Authors:
Wei Liu
1
;
Hongyi Jiang
2
;
Dandan Che
1
;
Lifei Chen
3
and
Qingshan Jiang
2
Affiliations:
1
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, P.R. China, Shenzhen School of Advanced Technology, University of Chinese Academy of Sciences, Shenzhen, P.R. China
;
2
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, P.R. China
;
3
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, Shenzhen, P.R. China, Digital Fujian IoT Laboratory of Environmental Monitoring, Fujian Normal University, Fuzhou, P.R. China
Keyword(s):
IoT, Real-time Data, Anomaly Detection, Smoothed Z-Score Algorithm, Dynamic Threshold.
Abstract:
Temperature control plays a vital part in medical supply management, of which effective monitoring and anomaly detection ensure that the medication storage is maintained properly to meet health and safety requirements. In this paper, an unsupervised temperature anomaly detection method, called DTAD (Dynamic Threshold Anomaly Detection), is proposed to detect anomalies in real-time temperature time series. The DTAD sets dynamic thresholds based on the Smoothed Z-Score Algorithm, rather than set fixed thresholds of a temperature range by experience. The comparative evaluation is performed on the DTAD and four other commonly employed methods, the results of which shows that the DTAD reaches a higher accuracy and a better time efficiency. The DTAD is fully automated and can be used in developing a real-time IoT temperature anomaly detection system for medical equipment.