Abstract: There exists error accumulation in the multiclassification method, based on support vector machines and decision trees. It tends to decrease classification accuracy and results in a bad classification. With a careful analysis of error accumulation, it proposes a new multiclassification method, based on Huffman Tree and SVM. It divided a multiclassification problem into multiple binary classification problems, and gave classification priority, depending on the dissimilarity. At last, through an experiment with Lecast open source data sets, it verified the effectiveness. The experimental results show that the new method is superior to the traditional multiclassification method in classification speed and classification accuracy. 

Key words: decision tree, support vector machine, dissimilarity, Huffman tree