已有的非平衡数据分类算法主要采取直接对损失函数进行加权的方法.文中提出一种加权边缘的hinge损失函数并证明它的贝叶斯一致性,得到加权边缘支持向量机算法(WMSVM),并给出类似于SMO的求解方法.实验结果表明WMSVM在一些数据库上是有效的,从而从理论和实验上说明基于加权边缘的损失函数方法是已有代价敏感方法的一种较好补充. The existing unbalanced data classification algorithms mainly adopt the method of directly weighting the loss function.In this paper, we propose a hinge loss function of weighted edges and prove its Bayesian consistency, and obtain a weighted edge support vector machine (WMSVM) And gives the solution method similar to SMO.Experimental results show that WMSVM is effective in some databases, which theoretically and experimentally shows that the weighted edge-based loss function method is a good complement to the cost-sensitive method.