m5 forecasting data extractor
The dataset comes from the Kaggle Contest.
Remember to cite if you would like to use. Ref paper "The M5 Accuracy competition: Results, findings and conclusions".
The repo is to store the data extractor, which outputs the data InventoryManageEnv used from the raw data.
git clone https://github.com/Wu-Jingcheng/m5_data_extractor.git
cd m5_data_extractor
(unzip sales_train_evaluation.zip)
(modify the config or just leave it there)
python data_process.py
(enter `yes` to acccept, here we go)
Please see the data_process.py
for more details in comments at the config
bar.
The config here is essentially for filter the data and extract data.
Run the script and enter yes
as input, data and inventory_config.py will be generated.
Essentially, this script looks for items that meet the specified requirements, and extract a certain number (specified) of them to the test dataset.
For the unpicked remainder, they goes to the train dataset.
The directory will be like:
|root
|inventory_config.py
--|data_patch
----|test
------|sku_selection.txt (info that tells you what items are present here)
------|store{i}.csv (testing dataset)
----|train
------|sku_selection.txt
------|store{i}.csv (noted that training and testing datasets are different)