M5 Modelling
Background
Please look at the Kaggle competition for background! There is the Accuracy Competition and the Companion Competition for uncertainty.
Please also check out the Competitor's Guide.
This repository also tries to estimate elasticities and other effect sizes, rather than just focusing on accuracy.
Setup
-
Create a conda environment called
m5-comp
:conda env create
-
Then, install the library in editable mode:
pip install -e .
-
Set up your Kaggle API credentials, as outlined here.
-
Download and unzip the data:
kaggle competitions download -c m5-forecasting-accuracy -p data unzip data/m5-forecasting-accuracy.zip -d data
-
Run jupyter.
Runnin in Julia
-
Install Julia.
-
Activate the conda env, then run the following to install the required Julia packages:
]activate . instantiate
Backspace to exit package mode, then install the IJulia (Jupyter) kernel:
using IJulia installkernel("julia4threads", env=Dict("JULIA_NUM_THREADS"=>"4", "JULIA_PROJECT"=>pwd()))
-
Run
jupyter notebook
and select the kernel you want (either default or 4-thread kernel).