Anomaly Detection Algorithm for Multivariate Complex Data Based on Deep Neural Network

The conventional data anomaly detection structure is mostly single target execution, and the detection efficiency is relatively low, resulting in an increase in the final false detection rate. Therefore, the design and practical verification of the anomaly detection algorithm for multivariate complex data based on deep neural network are proposed. At present, the data anomaly detection feature extraction is carried out first, and the multi-target method is used to improve the detection efficiency and build a multi-target anomaly monitoring and calculation structure. Based on this, the deep neural network data anomaly detection and calculation model is designed, and the adaptive cascade scheduling method is used to achieve anomaly detection. The test results show that compared with YOLOv7 target detection algorithm and multiple data enhancement strategy MODIS data segmentation optimization and typical related forest algorithm. The final false detection rate of the multi complex data anomaly detection algorithm of the deep neural network designed this time is relatively low, which indicates that the detection algorithm designed for abnormal data is more efficient, the detection range is identifiable, targeted and has practical application value with the help and support of the deep neural network technology.