K-最近邻的改进及其在文本分类中的应用

采用K近邻算法(Knearest neighbors,简称KNN)进行分类时,如果训练样本数量太大,那么搜索测试样本的K个最近邻时,算法的计算量很大.本文针对KNN的不足提出了一种改进方法.改进的KNN算法通过定义样本的延拓类和延拓能力,保留延拓能力强的样本作为它延拓类中其它训练样本的代表,来缩减训练样本数量,达到减少算法计算量的目的.实验证明,改进的KNN算法具有很好的性能.

Improvement of K Nearest Neighbors and Applications in Text Classification

When K nearest neighbor (KNN) is used to judge which category new texts belong to in text classification, KNN method may need much calculation to search k nearest neighbors of new texts if the number of training set is large. Introduce extension category and extension capability of a sample in the training set, An improved KNN method is presented, which can reduce the amount of training sets and select good samples from the training set to make up of a new training set to solve the problem above. The experiment result shows that it has good performance.