Journal Text Categorization with the Combination of Local Linearity and One-Class

A research is proposed on journal text categorization with the combination of local linearity and one-class. Local linearity is introduced to determine the samples' low-dimensional manifold, which could be regarded as the distribution of the samples in low-dimensional mapping spaces. At the same time, the border of positive and negative samples is determined by one-class. Compared with Knearest algorithm, linear SVM and one-class SVM, the new algorithm of journal text categorization gives better results in high precision, simple parameter estimation and easy control of risks, which gives an effective approach for the solution of text categorization.