摘要
KNN是最著名的模式识别统计学方法之一。它是一种无参数分类方法,由于其分类的简单有效性,因此得到较为广泛的应用。但是对KNN分类系统的全面评价还有待进一步研究。提出的改进加权KNN算法相比之下具有更高和更加稳定的识别率。因为它在经典KNN算法基础上增加加权距离和类间相似度信息,比经典KNN这种单纯依靠投票的分类方法更加可靠,在分类识别研究中更具有研究和应用价值。
KNN is one of the most famous statistical methods of pattern recognition. It is a non-parametric classification method, due to the simple effectiveness of its classification, so it has been more widely used. Further research needs to be a comprehensive evaluation of the KNN classification system. Proposes an improved weighted KNN algorithm, which has a higher and more stable compared to the recognition rate. Because it increases the degree of similarity between the weighted distance and class information in the classic KNN algorithm, based on the classic KNN than relying solely on the classification of this vote is more reliable, more research and application in classification study.
基金
广州市高等学校第五批教育教项目(No.JG201337)
广东省高职教育教学管理委员会项目(No.JGW2013070)
关键词
KNN
改时加权
加权距离
类间相似度
KNN （K-Nearest Neighborhood）
Changed Weighted
Weighted Distance
Similarity Between Classes
