海明距离判别法分类准确率的影响因素

发布时间：2018-03-26 12:06

本文选题：海明距离判别法　切入点：属性层级　出处：《江西师范大学学报(自然科学版)》2017年04期

【摘要】：为探讨海明距离判别法(HDD)的非参数优势,通过一个5因素混合实验,考察了4个因素(属性层级、测验长度、样本容量、知识状态分布)对HDD的3种判别方法(R方法、B方法、W方法)分类准确率的影响.结果表明:1)属性层级和测验长度均会影响HDD判准率,属性层级越紧密、测验长度越长,HDD判准率越高;2)HDD对样本容量无依赖,可适于小样本评估;3)HDD的R方法、B方法、W方法的分类准确率无差异;4)HDD无需被试知识状态分布的正态性假设,更适于均匀分布.
[Abstract]:In order to investigate the nonparametric advantage of Hamin distance discriminant method (HDD), four factors (attribute hierarchy, test length, sample size) were investigated by a 5-factor mixed experiment. The effect of knowledge state distribution on the classification accuracy of HDD's three discriminant methods / R method / B method / W method. The results show that both the attribute level and the test length will affect the HDD accuracy rate, and the closer the attribute hierarchy is, the closer the classification accuracy is. The longer the length of the test, the higher the accuracy of HDD. The higher the accuracy of HDD is, the higher HDD is independent of sample size, so it is suitable for small samples to evaluate HDD. There is no difference in the classification accuracy of the R method / B method / W method for evaluating HDD in small samples. The assumption that HDD does not need the normal distribution of knowledge states of the subjects is more suitable for uniform distribution.
【作者单位】：浙江师范大学教师教育学院;
【基金】：教育部人文社会科学研究一般课题(16YJA190002)资助项目
【分类号】：B841.7

【相似文献】