An ARIMA Based Real-time Monitoring and Warning Algorithm for the Anomaly Detection

With the urgent demands of multi-parameter testing under the extreme environment,such as the deep water, upper air and deep underground etc., the fiber mechanical and thermal multi-parameter instrument is developed with its redominant advantages of accuracy, reliability, sensitivity and convenience. Most existing system usually defined a fixed threshold for the accident warning, which delayed reactions to the emergency. Thus, selecting an appropriate and adjustable threshold for anomaly detection is very necessary. In order to tackle this problem, a modified time series prediction model M-ARIMA is proposed in this paper. M-ARIMA can detect the emergency and achieve high real-time alarm rate. M-ARIMA is based on the dynamic variance and reduces the error rate of early warning caused by the normal fluctuations. Experimental results show that M-ARIMA can detect abnormities with a median accuracy of more than 92% and a median error of less than 10%.