waS better than original data during using support vector regression (SVR) algorithm. Parameters range waS obtained baSed on index λ, providing reference for similar engineering application.
出处
《四川建筑科学研究》
北大核心
2008年第2期148-151,共4页
Sichuan Building Science