方法对比

	Temporal Fusion Transformer	DeepAR	Informer
领域	深度学习	深度学习	深度学习
方法族	Machine learning	Machine learning	Machine learning
起源年份	2021	2020	2021
提出者	Lim, B., Arık, S. Ö., Loeff, N. & Pfister, T.	Salinas, D., Flunkert, V. & Gasthaus, J. (Amazon)	Zhou, H. et al.
类型	Attention-based deep learning forecasting architecture	Autoregressive recurrent neural network (probabilistic forecasting)	Transformer (ProbSparse self-attention)
开创性文献	Lim, B., Arık, S. Ö., Loeff, N. & Pfister, T. (2021). Temporal Fusion Transformers for Interpretable Multi-Horizon Time Series Forecasting. International Journal of Forecasting, 37(4), 1748–1764.	Salinas, D., Flunkert, V., Gasthaus, J. & Januschowski, T. (2020). DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks. International Journal of Forecasting, 36(3), 1181–1191.	Zhou, H. et al. (2021). Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting. AAAI.
别名	Temporal Fusion Transformer (TFT), TFT, interpretable multi-horizon forecasting transformer	DeepAR — Olasılıksal RNN Tahmini, probabilistic autoregressive RNN forecasting, Amazon DeepAR	Informer — Uzun Dizi Transformer Tahmini, Informer transformer, ProbSparse attention forecaster
相关	6	5	5
摘要	The Temporal Fusion Transformer (TFT), introduced by Lim, Arık, Loeff and Pfister in 2021, is an interpretable deep learning architecture for multi-horizon time series forecasting. It combines variable selection, gating, multi-horizon attention and quantile outputs, processing static, past and known-future inputs together to produce multi-step forecasts.	DeepAR is Amazon's industrial forecasting model, introduced by Salinas, Flunkert and Gasthaus (2017; published 2020), that uses an autoregressive recurrent neural network to estimate the parameters of a probability distribution at each step, producing a confidence interval rather than a single point forecast. It can model many related time series jointly within one model.	Informer is a Transformer-based model introduced by Zhou et al. in 2021 for long-sequence time-series forecasting, using a ProbSparse self-attention mechanism that lowers the computational complexity of the standard Transformer to O(L log L). It is built for problems that demand predictions across thousands of future steps.