COVID-19 Growth Forecast
Web UI (data is updated daily): https://app.katanaml.io/covid19/
Katana ML Docker container with backend code: https://hub.docker.com/r/katanaml/light-runtime
Article I: https://bit.ly/2XHjL4P
Article II: https://bit.ly/2RMB14V
About: The forecast is based on two equations: Logistic and Hill. Both generate sigmoid, Hill in most of the cases generate a better result, it is producing not symmetric sigmoid. Backtesting is done for both methods by forecasting for five days in the past. If the backtesting forecast is above the current forecast, means the current situation is better than it was forecasted. If the backtesting forecast is below the current forecast, means the current situation is worse than it was forecasted..
Technology: Python, Prophet
Author: Katana, Red Samurai Consulting, Andrej Baranovskij
covid19_endpoint.py - Flask endpoint
covid19_scheduler.py - Re-training scheduler
covid19_model.py - model construction
(POST) curl --data "country=Lithuania_cases" https://app.katanaml.io/katana-ml/api/v1.0/forecast/covid19
(GET) curl https://app.katanaml.io/katana-ml/api/v1.0/forecast/covid19/countries
(GET) curl https://app.katanaml.io/katana-ml/api/v1.0/forecast/covid19/stats
Licensed under the Apache License, Version 2.0. Copyright 2020 Katana ML. Copy of the license.