海量数据分类问题是目前SVM学习算法研究的一个热点。传统的SVM方法是通过计算二次规划问题求解的,当训练样本数较大时存在一定的困难。文章介绍一种新的方法,该方法把二分类问题看作特殊的软最小球覆盖问题,提出求解SVM问题的全新的最小球覆盖几何解释,通过软最小球算法进行求解,避免了计算复杂度高的问题,可以用来求解l1目标函数海量数据ν-SVM分类问题。实验结果表明,文章提出的算法与同类算法相比具有相当的竞争力。 Mass data classification is a hot spot in SVM learning algorithms. The traditional SVM method is calculated by solving the quadratic programming problem, which has some difficulties when the training samples are large. In this paper, we introduce a new method which treats the dichotomous problem as a special soft minimum sphere covering problem. We propose a new minimum spherical covering geometric solution to the SVM problem, which is solved by the soft minimum sphere algorithm, which avoids the computational complexity The high problem can be used to solve l1 objective function mass data ν-SVM classification problem. Experimental results show that the proposed algorithm is quite competitive with similar algorithms.