DeepAR: Probabilistic Forecasting with Autoregressive Recurrent Networks
By Valentin Flunkert, David Salinas, Jan Gasthaus. International Journal of Forecasting, 2017
3,009 citations recorded by the source corpus.
Probabilistic forecasting, i.e. estimating the probability distribution of a time series' future given its past, is a key enabler for optimizing business processes. In retail businesses, for example, forecasting demand is crucial for having the right inventory available at the right time at the right place. In this paper we propose DeepAR, a methodology for producing accurate probabilistic forecasts, based on training an auto regressive recurrent network model on a large number of related time series. We demonstrate how by applying deep learning techniques to forecasting, one can overcome many of the challenges faced by widely-used classical approaches to the problem. We show through extensive empirical evaluation on several real-world forecasting data sets accuracy improvements of around 15% compared to state-of-the-art methods.
Subfields: Management Science and Operations Research, Signal Processing, Time-Series & Forecasting
View Paper: https://arxiv.org/abs/1704.04110
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Management Science and Operations Research
- Rob J Hyndman
- Fotios Petropoulos
- Konstantinos Nikolopoulos
- Nikolaos Kourentzes
- Vassilios Assimakopoulos
- Spyros Makridakis
- Christoph Bergmeir
- Evangelos Spiliotis
In Signal Processing
In Time-Series & Forecasting
- Fotios Petropoulos
- Rob J Hyndman
- Bryan Lim
- Bernie Wang
- Jan Gasthaus
- Konstantinos Nikolopoulos
- Tim Januschowski
- Christoph Bergmeir
Citations in the corpus
- Time-series forecasting with deep learning: a survey (2020, methodology): “Given the flexibility of neural networks, deep neural networks have been used to model both discrete [55] and continuous [37,56] targets—by customizing of decoder and output layer of the neural network to match the desired target type.”
- N-BEATS: Neural basis expansion analysis for interpretable time series forecasting (2019, background): “Somewhat independently, the work in the modern deep learning TS forecasting developed based on variations of recurrent neural networks (Flunkert et al., 2017; Rangapuram et al., 2018; Toubeau et al., 2019; Zia & Razzaq, 2018) being largely dominated by the electricity load forecasting in the multi-v”
- Deep Factors for Forecasting (2019, background): “Deep Factor with one factor and no random effects, accompanied with autoregressive inputs, reduces to DeepAR (Flunkert et al., 2017).”
- Recurrent Neural Filters: Learning Independent Bayesian Filtering Steps for Time Series Prediction (2019, methodology): “Benchmarks: We compare the VRNF-KF, VRNF-NN and standard RNF against a range of RNN, autoregressive and RVAE benchmarks – including the standard LSTM, DeepAR Model [13], Deep State Space Model (DSSM) [37], Variational RNN (VRNN) [7], and Deep Kalman Filter (DKF) [27].”
- FEDformer: Frequency Enhanced Decomposed Transformer for Long-term Series Forecasting (2022, methodology): “Despite the impressive results achieved by RNN-type methods (Rangapuram et al., 2018; Flunkert et al., 2017), they often suffer from the problem of gradi-* Despite the progress made by Transformer-based meth-ods for time series forecasting, they tend to fail in capturing the overall…”
- Forecasting in social settings: The state of the art (2020, background): “For neural nets, good forecasts have been obtained when there are enormous collections of available data (Salinas et al., 2017).”
- Adaptive Graph Convolutional Recurrent Network for Traffic Forecasting (2020, background): “A majority of such studies [16, 10, 8, 17, 18, 19, 20] rely on LSTM or GRU to model the temporal dynamics in the time series data.”
- Darts: User-Friendly Modern Machine Learning for Time Series (2021, methodology): “…al., 2017), Theta (Assimakopoulos and Nikolopoulos, 2000), Prophet (Taylor and Letham, 2018), FFT-based forecasting, RNN models similar to DeepAR (Flunkert et al., 2017), N-BEATS (Oreshkin et al., 2019), TCN (Bai et al., 2018), TFT (Lim et al., 2021) and generic regression models that can wrap…”
- Time Series Data Augmentation for Deep Learning: A Survey (2020, methodology): “In this subsection we demonstrate the practical effectiveness of data augmentation in two popular deep models DeepAR [Salinas et al., 2019] and Transformer [Vaswani et al.”
- Transformers in Time-Series Analysis: A Tutorial (2022, methodology): “The model performed better than state-of-the-art (SOTA), including Autoregressive Integrated Moving Average (ARIMA) [85], Prophet [86], LSTMa [87], LSTnet [88], and DeepAR [89].”
- A Survey on Bayesian Deep Learning (2016, methodology): “Forecasting DeepAR [21] LV DeepState [90] LV Spline Quantile Function RNN [27] LV DeepFactor [124] LV”
- Time-MoE: Billion-Scale Time Series Foundation Models with Mixture of Experts (2024, neutral): “Deep learning models have become powerful tools for time series forecasting over the past decade, which can be broadly categorized into two types: (1) univariate models , such as DeepState (Rangapuram et al., 2018), DeepAR (Salinas et al., 2020), and N-BEATS (Ore-shkin et al., 2020), which focus on ”