基于不同惩罚系数的SMO改进算法

Improved SMO algorithm with different error costs

为了解决Keerthi改进的序贯最小优化(SMO)算法在处理非平衡数据集时,整体分类性能较低、稳定性差等问题,对两个类别施加不同的惩罚系数的方法对算法作进一步改进,同时给出计算公式及算法步骤。实验结果表明,该算法不但提高了处理非平衡数据集的能力,也进一步提高了其稳定性。

When Keerthi＇s Sequential Minimal Optimization （SMO） algorithm is applied to the classification of unbalanced datasets, it not only leads to a poor classification performance but makes the result unstable, In order to overcome the difficulty, an improved SMO algorithm that used different error costs for different class was presented. Besides, the formula and the steps of the improved SMO algorithm were given. Experimental results show that our algorithm＇ s ability of dealing with unbalanced datasets can be improved and its stability also can be intensified.