基于小样本集弱学习规则的KNN分类算法

摘要

KNN and its improved algorithms identify the class labels of the unlabeled datasets Du by using the labeled datasets Dl, if the data objects in Dl are very little, and this will influence the accuracy of classification.Improving the accuracy of classification was the goal of KNN classification algorithm based on the rule of weak learning on small sample sets, which learned the label information of objects in Dl based on Dl firstly, and then selected some data objects in Du and labeled them by using the learned label information, finally labeled the objects in Du based on the expanded labeled datasets Dl.The accuracy of the presented method is demonstrated with standard datasets, and obtains a satisfying result.%KNN及其改进算法使用类标号已知的数据集Dl对类标号未知的数据集Du进行类别标志,如果Dl中的数据数量过少,将会影响最后的分类精度.基于小样本弱学习规则的KNN分类算法旨在提高基于小样本集的KNN算法的分类精度,它首先对Dl中的数据对象进行学习,从中选取一些数据,利用学到的标签知识对其进行类别标号,然后将其加入到Dl中;最后利用扩展后的Dl时Du中的数据对象进行类别标志.通过使用标准数据集的测试发现,该算法能够提高KNN的分类精度,取得了较满意的结果.