A Multi-Horizon Quantile Recurrent Forecaster

Ruofeng Wen, Kari Torkkola, Balakrishnan Narayanaswamy, Dhruv Madeka
11/29/2017
stat.ML

Official code: tianchen101/MQRNN (68 stars, last commit 2020-11-20).

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.

View on arXivView PDF

Code Implementations(1)

tianchen101/MQRNNOfficial100%

Multi-Quantile Recurrent Neural Network for Quantile Regression

6819Python, Jupyter NotebookNov 20, 20205 years ago

Similar Papers

Seasonal Encoder-Decoder Architecture for Forecasting

84% match2022

Time-Series Classification for Dynamic Strategies in Multi-Step Forecasting

84% match2024

Probabilistic Forecasting with Temporal Convolutional Neural Network

83% match1 repo1582019

SutraNets: Sub-series Autoregressive Networks for Long-Sequence, Probabilistic Forecasting

83% match2023

TiRex-2: Generalizing TiRex to Multivariate Data and Streaming

82% match2026

Some variation of COBRA in sequential learning setup

82% match2024

Cite this paper

@article{wen2017multihorizon,
  title  = {A Multi-Horizon Quantile Recurrent Forecaster},
  author = {Ruofeng Wen and Kari Torkkola and Balakrishnan Narayanaswamy and Dhruv Madeka},
  year   = {2017},
  eprint = {1711.11053},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1711.11053}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1711.11053.svg)](https://codeofpaper.com/paper/1711.11053)

Discussion