Data abnormal fluctuation detection and early warning
Chunyi Xu, Xuechen Sun, Yitong Yan · 2022 IEEE Conference on Telecommunications, Optics and Computer Science (TOCS) · 2022
This paper focuses on abnormal data evaluation index, tries to use quantitative abnormal data evaluation index and abnormal degree, so as to realize abnormal fluctuation detection and early warning.firstly analyzes whether the sensor data have fluctuation or not. By comparing the time series data difference and the average value of the difference, the fluctuating data, i. e. the abnormal data, can be filtered out. According to the five definitions and measurement standards of volatility characteristics, the risk assessment model of abnormal data is established, and the judgment of non-risk and risk is realized.And we set the two risk volatility characteristics as two quantitative indicators of abnormal degree of risk anomaly data, so we focus on the quantitative analysis of abnormal degree. By using entropy weight method, the weight of these two indexes is assigned, and then the abnormal degree is quantified by mechanism analysis.