KD-KNN Text Categorization Method Based On Improvement TFI-DF

Text categorization under the assigned classification system, uses natural languages to express texts that sort automatically into the pre-set type set according to its contents and makes a piece of text correspond to one (several) type(s) of type set. In order to enhance the categorization performance, this paper proposes the improved TFI-DF feature extraction method and kernel-based distance-weighted KNN algorithm with the improved TFI-DF feature selection method to preprocess the data source and set up the space vector model, which provides a convenient data structure for text categorization, and enhances the precision of the feature selection method. The KNN algorithm (kernel-based distance-weighted KNN) has solved the multi-peak distribution and boundary overlap problems of samples and the precise classified decision problem of classifiers. The experimental system has proved the effectiveness of the new method.