LAUSR

The real-time sensing data in sensor networks are mostly data stream, and due to the inevitable factors such as external natural environment and man-made interference of sensors deployment, the phenomenon of data drift like slowing down or aggravation often appears when the perceived data stream propagates in the time domain, making it impossible to match the real-time data with standardized event templates in real time. Thus, it is difficult to identify the results through the existing data stream monitoring approaches before the ends of disaster events, and the accuracy is extremely low. Therefore, aiming at the deficiencies of the existing real-time identification approaches, this paper proposes a multi-stage real-time identification approach (MRIA) for data stream events with data drift feature based on dynamic time warping. First, the initial identification domain of data stream events is determined, and an anti-aliasing model based on dynamic time warping is constructed for the drift feature of the data stream, realizing the real-time similarity matching between the real-time data and event template. Second, a variable sliding window mechanism is introduced to determine the starting position of events, and an optimized matching approach for the incremental sequence is proposed to reduce the computational cost of re-matching in the process of the matching, and it obtains the identification benchmark of the similarity matching between the real-time data and the event template dynamically by dynamic threshold setting, which can improve the accuracy of matching. On this basis, an event multi-stage real-time identification approach based on identification proportion allocation is proposed, which can obtain the possibilities of events occurrence and the information of final disaster events quickly through the initial real-time identification and the final real-time identification process. The data compression optimization strategy based on the piecewise aggregate approximation approach is proposed to reduce the data scale, which further improves the identification efficiency. Furthermore, it provides an effective way for real-time identification of data stream events. The experimental results show that the approach proposed in this paper has great advantages in the efficiency and accuracy of data stream events identification.