Epidemic growth rate modelling and prediction for current 2019-nCoV outbreak


We are a group of passionate createives and coders who want to help fighting the spread and the consquences of the Spread of 2019-nCov better known as Corona virus or under the disease it is causing - COVID-19.

Problem description

Many problems arise with COVID-19 and the battle against it. We can only address one. One huge issue is the introduction of interventions by the governments of many countries. But what are these interventions really achieving? Also we are interested in external factors influencing the spreading of the virus as well as time series prediction of the near future based on google trends.

Solution description

Prediction Model: We calculated the initial growth rate until the infliction date, where We collected much data and did a prediction analysis for the initial growth rate when the virus starts spreading in a country. This way we can predict the spread for example in developing countries, where the virus has not yet been found.  Additionally we were able to analyse our model qualitatively to see the influence of the different parameters.

By doing a ton of research we were able to put together a unique dataset holding the interventions made by governments and the date they came into effect. This allowed us to expand our previous model and predict the mean growth rate after governmental measures.

Time Series Analysis: We used the google trends api to get time series data of important search words for wach country. A powerful Machine Learning approach then was made to predict case numbers of the next seven days.

what is new about this approach?

  • explain how effective certain measures are through new dataset
  • predict the spread of the virus in e.g. developing countries where we see only few cases today


Data Sources

example data included in the model

  • population of the countries
  • relative age distribution of the population
  • quiality of the health care system
  • Interventions introduced
  • sentiment of citizens about the quarantine
  • ...


This was "only" a 72 hour Hackathon, so we got our Prototype model working. In the future we plan to check further data sources, if there might be variables even more important for our model. Also we already worked on a data flow architecture, which updates our model automatically with the current Johns Hopkins data.

Try It out



data, jupyter-notebook, love, python

Devpost Software Identifier