摘要
为研究机器学习系统在供水网络管理中的应用趋势,发现供水网络数据中的奇点,分析了8项国内外机器学习在供水网络管理中的研究,从数据处理到模型验证,对每项研究中使用的方法进行了分析。结果表明,类别不平衡问题主要是来自供水网络的典型数据,建议使用抽样方法来训练分类器解决此类问题;此外变量的缩放和转换通常会对模型的性能产生积极影响。
In order to study the application trend of machine learning system in water supply network management and discover the singularities in water supply network data, 8 domestic and foreign researches on machine learning in water supply network management were analyzed. From data processing to model verification, each study the methods used in were analyzed. The results show that the problem of category imbalance is mainly from the typical data of the water supply network. It is recommended to use sampling methods to train the classifier to solve such problems;in addition, the scaling and transformation of variables usually have a positive impact on the performance of the model.
作者
卢冰锋
Lu Bingfeng(Water-supply Branch,Jinneng Holdings Coal Industry Group,Datong 037003,China)
出处
《山西建筑》
2022年第15期123-126,共4页
Shanxi Architecture
关键词
供水网络
机器学习
预测模型
water supply network
machine learning
predictive model
