GCN Network Optimization for COVID-19 Predictions

This project optimizes a Graph Convolutional Network (GCN) model with Optuna to predict COVID-19 cases in various regions, incorporating additional features alongside historical data for improved accuracy.

Table of Contents

• Dataset
• Prerequisites
• Installation
• Usage
• Contributing
• License

Dataset

The primary dataset "confirmed_cases_by_region_and_date.json" contains daily confirmed COVID-19 cases across multiple regions. Complementary datasets for population demographics, hospital resources, and water supply information are also utilized as features.

Prerequisites

Ensure the installation of the following software:

• Python 3.9.4 or above
• Pyenv 2.3.17 or above

Installation

After installing Python and Pyenv, clone the repository and install the necessary Python packages. Use the commands below:

# Create a virtual environment (optional but recommended)
pyenv virtualenv 3.9.4 gcn
pyenv local gcn

# Clone the repository
git clone https://github.com/RianBrug/gcn_for_covid_sc_brazil.git

# Navigate to the project directory
cd gcn_for_covid_sc_brazil

# Install the required packages
pip install -r requirements.txt

Usage

Prepare the dataset: Ensure "confirmed_cases_by_region_and_date.json" and additional feature files are located in the "assets" folder.

Activate the virtual environment (if applicable):

pyenv activate gcn
Run the optimization:

python gcn_network.py
or
python gat_network.py

Analyze the results: Optuna Dashboard

Optuna comes with an interactive dashboard, which provides a rich interface for visualizing the optimization process. This can be very useful for understanding and interpreting the model's behavior and performance.

To launch the dashboard:

1. Start the Optuna dashboard server with your optimization database:

   optuna dashboard --storage sqlite:///gcn

   Replace "example.db" with the name of your SQLite database file that was used for storing the optimization results.

2. Open your web browser and navigate to the displayed URL (usually localhost:8080). The Optuna Dashboard will appear, displaying a variety of interactive plots about your optimization process.

Please note that the Optuna Dashboard is read-only and does not support modifying the database. Always remember to save and backup your database.

Contributing

Contributions to the project are always welcome. Feel free to submit bug reports, feature requests, or pull requests.

License

https://creativecommons.org/licenses/by-nc/3.0/br/deed.en This project is licensed under CC BY-NC 3.0