The application of digital archives classification with progressive M-SVM to wisdom school building

Abstract

When using the SVM algorithm, the training set is so large that the traditional classification methods can't satisfy the real-time requirements, how to design a more efficient SVM algorithm is one of the important study problems. We improve the method of the building about the digital archive's corpus and also improve the course of Chinese participle and the multiprocessing of text feature selection with TF, IDF and Information Gain. The experiment shows that this improved method about M-SVM has obtained a better result. © 2012 Springer Science+Business Media B.V.