A Smart Governance Anomaly Detection Method and System Based on Dynamic Spatiotemporal Hypergraph Evolution

Abstract

This invention relates to the field of data anomaly detection technology, and in particular to a smart governance anomaly detection method and system based on dynamic spatiotemporal hypergraph evolution. Based on the acquired raw observation data, adaptive modal decoupling and multi-view embedding are performed to obtain initial node representations, including signal decoupling based on variational modes and multi-view spatiotemporal embedding encoding. Based on the obtained initial node representations, dynamic evolutionary hypergraph structure learning is performed to obtain deep feature tensors, including dynamic hyperedge generation based on metric learning and spatiotemporal hypergraph convolutional evolution. Based on the deep feature tensors, multi-scale temporal prototype memory prediction is performed, including multi-scale temporal feature extraction, prototype memory retrieval and reconstruction, and future prediction of reconstructed features. This invention solves the problems of difficulty in predicting sudden anomalies and the lack of interpretability of deep models.

Images