Main scripts

In the following we describe the main scripts of and how they work

Dataset creator

The script dataset_creator.py, when executed, reads the information contained in the configuration files and creates a dataset witht he specified features and time range in a folder in the directory output/.

Relevant parameters are listed in the Dataset creator settings section

To execute, use the commands

venv/bin/python3 dataset_creator.py -c conf/oasi_datasets.json -t MOR -l logs/datasets.log
venv/bin/python3 dataset_creator.py -c conf/oasi_datasets.json -t EVE -l logs/datasets.log

Features selector

The script features_selector.py creates the dataset as before, and in addition it performs a feature selection according to the parameters in the configuration file

Relevant parameters are listed in the Feature analyzer settings section

To execute, use the commands

venv/bin/python3 features_selector.py -c conf/oasi_datasets.json -t MOR -l logs/datasets.log
venv/bin/python3 features_selector.py -c conf/oasi_datasets.json -t EVE -l logs/datasets.log

Final model creator

The script final_model_creator.py, given a set of weights \(\{w_1, w_2, w_3 \}\) in the configuration files, will create the final NGBoost and QRF models. The first is used for the daily single value prediction, while the second for the probabilistic prediction through the quantiles.

To execute, use the commands

venv/bin/python3 final_model_creator.py -c conf/oasi_datasets.json -t MOR -l logs/datasets.log
venv/bin/python3 final_model_creator.py -c conf/oasi_datasets.json -t EVE -l logs/datasets.log