A Multi-Horizon Quantile Recurrent Forecaster (2017-11-29T00:00:00.000000Z)

TL;DR

A framework for general probabilistic multi-step time series regression that exploits the expressiveness and temporal nature of Recurrent Neural Networks, the nonparametric nature of Quantile Regression and the efficiency of Direct Multi-Horizon Forecasting is proposed.

Abstract

We propose a framework for general probabilistic multi-step time series regression. Specifically, we exploit the expressiveness and temporal nature of Recurrent Neural Networks, the nonparametric nature of Quantile Regression and the efficiency of Direct Multi-Horizon Forecasting. A new training scheme for recurrent nets is designed 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.

Authors

Ruofeng Wen

1 papers

K. Torkkola

1 papers

Balakrishnan Narayanaswamy

1 papers

A Multi-Horizon Quantile Recurrent Forecaster

TL;DR

Abstract

Authors

References25 items

Position-Based Content Attention for Time Series Forecasting with Sequence-to-Sequence RNNs

Conditional Time Series Forecasting with Convolutional Neural Networks

Analyzing and Exploiting NARX Recurrent Neural Networks for Long-Term Dependencies

Revisiting NARX Recurrent Neural Networks for Long-Term Dependencies

Predicting Surgery Duration with Neural Heteroscedastic Regression

Quantile autoregression neural network model with applications to evaluating value at risk

Professor Forcing: A New Algorithm for Training Recurrent Networks

WaveNet: A Generative Model for Raw Audio

Probabilistic energy forecasting: Global Energy Forecasting Competition 2014 and beyond

Learning to Diagnose with LSTM Recurrent Neural Networks

Scheduled Sampling for Sequence Prediction with Recurrent Neural Networks

Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation

A review of unsupervised feature learning and deep learning for time-series modeling

Generating Sequences With Recurrent Neural Networks

Direct Multi-Step Estimation and Forecasting

Learning to Forget: Continual Prediction with LSTM

A Quantile Regression Neural Network Approach to Estimating the Conditional Density of Multiperiod Returns

Recurrent Networks and NARMA Modeling

Finding Structure in Time

Time Series Analysis: Forecasting and Control

Time Series Analysis: Forecasting and Control

A Bias and Variance Analysis for Multistep-Ahead Time Series Forecasting

Regression Quantiles

Mixture Density Networks

Courville , and Yoshua Bengio . “ Professor forcing : A new algorithm for training recurrent networks . ” In Advances In Neural Information Processing Systems

Field of Study

Journal Information

Name