| 一种基于ARIMA模型与3?σ准则的取水异常检测方法 |
| |
| 作者姓名： | 赵和松 王圆圆 孙爱民 |
| |
| 作者单位： | 水利部信息中心,北京 100053;北京金水信息技术发展有限公司,北京 100053;河海大学计算机与信息学院,江苏 南京 211100 |
| |
| 摘 要： | 为提高取水预测数据的准确性,针对现有部分取水数据异常且难以进行人工判别的问题,提出一种基于ARIMA模型与3σ 准则的取水异常检测方法.分析每个取水点每年的日取水量的时间序列数据,使用时间序列的ARIMA模型和高斯分布的3σ 准则判断日取水量是否为异常值;通过时间序列分解算法分析异常值附近取水点的趋势,判断异常值附近是...
|
| 关 键 词： | 取水异常检测 机器学习 ARIMA模型 3σ准则 时间序列分解算法 |
| 收稿时间： | 2021/1/11 0:00:00 |
| 修稿时间： | 2021/9/29 0:00:00 |
A water intake anomaly detection method based on ARIMA model |
| |
| Authors: | ZHAO Hesong WANG Yuanyuan SUN Aimin |
| |
| Affiliation: | Information Center,Ministry of Water Resources,Beijing 100053 ,China;Beijing Jinshui Information Technology Development Co.,Ltd.,Beijing 100053 ,China; School of computer and information,Hohai University,Nanjing 211100 ,China |
| |
| Abstract: | Due to the different water intake equipment at different water intake points and the different water intake environment and operation environment, some abnormal data of water intake appear, which is difficult to be distinguished manually. In this paper, the time series data of daily water withdrawal in each year at each water intake point are analyzed, and the ARIMA model of time series and 3 sigma criteria of gaussian distribution are used to determine whether the daily water withdrawals are abnormal values; then, the trend of water intake point near the outliers is analyzed by STL to determine whether there are other undetected outliers around the outliers. Finally, the modified reference values of the outliers are given. |
| |
| Keywords: | Keywords Anomaly detection Machine learning Gaussian distribution STL |
| 本文献已被 万方数据 等数据库收录！ |
| 点击此处可从《》浏览原始摘要信息 |
|
点击此处可从《》下载全文 |
|
