A Multi-Horizon Quantile Recurrent Forecaster

Abstract

We propose a framework for general probabilistic multi-step time series regression. Specifically, we exploit the expressiveness and temporal nature of Sequence-to-Sequence Neural Networks (e.g. recurrent and convolutional structures), the nonparametric nature of Quantile Regression and the efficiency of Direct Multi-Horizon Forecasting. A new training scheme, *forking-sequences*, is designed for sequential nets to boost stability and performance. We show that the approach accommodates both temporal and static covariates, learning across multiple related series, shifting seasonality, future planned event spikes and cold-starts in real life large-scale forecasting. The performance of the framework is demonstrated in an application to predict the future demand of items sold on Amazon.com, and in a public probabilistic forecasting competition to predict electricity price and load.

Published @ 31st Conference on Neural Information Processing Systems (NIPS 2017), Time Series Workshop. Long Beach, CA, USA

References in corpus (6)

• WaveNet: A Generative Model for Raw Audio
• Conditional Time Series Forecasting with Convolutional Neural Networks
• Professor Forcing: A New Algorithm for Training Recurrent Networks
• Analyzing and Exploiting NARX Recurrent Neural Networks for Long-Term Dependencies
• Predicting Surgery Duration with Neural Heteroscedastic Regression
• Position-based Content Attention for Time Series Forecasting with Sequence-to-sequence RNNs

Cited by in corpus (51)

• Time Series Forecasting With Deep Learning: A Survey
• Recurrent Neural Networks for Time Series Forecasting: Current Status and Future Directions
• Enhancing the Locality and Breaking the Memory Bottleneck of Transformer on Time Series Forecasting
• Autoformer: Decomposition Transformers with Auto-Correlation for Long-Term Series Forecasting
• Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting
• Deep Learning for Time Series Forecasting: Tutorial and Literature Survey
• Neural basis expansion analysis with exogenous variables: Forecasting electricity prices with NBEATSx
• LSTM-MSNet: Leveraging Forecasts on Sets of Related Time Series with Multiple Seasonal Patterns
• TCCT: Tightly-Coupled Convolutional Transformer on Time Series Forecasting
• Think Globally, Act Locally: A Deep Neural Network Approach to High-Dimensional Time Series Forecasting
• GluonTS: Probabilistic Time Series Models in Python
• Deep Factors for Forecasting
• A Spatiotemporal Deep Neural Network for Fine-Grained Multi-Horizon Wind Prediction
• TCGAN: Convolutional Generative Adversarial Network for Time Series Classification and Clustering
• Weather-Informed Probabilistic Forecasting and Scenario Generation in Power Systems
• Short-term daily precipitation forecasting with seasonally-integrated autoencoder
• Bridging Physics-based and Data-driven modeling for Learning Dynamical Systems
• Probabilistic Forecasting with Temporal Convolutional Neural Network
• Deep Factors with Gaussian Processes for Forecasting
• Domain Adaptation for Time Series Forecasting via Attention Sharing
• Adjusting for Autocorrelated Errors in Neural Networks for Time Series
• Topological Attention for Time Series Forecasting
• ScoreGrad: Multivariate Probabilistic Time Series Forecasting with Continuous Energy-based Generative Models
• Deep Generative Quantile-Copula Models for Probabilistic Forecasting
• Learning Quantile Functions without Quantile Crossing for Distribution-free Time Series Forecasting
• Interpretable Water Level Forecaster with Spatiotemporal Causal Attention Mechanisms
• Intermittent Demand Forecasting with Deep Renewal Processes
• Demand Forecasting of Individual Probability Density Functions with Machine Learning
• Learning Interpretable Deep State Space Model for Probabilistic Time Series Forecasting
• Simultaneously Reconciled Quantile Forecasting of Hierarchically Related Time Series
• Hierarchically Regularized Deep Forecasting
• A Scalable and Transferable Time Series Prediction Framework for Demand Forecasting
• MixSeq: Connecting Macroscopic Time Series Forecasting with Microscopic Time Series Data
• MQTransformer: Multi-Horizon Forecasts with Context Dependent and Feedback-Aware Attention
• Multi-stream RNN for Merchant Transaction Prediction
• Few-Shot Learning Patterns in Financial Time-Series for Trend-Following Strategies
• Online Multi-horizon Transaction Metric Estimation with Multi-modal Learning in Payment Networks
• Knowledge Enhanced Neural Fashion Trend Forecasting
• Long-term series forecasting with Query Selector -- efficient model of sparse attention
• Probabilistic Hierarchical Forecasting with Deep Poisson Mixtures
• Multi-future Merchant Transaction Prediction
• A Local Regret in Nonconvex Online Learning
• Unified recurrent neural network for many feature types
• Probability Paths and the Structure of Predictions over Time
• On the benefits of maximum likelihood estimation for Regression and Forecasting
• Quantifying Uncertainty in Deep Spatiotemporal Forecasting
• DMIDAS: Deep Mixed Data Sampling Regression for Long Multi-Horizon Time Series Forecasting
• Graph Deep Factors for Forecasting
• Synergetic Learning of Heterogeneous Temporal Sequences for Multi-Horizon Probabilistic Forecasting
• Take a NAP: Non-Autoregressive Prediction for Pedestrian Trajectories
• Long-Term Effect Estimation with Surrogate Representation