covid-dl

Machine Learning models for COVID-19 risk prediction

WARNING Keep in mind this is a project still in development, and the predictions of the models are not always reliable. Use at your own discretion.

We provide a simple Google Colab notebook so that you can have a quick glance at the results of the model without having to install anything. It will show different metrics of how well the different models we trained performed. Just run it and play with the Dash graphs:

Usage

Requirements

Install the module's requirements:

pip install -r requirements.txt

Data fetching

Copy the provinces-incidence-mobility.csv file (obtained from the covid-risk-map repo) to the data/raw folder.

Alternatively, you can download the original dataset we used to perform the analysis:

wget -O covid-dl/data/raw/provinces-incidence-mobility.csv https://api.cloud.ifca.es:8080/swift/v1/covid/provinces-incidence-mobility.csv

Training a model

If you want to train a model on your data run:

python train.py

If you already have a trained model (in models/feedforward) you can go directly to the prediction step.

Predicting future incidence

To make a prediction into the future from the last available date run:

python predict.py

This will write a predictions.csv file in data/predictions.

To view the results interactively, run a plotly instance with:

python visualize.py