Abstract: With the development of World Wide Web, text classification has become a key technology in organizing and processing large amount of document data. It’s a simple, effective and nonparametric classification method. This paper proposes an algorithm PIM-KNN(Parameter Iteratively Modified-KNN) to adjust parameter in classifier according to results of close test of the KNN algorithm: the sample of wrong judged should reduce the distance between itself and the class which it belongs to, enlarge the distance between itself and the class which wrong judged. The experiments results show that the classification results can be improved significantly by adjusting parameter of the PIM-KNN.

Key words: text classification, KNN, iterative, distance