K-近邻算法的改进及实现

利用k-近邻算法进行分类时。如果属性集包含不相关属性或弱相关属性，那么分类精度将会降低。研究了k-近邻分类器，分析了k-近邻分类器的缺点，提出了一种利用随机属性子集组合k近邻分类器的算法。通过随机的属性子集组合多个k近邻分类器，利用简单的投票，对多个k-近邻分类器的输出进行组合，这样可有效地改进k-近邻分类器的精度。

Improvement and Implementation of K Nearest Neighbors Algorithm

When k nearest neighbors is used to judge which category new objects belong to in classification, the classification precision is reduced if the attribute collection contains the uncorrelated attribute or the weak correlated attribute. After discussing k Nearest Neighbors and analyzing the shortcomings of k Nearest Neighbors, the paper proposes an algorithm of MKNN using a random subsets of attributions. With the simple voting method, the multiple nearest neighbor classifiers are combined via a random attribution set and the outputs of the multiple KNN classifiers are combined. The method of MKNN can improve on the classification precision.