TSAD: Temporal-spatial association differences-based unsupervised anomaly detection for multivariate time-series

Hanbing Zhu, Nan Xiao, Hefei Ling, Zongyi Li, Yuxuan Shi, Chuang Zhao, Hongxu Ji, Ping Li, Hui Liu

Abstract: Highlights•Proposes two-branch Transformer architecture for temporal–spatial anomaly differentiation.•Introduces anomaly scoring integrating reconstruction errors and association differences.•Achieves SOTA performance on four industrial anomaly detection benchmarks.