基于混淆交叉的多分类支撑向量机树

Multi-classification Tree-structured Support Vector Machine Based On Confusion Cross

本文针对复杂模式分类和多分类问题,提出了一种基于混淆交叉的多分类支撑向量机树模型,其整体结构为二叉树,在树的每个中间节点上嵌入了支撑向量机。在训练阶段,引入混淆交叉因子,在同属一个父节点的中间节点样例间进行样例的混淆交叉,将那些对分类曲面有较大影响的样例纳入树型结构更深层次的训练过程,参与更精细的分类超曲面的构建。本文将提出的支撑向量机树与现有的其他方法作了比较,实验结果说明了本文提出的模型在解决复杂模式识别问题及多分类问题上具有高效准确性及优越的泛化性能。

In this paper, the problems associated with complex pattern recognition and multi-classification are addressed and a tree-structured Support Vector Machine （TSSVM） with confusion cross is presented. A TSSVM is overall a binary tree, whose internal nodes are modular SVMs. Those two non-terminal nodes generated from the same parent node perform discounted confusion crossover controlled by a confusion cross factor. The presented approach is evaluated against other classifiers. The experimental results show that the proposed approach is competitive in dealing with complex pattern recognition problems and multi-classification problems.