A modeling platform for generating and implementing ML models for bathing water quality predictions

Platform for project Digital Water City

In the project Digital Water City (DWC) a platform will be developed that supports bathing water managers with setting up machine learning models for bathing water quality prediction.

This is a Django app. Django is a web development framework for Python. To run the app you need to have Python and you need to have the Django framework installed.

The following instructions were tested on Ubuntu Ubuntu 20.04.3 LTS (found with lsb_release -a).

Create private and public SSH key files

ssh-keygen -t rsa -b 4096 -C "your-email-address"

Press three times Enter:

1. Accept the default location for the files,
2. do not use a passphrase,
3. confirm not to use a passphrase.

Copy the output of the following command to the clipboard:

cat ~/.ssh/is_rsa.pub

Login to github.com and open https://github.com/settings/ssh/new. Give a "Title" of your choice. Paste the content of the clipboard into the text field "Key" of the form. Submit the form by pressing the button "Add SSH Key".

Change to a directory into which to clone this repository.

Clone the repository from GitHub into that directory.

git clone [email protected]:wseis/swim-ai.git

This should create a folder swimai in your current directory.

sudo apt-get install python3-pip

Virtual environments allow each project to use its own set of required packages. There are different solutions for managing virtual environments. I found two solutions:

Install Python package "virtualenv"

sudo pip3 install virtualenv

Create a new virtual environment

virtualenv venv

Activate your virtual environment

source venv/bin/activate

(To deactivate the virtual environment, run deactivate venv.)

Let the administrator install conda, e.g. to /opt/miniconda3.

Initialise the usage of conda

/opt/miniconda3/bin/conda init bash

List available conda environments

conda info --envs

Create a new environment with the python version you want

conda create --name dwc python=3.10

Activate the new environment

conda activate dwc

(To deactivate the environment, run conda deactivate)

The Django web application depends on many Python packages each of which is listed in the file requirements.txt in the project folder. The command pip3 allows to install all required packages from that file at once:

pip3 install -r requirements.txt

• GDAL_LIBRARY_PATH (on Ubuntu not required)
• CLIENT_ID
• CLIENT_SECRET
• REDIRECT_URI
• KEYROCK_ADMIN
• KEYROCK_PW
• APP_ID
• AUTH_BASIC

Run the following command to create all required tables in the database:

python manage.py migrate

Some tables need to be populated with data.

This can be done via Jupyter notbooks. They can be seen as tutorials containing text and Python code.

Run the web application Jupyter with

python manage.py shell_plus --notebook

Navigate to the notebook

notebooks/01_CreateFeatureTypes.ipynb

and run all of its code blocks. This creates all "feature types".

Stop the notebook server with Control-C.