摘要
K近邻算法作模式识别研究热点之一,其样本点最近邻个数的选取一直受到广泛的关注。传统的KNN算法再分配为样本点分配最近邻个数时没有考虑样本空间分配不均的情况。针对这种情况提出一种基于高斯函数加权的自适应KNN算法,根据数据总体分布特点为每一个样本点自适应地分配最近邻个数。实验结果表明,与现有的典型算法相比,该算法表现出较好的分类精度。
K-nearest neighbor algorithm is one of the hottest topics in pattern recognition, the selection of the nearest neighbor number of sample points has drawn much attention. When the traditional KNN algorithm is redistributed, the distribution of the nearest neighbor in the sample points does not consider the uneven distribution of sample space. Proposes an adaptive KNN algorithm based on Gaussian function weighting. The nearest neighbor number is adaptively assigned to each sample point according to the overall distribution of data. Experimental results show that the proposed algorithm shows better classification accuracy than the existing typical algorithms.
作者
李昂
LI Ang(Shanghai Maritime University, Shanghai 201306)
关键词
KNN算法
自适应KNN算法
最近邻个数
高斯函数
K-nearest Neighbor Algorithm
Adaptive K-nearest Neighbor Algorithm
Nearest Neighbors
Gaussian Function
