Re-estimated forecast – up to 250,000 infections and slow down around April 10th. Forecast of coronavirus epidemic in the US – ExMetrix model March 30, 2020 – Posted in: analyses, carousel, lifestyle, Our forecast, posts, Uncategorized

The model developed by the ExMetrix, company that deals with economic and social forecasting, shows that the number of people infected in the US can grow up to 250,000 cases and should be reached around April 15, 2020.

Previous version of model was prepared when number of infections reached 43k and suggested up to 160k of infections. Model had to be re-estimated mainly because mobility of citizens assumed more enforced social distancing, which did not take place, and this parameter had most significant impact on error.
“Lack of strong social distancing enforcements is visible especially when comparing number of infections on tests performed. In Poland this number is 1 infection on 25 tests in US 1 on 7 tests!” – comments Ryszard Lukos.

The model therefore suggests that the duration of the US epidemic will generally be longer than in China, but its dynamics will have a different character. After a very long period of relatively small number of cases, there is a period of rapid growth with a much greater slope than it was in China. The fastest growth of infected people can be as high as 10,000 cases per day and will occur in the period from 03/22/2020 to 04/03/2020. After April 3rd, the number of cases should increase at a slower rate, while very clear inhibition should be visible around April 10th.

How the prognostic model was created

While building the epidemic model for the US, the course in countries where COVID-19 developed earlier than in the USA, i.e. in China, South Korea, Japan and some European countries was analyzed.

The following information areas were included in the construction of the virus epidemic model:

  1. The pandemic to date in other countries
  2. Meteorological factors
  3. The state of health care in a given country
  4. The state of health condition of the society, with particular emphasis on risk factors such as oncological, cardiological diseases, diabetes and obesity
  5. Demographic structure of society. We especially paid attention to the percentage of elderly people
  6. Level of restrictions applied by the authorities and compliance by citizens

In the case of the US, it turned out that overweight or obesity among both men and women had a significant impact on the results (the highest percentage among the countries studied). The result is the lowest assessment indicator of the countries surveyed, + 0.31 (for comparison, this indicator for Japan is +0.494, for South Korea +0.448, and for China +0.43).

Simply put, a low indicator value means a higher estimated public susceptibility to infection. Of the countries surveyed, only the United Kingdom has an indicator similar to the USA. For most European countries, it ranges from +0.35 to +0.38.

The model was built using artificial intelligence technologies (including, but not limited to proprietary neural networks and simulated annealing algorithms). This technology allowed to estimate the impact of information from areas listed above on such parameters of the infection curve as:

  • slope – reflecting the increase in the number of infected at a given moment
  • spread – allowing to determine the maximum ceiling for the number of cases and the duration of the epidemic

Assumptions, risk of verifiability:

When building the model, we assumed that statistics in COVID-19 affected countries were not concealed or artificially modified.

Source of data on the number of infections:

Johns Hopkins University

https://github.com/CSSEGISandData/COVID-19/tree/master/csse_covid_19_data/csse_covid_19_time_series

Data sources describing individual areas included in the model:

OECD, World Bank, http://population.city/ , https://www.accuweather.com

As the situation develops and new data flows, especially from countries where the epidemic started earlier than in the US, we will update the model and take into account any additional information to make it as precise as possible.

ExMetrix is ​​a company creating a new generation of software for statistical data analysis and forecasting. ExMetrix is ​​a data analysis system using the latest statistical and numerical models based on Machine Learning and Artificial Intelligence. ExMetrix gives access to ready made predictive models and allows to create own models based on 80 million data streams available in the system or any of data sets provided. Dozens of organizations of various sizes use ExMetrix, including Azoty Group. The software is also used to educate students at top Polish Universities – UEK and UMCS.

Ryszard Łukoś, Chief Analyst at

ExMetrix

« Re-estimated forecast for US shows that restrictions were introduced to late
160,000 infections and slowdown around April 10th. COVID-19 ExMetrix forecasting model for the US »