关键词: 文档分类, 最大间隔投影, 数据挖掘, 流形学习

Abstract: To effectively resolve the following drawbacks of the traditional vector representation-based document dimensionality reduction algorithms，such as the curse of dimensionality and singular value problems，a novel Web document classification algorithm based on tensor maximum margin projection is proposed in this paper.This algorithm can improve the discriminant ability of dimensionality reduction algorithm by fully using the structure and association information of documents.The experimental results on the WebKB and 20NG document collections show that the proposed algorithm achieves much better performance than other conventional document classification algorithms.

Key words: document classification, maximum margin projection, data mining, manifold learning