Multivariate Time Series
Multivariate time series analysis focuses on understanding and predicting the interconnected behavior of multiple time-dependent variables. Current research emphasizes developing advanced deep learning models, such as Transformers, state-space models (like Mamba and its variants), and convolutional neural networks, to effectively capture complex temporal and cross-variable dependencies, including lead-lag dynamics and non-stationary patterns. These advancements are crucial for improving forecasting accuracy in diverse applications, from environmental monitoring (e.g., smog prediction) to healthcare (e.g., anomaly detection in medical time series) and telecommunications. Furthermore, significant effort is dedicated to enhancing model robustness against attacks and improving interpretability through techniques like influence functions and counterfactual explanations.
Papers
Explaining Time Series via Contrastive and Locally Sparse Perturbations
Zichuan Liu, Yingying Zhang, Tianchun Wang, Zefan Wang, Dongsheng Luo, Mengnan Du, Min Wu, Yi Wang, Chunlin Chen, Lunting Fan, Qingsong Wen
Deep Learning-based Group Causal Inference in Multivariate Time-series
Wasim Ahmad, Maha Shadaydeh, Joachim Denzler
Improving age prediction: Utilizing LSTM-based dynamic forecasting for data augmentation in multivariate time series analysis
Yutong Gao, Charles A. Ellis, Vince D. Calhoun, Robyn L. Miller
Dance of Channel and Sequence: An Efficient Attention-Based Approach for Multivariate Time Series Forecasting
Haoxin Wang, Yipeng Mo, Nan Yin, Honghe Dai, Bixiong Li, Songhai Fan, Site Mo