Skip to content
No description, website, or topics provided.
Branch: master
Clone or download
robjhyndman Merge pull request #10 from mitchelloharawild/master
Fixed missing end-quote in readme
Latest commit 61ddc71 Jan 5, 2019

This page contains the explanation of our forecast method for the M4 competition, authored by Pablo Montero-Manso, Thiyanga Talagala, Rob J Hyndman and George Athanasopoulos.

As part of our submission, we are producing the R package M4metalearning, see installation instructions below. This package is intended to facilitate reproducing the results of our submission, but can also be used on its own to apply our approach to other datasets, either using the pretrained model submitted to the M4 competition or by training new models.

Additionally, the authors have produced another R package M4comp2018, which facilitates the users the access to the M4 competition dataset. The M4comp2018 package can be found here

The description is divided into three sections, the first t

  1. Methodology
  2. Reproducing the results
  3. Usage example of the package


The goal of the M4Metalearning package is to provide forecasting metalearning tools based on our submission to the M4 competition.


You can install M4metalearning from github with:

# install.packages("devtools")


M4metalearning is using, for the time being, a slight modification of the tsfeatures package. Please install it from:

# install.packages("devtools")

Also, a custom version of the xgboost package is required. It is installed automatically when calling the training functions that use it, since it does not break compatibility. It just supports customized multiclass objective functions. You may install it manually from:

# install.packages("devtools")


For an example of the usage of the package see this page

You can’t perform that action at this time.