Abstract: With the rapid development of World Wide Web, text classification has become the key technology in organizing and processing large amount of document data As a simple, effective and nonparametric classification method, k NN method is widely used in document classification But k NN classifier not only has large computational demands, but also may decrease the precision of classification because of the uneven density of training data In this paper, a density based method for reducing the amount of training data is presented, which solves two problems mentioned above It not only reduces the computational demands of k NN classifier, but also makes the density of training data even and decreases the wrong classification between the edge of classes The experiment also shows that it has good performance