计算机工程与应用 ›› 2018, Vol. 54 ›› Issue (19): 254-260.DOI: 10.3778/j.issn.1002-8331.1706-0234

• 工程与应用 • 上一篇    下一篇

基于滑动窗口和ARMA的Argo剖面数据异常检测算法

罗一迪,蒋  华,王慧娇,王  鑫   

  1. 桂林电子科技大学 计算机与信息安全学院,广西 桂林 541004
  • 出版日期:2018-10-01 发布日期:2018-10-19

Anomaly detection algorithm of Argo profile based on Sliding Window and ARMA

LUO Yidi, JIANG Hua, WANG Huijiao, WANG Xin   

  1. College of Computer and Information Security, Guilin University of Electronic Technology, Guilin, Guangxi 541004, China
  • Online:2018-10-01 Published:2018-10-19

摘要: 针对由于Argo浮标的随机性和抛弃性而导致难以保证剖面数据质量的问题,根据滑动窗口(Sliding Window,SW)与自回归移动平均(AutoRegressive Moving Average,ARMA)模型的特点,提出了一种基于滑动窗口和ARMA的Argo剖面异常检测算法。利用滑动窗口将Argo剖面时间序列进行划分,再通过建立ARMA模型获取剖面的预测值,然后确定置信区间,最后通过判断观测数据是否在置信区间内实现异常检测。通过全球Argo浮标剖面数据进行实验,在滑动窗口宽度10~20,置信度在80%~90%时,敏感度可以达到85%以上,且准确度在99%以上。

关键词: Argo浮标, 滑动窗口, 自回归移动平均模型, 异常检测

Abstract: To solve the problem that it is difficult to guarantee the quality of data due to the randomness and abandonment of Argo buoy, according to the Sliding Window(SW) and AutoregRessive Moving Average(ARMA) model, the anomaly detection algorithm of Argo profile based on sliding window and ARMA is proposed. Firstly, the Argo profile series is split by using the sliding window, and then the ARMA model is established so as to get predicted value. Then the confidence interval is determined. Anomaly detection is achieved, finally, by judging whether the observed data is in the confidence interval. The corresponding result is validated with the real Argo profile data, and the sensitivity and accuracy of proposed method reaches at least 85% and 99% respectively when the wide of sliding window between 10 and 20 and the confidence level is in the condition of 80%~90%.

Key words: Argo profile, sliding window, AutoRegressive Moving Average(ARMA), anomaly detection