Ended 1 year ago
560 participants
902 submissions

## Materials (58 MB)

 2019 Novel Coronavirus COVID-19 (2019-nCoV) Data Repository by Johns Hopkins CSSEhttps://github.com/CSSEGISandData/COVID-19 55 MB countries.csvCountries list 1 MB sample_submission.csvSample submission 1 MB russia_regions.csvRussia regions list 1 MB

Input data source 2019 Novel Coronavirus COVID-19 (2019-nCoV) Data Repository by Johns Hopkins CSSE. XXX_time_series folders contain tables with information on confirmed cases, number of deaths, and recoveries by country/region. Note that the data is not entirely reliable; it is collected from different sources, including officials, press, etc. Solving the problem of real-time forecasting requires using the data as it is. Therefore the participants need to take into account all factors: sample data properties, data obtaining scheme of features of the states, measures taken, etc.
Please note that any available data can be additionally used if it provided that references to their sources are published in the contest official channel.

Update 14.04:
According to the second and third stage rules, the Russian regions list available in russian_regions.csv. This list contains data about the region's population, key - iso_code column.
Also, these regions were added in the sample_submission.csv file.
You can use https://github.com/grwlf/COVID-19_plus_Russia and COVID-19 Russia as data sources.

### Format of solutions

The participants must build a forecast for all countries on the list for each day, from the starting date of the competition to the end of the year. The forecast for each day must include the number of confirmed cases and deaths (total as of that day).

The forecast starts on day X+1 (X is a submit day), not taking into account previous dates. The leaderboard will be updated and include the results for the last week. The final leaderboard will take into account the submissions, including the full periods specified in the rules.

Submission example:

date,region,prediction_confirmed,prediction_deaths
2020-04-05,AFG,396,7
2020-04-06,AFG,449,8
2020-04-07,AFG,510,10
...


### Evaluation metrics

Mean Absolute Logarithmic Error - average module deviation of logarithms of predicted cases from real values, calculated as follows:

$$\text{MALE} = \left| \log_{10}\frac{\text{predicted}+1}{\text{actual}+1} \right|$$

Cookies help us deliver our services. By using our services, you agree to our use of cookies.