A Novel Multiclass Text Classification Algorithm Based on Multiconlitron

摘要

A novel multiclass text classification algorithm based on multiconlitron is proposed. The multiconlitron is constructed for each possible pair of classes in sample space, each of which is used to separate two classes. For the sample to be classified, every multiconlitron is used to judge its classman vote for the corresponding class. The final class of the sample is determined by the number of votes. The classification experiments are done on Reuters 21578, and the experimental results show that the proposed algorithm has better performance. Compared with multiclass support vector machines 1-a-1, 1-a-r and DAGSVM, while ensuring the classification accuracy, it improves the classification speed and training speed greatly.