论文部分内容阅读
介绍现有K-近邻分类法的基本思想和研究现状,并针对此方法在分类各类数据集分布不平衡时容易造成分类精度低的问题作相应的改进。改进的K-近邻分类法中引入类代表度和样本代表度,使得K-近邻分类法在相似度计算时选出的近邻样本更能代表其所在类,从而减小误判率。实验证明改进方法有效。
This paper introduces the basic idea and current research status of the existing K-nearest neighbor classification method, and improves the problem of low classification accuracy when the classification of various data sets is unbalanced. In the improved K-nearest neighbor classification, the class representativeness and the sample representativeness are introduced so that the nearest neighbor samples selected by the K-nearest neighbor classification method can better represent their class in the similarity calculation and thus reduce the false positive rate. Experiments show that the improvement method is effective.