Abstract: According to the maximal margin linear classifier and the contraction of closed convex hull，2-classification linearly non-separable problem can be transformed to linearly separable problem by using proposed contraction methods of closed convex hull.Multi-classification problem can be solved by contracting closed convex，and multi-classification algorithm based on the contraction of closed convex hull is presented.The geometric meaning of optimization problem is obvious.The shortcomings of complicated objective function in multi-classification are overcame，nonlinear separable multi-classification problem can be solved using kernel method.

Key words: Support Vector Machine（SVM）, multi-classification, closed convex hull

摘要： 在最大边缘线性分类器和闭凸包收缩思想的基础上，针对二分类问题，通过闭凸包收缩技术，将线性不可分问题转化为线性可分问题。将上述思想推广到解决多分类问题中，提出了一类基于闭凸包收缩的多分类算法。该方法几何意义明确，在一定程度上克服了以往多分类方法目标函数过于复杂的缺点，并利用核思想将其推广到非线性分类问题上。

关键词: 支持向量机, 多分类, 闭凸包