Research on Sample Distribution Based on Classification Method
Generally speaking,KNN classifier and neighborhood classifier can perform well when the samples are evenly dis-tributed,but in practical application,samples are easily affected by technology,environment and other factors during data acquisi-tion,resulting in uneven distribution.Neither of the two classifications can achieve satisfactory classification results because of the uneven distribution of samples.To deal with this problem,a k-neighborhood classifier is designed based on KNN classifier and neighborhood classifier.The core idea of such classifier is to judge the sample distribution in the local area of the sample to be test-ed,and then to improve the classification accuracy by limiting the number of samples in the high-density area and reducing the search space of samples in the low-density area.The experiments employ three norm on 10 data sets of UCI database and ORL face database,and the k-neighborhood classifier is compared with KNN classifier and neighborhood classifier.The experimental results imply that compared with the other two classifiers,k-neighborhood classifier can not only improve the classification accuracy,but also maintain well-matched classification stability.
万方数据
维普
