摘要
针对时间序列中异常点的检测计算问题,提出了一种基于ARIMA模型和基于K−Means模型的组合异常检测计算方法。首先测试训练采用差分自回归移动平均模型(ARIMA),之后采用滑动窗口配合ARIMA模型对测试集进行预测得到异常预测值,然后计算误差项以及误差项的置信区间,误差项在置信区间判定范围以外的,将其对应的原始值判定为异常值。在检测出异常值之后,采用K−Means算法对原始数据进行聚类,然后通过计算出状态转移概率,对检测出的异常的取值结果进行质量评估,最后确定出异常值。实验探讨了算法中的滑动窗口对异常检测的影响,并以NAB部分数据集对算法进行了验证。实验结果表明,与同类经典算法相比,该算法不仅能够有效检测出时间序列中的异常点,而且在提高精准率和正确率方面取得了很好的效果。
Aiming at the problem of detection and calculation of abnormal points in time series,a combined anomaly detection calculation method based on ARIMA model and K-Means model is proposed.Firstly,the differential autoregressive moving average model(ARIMA)is used for test training.Secondly,the sliding window combined with ARIMA model is used to predict the test set to obtain the abnormal prediction value.Finally,the error term and its confidence interval are calculated.If the error term is outside the confidence interval determination range,its corresponding original value is determined as abnormal value.After the outliers are detected,the K-Means algorithm is used to cluster the original data,and then the state transition probability is calculated to evaluate the quality of the detected abnormal values,and finally the outliers are determined.The influence of sliding window on anomaly detection is analyzed,and the algorithm is verified with NAB data set.Experimental results show that compared with similar classical algorithms,this algorithm can not only effectively detect outliers in time series,but also achieve good results in accuracy.
作者
李鹏翔
刘佳楠
LI Pengxiang;LIU Jianan(Shaanxi Yanchang Petroleum Balasu Coal Industry Co.,Ltd.,Yulin 719000,China)
出处
《陕西煤炭》
2021年第S02期90-94,98,共6页
Shaanxi Coal
作者简介
李鹏翔(1992-),男,山西吕梁人,2020年毕业于太原理工大学计算机技术专业,硕士,现从事智能化建设方面的工作。
