This project uses Keras LSTM to produce a feed forward neural net forecast of daily cases.
Navigate to the root diretory
Run the following:
pip3 install -r requirements.txt pip3 install ./covid_forecast
Open a python3 prompt and run the following:
import covid_forecast as cf cf.run_daily_stats(perspective='global' ,train_sample_size=0.8 ,series_type='Infections')
The perspective parameter may be either 'global' for global daily statistics; 'vic', 'nsw' or 'qld' for daily satistics for these states in Australia. The training sample is a value > 0 and less than or equal to one. If this value is set to 1 all observations are used to train the forecast. This is the percentage of observations held out for training. The most recent observations are used for testing the fit. The series_type parameter determines the series to be forecast and can be either 'Infections' or 'Case Fatality Rate'. By default this program "looks back" 10 days in training the model and forecasts forward 7 days. These values will be parametised in future development.