This repo contains the final project (currently under development) for my college degree as Bachelor of Production Engineering.
It consists of a Long Short-Term Memory Neural Network to predict average energy load of the next 5 weeks individually.
The operative week for ONS (Nacional Operator of the Interconnected System) begins at Fridays and ends at Thursdays.
Fisrt, tou need to install the dependecies. You can do it with the following command on the root folder:
pip install -r requirements/requirements.txt
This projects is structured using DVC. You can control the preprocess, featurize and model hyper-parameters through the params.yaml file. To run the worlflow, simply alter params.yaml at will and type into CLI from project root:
dvc repro
DVC will check if the dependecies declared for each step of the pipeline in dvc.yaml changed vs last run and re-run only the necessary steps. My experiments are tracked through MLFlow and registered in the "Experiments" tab in dagshub project repo.
Validation visuals are stored to valuation folder.