Anomaly Detection of UAV Data in Graph-based Generative Adversarial Networks
UAV anomaly state detection is one of the important ways to ensure the safety of UAV flight,a-mong which anomaly detection based on UAV flight data is the most commonly used means.Therefore,this paper proposes a graph-based generative adversarial network for anomaly detection algorithm(TGAN-GAT).In this paper,TCN is used as the base network of generative adversarial network,which solves the problem that traditional recurrent neural networks cannot be computed in parallel and improves the training speed.And the graph attention mechanism is introduced to clearly capture the relationship be-tween different time series.The anomaly detection method is then based on the anomaly score due to the reconstruction error and the discrimination error.Experiments show that this anomaly detection algorithm improves the recall and F1 score by 10.68%and 7.02%relative to MTAD-GAT compared to the second place.
multivariate time seriesgenerative adversarial networksanomaly detectiongraph attention mechanismsunsupervised learning
万方数据
维普
