【摘要】 利用基于阈值聚类算法首先对带类标记的样本数据集进行有指导性聚类,其主要目的是压缩训练数据集,解决KNN分类算法的样本选择问题以及孤立点的发现,用少量的更具代表性的聚类中心替代KNN算法中巨大的样本集,然后利用聚类密度改进KNN分类算法,从而提高KNN分类检测的准确度和速度.更多还原

【Abstract】 A guidance clustering of labeled empirical data is studied using threshold-based clustering.The main target of clustering is to compress huge training dataset to the limited number of clustering centroids,to solve the problem of sample selection in KNN classification algorithm,and discover isolated point.A smaller number of more representative clustering centroids are used in place of the original huge sample dataset of KNN algorithm.Then,the clustering density is used to revise KNN classification algorithm to improve the detection accuracy and detection speed.更多还原