Text classifier based on support vector machine and output coding

A new kind of text category algorithm based on SVM(Support Vector Machine) theory and output coding is given. Six combined multiclassifiers which cover three encoding styles (one-to-many, one-to-one, ECOC) and two different similarity calculating methods (hamming similarity , losing-based similarity)is used to classify the documents . Experiments with these multiclassifiers show that the combination of one-to-many with losing-based similarity has the highest recall rate and the highest precision rate.