This self-explaining commented notebook (paired with example .csv datasets) has been shared as Jupyter notebook (tested in Google Colab and linked to Gdrive folder but can be easily adapted to run in other platforms).

The notebook is structured in steps and guides the user in setting up a forecast system to predict next value in a time series of discharge (TARGET variable) using a few input features (Please refer to the example datasets available on the github page).

Main steps encoded in the notebook are:

- DATA READ AND PREPARATION – loading the dataset with target and input features, performing basic operations such as filling missing values, normalizing the variables, dealing with the time information.
- FEATURE DROP to remove uninformative features from database, this can be done iteratively after running the ML models to identify most informative features
- MODEL Tuning to test and tune many ML algorithms and identify the most performing one, also to identify informative features, ( in this example exploiting H2O AutoML libraries), please refer to H2O documentation (https://docs.h2o.ai/h2o/latest-stable/h2o-docs/automl.html) if you want to fine tune the parameters and use different models
- DEPLOY MODELS to save best performing models for operational deploying
- LOAD MODELS to load a saved model and run it over a test dataset

Example datasets are provided ( available on the github page "Data" folder):

Discharge.csv ,  2 weeks dataset (default dataset for the notebook)

Discharge_m.csv  ,  1 month  dataset

daily_master dataset.csv  ,   daily dataset of the target and input variables as retrieved 

Once trained the forecast algorithm  can be deployed as a service, an example is provided as  AI4EUexperiment as well as Docker container 

On the github page ("Retrieval" folder)   also the python functions for data retrieval have been uploaded.

This asset has been developed in the 1st Open Call of I-NERGY (This project has received funding from the European Union's Horizon 2020 research and innovation programme within the framework of the I-NERGY Project, funded under grant agreement No 101016508)