Multi-Horizon Time Series Forecasting with Temporal Attention Learning

By Chenyou Fan, Yuze Zhang, Yi Pan, Xiaoyue Li, Chi Zhang, Rong Yuan, Di Wu, Wensheng Wang, Jian Pei, Heng Huang. 2019

235 citations recorded by the source corpus.

We propose a novel data-driven approach for solving multi-horizon probabilistic forecasting tasks that predicts the full distribution of a time series on future horizons. We illustrate that temporal patterns hidden in historical information play an important role in accurate forecasting of long time series. Traditional methods rely on setting up temporal dependencies manually to explore related patterns in historical data, which is unrealistic in forecasting long-term series on real-world data. Instead, we propose to explicitly learn constructing hidden patterns' representations with deep neural networks and attending to different parts of the history for forecasting the future. In this paper, we propose an end-to-end deep-learning framework for multi-horizon time series forecasting, with temporal attention mechanisms to better capture latent patterns in historical data which are useful in predicting the future. Forecasts of multiple quantiles on multiple future horizons can be generated simultaneously based on the learned latent pattern features. We also propose a multimodal fusion mechanism which is used to combine features from different parts of the history to better represent the future. Experiment results demonstrate our approach achieves state-of-the-art performance on two large-scale forecasting datasets in different domains.

Subfields: Computer Vision and Pattern Recognition, Management Science and Operations Research, Model Architectures, Signal Processing, Time-Series & Forecasting

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

1. Bryan Lim
2. Artur Dubrawski
3. Tomas Pfister
4. Kin G. Olivares
5. Max Mergenthaler-Canseco
6. Federico Garza
7. Sercan Ö. Arik
8. Boris N. Oreshkin

In Management Science and Operations Research

1. Tomas Pfister
2. Sercan Ö. Arik
3. Bryan Lim
4. Artur Dubrawski
5. Boris N. Oreshkin

In Model Architectures

1. Bryan Lim
2. Tomas Pfister
3. Nicolas Loeff
4. Sercan Ö. Arik
5. Boris N. Oreshkin

In Signal Processing

1. Tomas Pfister
2. Sercan Ö. Arik
3. Bryan Lim
4. Artur Dubrawski
5. Kin G. Olivares
6. Max Mergenthaler-Canseco
7. Federico Garza
8. Boris N. Oreshkin

In Time-Series & Forecasting

1. Tomas Pfister
2. Sercan Ö. Arik
3. Bryan Lim
4. Boris N. Oreshkin

Citations in the corpus

• Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting (2019, methodology): “DNNs for Multi-horizon Forecasting: Similarly to traditional multihorizon forecasting methods [18, 19], recent deep learning methods can be categorized into iterated approaches using autoregressive models [9, 6, 12] or direct methods based on sequence-to-sequence models [10, 11].”
• N-HiTS: Neural Hierarchical Interpolation for Time Series Forecasting (2022, methodology): “Neural long-horizon forecasting research has mostly focused on attention efficiency making self-attention sparse (Child et al. 2019; Li et al. 2019; Zhou et al. 2020) or local (Li et al. 2019).”