针对K-nn算法应用于文本分类时存在的问题进行改进,提出利用聚簇进行训练样本约减的方法,并用神经网络实现改进的k-nn分类器。实验表明,分类所用的时间比传统k-nn算法有明显减少,一定程度上提高了分类效率,同时分类准确性有所提高。 In order to improve the existing problems when K-nn is applied to text classification, this paper proposes a method of reducing the training samples by clustering, and realizes an improved k-nn classifier by using neural network. Experiments show that the classification time is significantly reduced than the traditional k-nn algorithm, to a certain extent, improve the classification efficiency, while the classification accuracy has increased.