多项式光滑的支撑向量机

数据分类问题是数据挖掘研究的一个热门课题．它是根据对数据样本集合建模，得到最优的分类器，从而可以对未知数据进行分类．支撑向量机是二分类问题的一个分类模型，模型的结果表现为支撑向量．Lee和Mangasarian在2001年提出了使用Sigmoid函数的积分函数作光滑的支撑向量机模型SSVM．该文研究了用多项式函数作光滑的支撑向量机(PSSVM)模型，并提出了两个用于光滑多项式的函数．根据模型特点，应用BFGS方法以及Newton Armijo方法进行求解，数值实验结果表明PSSVM模型在分类性能上优于SSVM模型．

Polynomial Smooth Support Vector Machine(PSSVM)

Data classification is an important issue of research on data mining. According to the sample dataset, we can build an mathematical model and get the optimal classifier. Then use this classifier to classify the unclassified data points. Support vector machine(SVM) is the main classification model of two classification. The result of support vector machine model is separating surface called support vector. In 2001, Lee and Mangasarian presented the smooth support vector machine(SSVM) which used the integral of Sigmoid function as smoothing function. In this paper, authors research the so called PSSVM which uses the polynomial functions to smoothen the objective function and present two polynomial functions. According to the features of PSSVM, authors use the BFGS and Newton Armijo methods to implement the experiment and show that PSSVM is better than SSVM.