摘要
根据生产制造企业网络的特点,在智能故障诊断中,提出了一种基于分类采样的随机森林算法(CSRF).该算法结合随机森林算法基本原理,使用分类采样技术生成所需的训练样本,很大程度上解决了数据不均衡带来的问题.该算法为随机森林的每一棵分类回归树(CART)生成相应的训练数据,缓解了采样偏置,提高了算法的性能.实验表明:该算法与随机森林算法相比在准确率上提升了约4%,有效降低了故障诊断的风险.
An intelligent fault diagnose algorithm(Classified Sample Random Forest,CSRF)is proposed according to the characteristics of manufacturing enterprise network.The algorithm combines the basic principle of random forest algorithm,using classified sampling technology to generate the required training samples,and largely solved the problem caused by data imbalance.The algorithm can generate the corresponding training data for each classification and regression tree(CART)in random forest,which alleviates the sampling bias and improves the performance of the algorithm.Experiments show that the algorithm improves the accuracy rate by about 4%compared with the random forest algorithm,and reduces the risk of fault diagnosis effectively.
出处
《北京交通大学学报》
CAS
CSCD
北大核心
2017年第2期54-58,共5页
JOURNAL OF BEIJING JIAOTONG UNIVERSITY
基金
国家科技支撑计划项目(2015BAF08B02
2015BAG12B00)~~
关键词
工业以太网
智能算法
采样算法
随机森林
industrial ethernet
intelligent algorithm
sampling algorithm
random forest
