关键词: 类电磁机制算法, 支持向量机, 多分类, 最大分类间隔, 决策树

Abstract:

Based on the characteristics of the classification problem, a modified electromagnetism-like mechanism (EM) algorithm is designed, which is suitable for classification. Then a modified EM-based optimal decision-tree algorithm is proposed to deal with the SVM multi-class classification problem. First, EM is used to create an optimal or near-optimal decision tree automatically, which makes the margin between two classes maximal at every decision node. Then at every decision node, standard SVM is used to make binary classification. Finally, the SVM decision tree achieves multi-classification. Experimental results show that the proposed method is better than the traditional methods such as “1-a-1”, “1-a-r”, “DAG-SVM”,“DT-SVM” and “GADT-SVM”.

Key words: electromagnetism-like mechanism algorithm, support vector machine, multi-classification, maximal margin, decision tree