摘要
最近邻算法由于操作简单,效果显著,无论在科研还是实际生活中都具有广泛应用。文章首先解释了基于欧式距离的最近邻算法在计算两个记录之间距离方面的不足,然后提出了基于马氏距离的最近邻算法,真实数据集的实验结果显示,改进后的最近邻算法能取得较好的成绩。
Nearest neighbor(NN) algorithm is applied widely in both scientific research and real application because it can be operated easily and the algorithm's performance usually is excellent than the corresponding methods.In this paper, we analyze the advantages of Euclidean-based NN algorithm, then propose Mahalanobis-based NN algorithm in which Mahalanobis distance metric is designed to replace the Euclidean distance for computing the distance between two records.Finally, the experimental results on real datasets show the improved method outperform the original one.
出处
《微计算机信息》
2010年第9期225-226,215,共3页
Control & Automation
基金
基金申请人:刘星毅
项目名称:工业数据集缺失数据的填充研究
基金颁发部门:广西科技厅(桂科自0899018)
基金申请人:刘星毅
项目名称:社会调查中缺失数据的研究
基金颁发部门:广西教育厅(200808MS062)
关键词
最近邻算法
数据缺失填充
马氏距离
nearest neighbor algorithm
missing data imputation
Mahalanobis distance