Half-Against-Half Multi-Class Text Classification Using Progressive Transductive Support Vector Machine

Progressive Transductive Support Vector Machine extends Transductive Support Vector Machine in different class distribution. It is the solution to the problem that it has to estimate the ratio of positive negative examples from the sets which are not an easy task to deal with. The paper introduces a Half-Against-Half Multi-Class Text Classification algorithm Using Progressive Transductive Support Vector Machine. It shows both in theoretical estimation and experimental results on Reuters-21578 data set that HAH has significant advantages over the methods of OAA, OAO and DDAG in the testing speed, the training speed and the size of the classifier model, and the accuracy of classification is close to OAA, OAO and DDAG. It is a promising approach to solving the issues of large-scale multi-class text classification.