Abstract:
Aimed at the multiple model problem of data mining, the theory and technology of combination model is discussed and the application of combination theory to the nearest neighbor is studied. The paper proposes an algorithm of MNN (multiple nearest neighbor) classifiers using a random subset of attributions. With the simple voting method, the multiple nearest neighbor classifiers are combined via a random attribution set and the output of the multiple NN classifiers is combined. The method of MNN can improve on the classification precision. Comparing the MNN method to NN-ECOC, two strongpoint are obtained: (1)MNN is a more simple method; (2)MNN is not limited by multiple classes.
Key words:
Data mining,
Classification model,
Multiple model
摘要: 针对数据挖掘的组合模型问题,研究了组合模型的理论和技术,分析了组合理论在近邻法的应用现状,提出了一种通过随机属性子集组合近邻分类器的算法MNN,利用简单的投票方法,通过一个随机的属性子集来组合多重近邻分类器,对多重NN分类器的输出进行组合,MNN方法能有效地改进近邻法的分类精度。MNN方法与NN-E000相比,有两个主要的优点:(1) MNN是一个更简单的方法;(2) MNN不受多类问题的限制。
关键词:
数据挖掘,
分类模型,
组合模型
ZHENG Hongzhen; LIU Yang; ZHAN Dechen. Multiple Nearest Neighbor Algorithm Based on Data Mining[J]. Computer Engineering, 2007, 33(03): 48-49.
郑宏珍;刘 扬;战德臣. 基于数据挖掘的组合近邻模型算法[J]. 计算机工程, 2007, 33(03): 48-49.