Dynamic Detection Method for Spatiotemporal Data Based on Hybrid Model and Singular Spectrum Analysis

Abstract

As internet technology advances, processing a large amount of network data has become an important part of network work. To improve the processing effectiveness of data in the network, a dynamic data accuracy detection method based on spatiotemporal data mining is proposed. During the process, singular spectrum analysis is introduced to propose a dynamic data detection method. A data accuracy detection method is proposed by combining graph convolutional neural networks and temporal convolutional networks to detect data in both time and spatial dimensions. Finally, the effectiveness of the research method is analyzed. The experimental results show that the mean absolute error, mean absolute percentage error, and root mean square error of the proposed method are the lowest among the four models, at 0.16, 0.18, and 0.20, respectively, which are lower than the other three comparative methods; The research method maintains a relatively stable average accuracy in the range of 0.75~0.80 when dealing with different tasks. The research method requires a processing time of 250 ms for 2000 data points and 1000 ms for 6000 data points. Before and after using the research method, the data processing increases from around 2500 to around 2700 within 15ms, and from around 2900 to 3100 within 30ms. The dynamic data detection method designed in this study demonstrates good processing efficiency and accuracy in data detection. Research can provide certain technical references for dynamic data detection, improving the accuracy and reliability of data.