摘要
为了克服用重复剪辑近邻法筛选训练样本集产生样本被误剔除进而增大决策树的判决风险和误判概率的问题,提出一种新的方法-引入拒绝阀值的重复剪辑近邻法,并从理论上分析了它降低判决风险和误判概率的原理。通过实验比较这两种方法发现,引入拒绝阀值的重复剪辑近邻法在降低判决风险和误判概率上要优于重复剪辑近邻法;而在决策树的规模和分类错误率上,重复剪辑近邻法的精度要优于引入拒绝阀值的重复剪辑近邻法。
Mistakenly removing sample thereby increasing the risk in the judgement and the probability misjudgment of the decision tree is a problem with multi-edit-nearest-neighbor algorithm in selecting training samples. For solving this problem, a new algorithm, refusing threshold multi-edit-nearest-neighbor algorithm, is proposed. The mechanism how it avoided this default is analysed. By experiments comparing to multi-edit-nearest-neighbor algorithm show that the refusing threshold multi-edit-nearest-neighbor algorithm in reducing the risk the judgement and probability misjudgment is superior to multi-edit-nearest-neighbor algorithm. However, in term of the accuracy rate between decision trees and the classification, the multi-edit-nearest-neighbor algorithm is better than the refusing threshold multiedit-nearest-neighbor algorithm.
出处
《计算机工程与设计》
CSCD
北大核心
2008年第15期3989-3990,4057,共3页
Computer Engineering and Design
关键词
数据挖掘
决策树
引入拒绝阀值的重复剪辑近邻法
重复剪辑近邻法
样本筛选
data mining
decision tree
refusing threshold multi-edit-nearest-neighbor algorithm
multi-edit-nearest-neighboralgorithm
sample selection