An Improving Mulit-category Classification Method Based on the Binary Tree Support Vector Machine

Aiming at the problems of the slow convergence speed of general partial binary tree support vector machine (SVM) classifier and the fault samples easy to accumulate caused by complete binary tree and partial binary tree SVM classifier. The thesis proposes a multi-category classification method based on the unbalanced binary tree support vector machine (SVM),which construct a unbalanced binary tree SVM, making the easy to distinguish category can split out step by step from the root node, and reducing the accumulated errors caused by previous classification by analyzing the distribution of sample space. The results show that, comparing this method with the method of completeand partialbinary tree ,an unbalanced binary tree SVM built in this paper has a strong ability of autonomous learning, and can easily distinguish separate classes first, thus improving classification accuracy.