摘要
针对传统决策树算法应用于电梯液压缓冲器隐患智能识别准确率有待提高的问题,提出一种改进传统决策树ID3与主成分分析法(PCA)融合的智能识别方法(PAC-ID3)。针对传统ID3算法倾向于选择取值较多的属性缺点,引进属性阈值和信息增益率,对传统ID3算法进行改进优化;样本数集通过主成分分析方法,解决决策树存在多值倾向问题,选出更具有代表性的决策属性,提高决策树的建模效率和准确率;通过对改进优化融合前后算法进行了比较,实验结果表明,改进融合后的算法提高了隐患识别的精确率。
Aiming at the problem that the accuracy of traditional decision tree algorithm applied to the classification and prediction of elevator hydraulic buffer failure types needs to be improved,this paper proposes a method for predicting failure type classification that improves the integration of traditional decision tree ID3 and principal components analysis(PAC).Aiming at the shortcomings of traditional ID3 algorithms that tend to choose more attributes,the attribute threshold and information gain rate are introduced to improve and optimize the traditional ID3 algorithm.The sample number set uses the principal component analysis method to solve the problem of the multi-value tendency of the decision tree.To select more representative decision attributes to improve the modeling efficiency and accuracy of the decision tree.By comparing the algorithms before and after the improved optimization fusion,the experimental results show that the improved fusion algorithm improves the accuracy of the prediction classification rate.
作者
梁敏健
刘德阳
LIANG Minjian;LIU Deyang(Zhuhai Branch,Guangdong Institute of Special Equipment Inspection and Research,Guangdong Zhuhai 519002,China)
出处
《工业仪表与自动化装置》
2021年第5期94-100,共7页
Industrial Instrumentation & Automation
基金
广东省市场监督管理局科技项目(2020CT03,2018CT10)
广东省特种设备检测研究院科技项目(2020JD-2-05,2020JD-2-04,2021JD-1-05)。
