To reproduce:
- download data from link (see Data chapter)
- copy data from csse_covid_19_daily_reports to data dir
- install requirements from requirements.txt (could use Pipenv with pipfile, but for this project I just decided to stick to virtualenv)
- run data_merge.py in order to get a single file for each country with all necessary aggregated info. in addition, file with Latitude and Longitude will be generated containing info about countries coordinates. with this file we'll be able to aggregate info about situation with COVID for each country using neighbouring countries
- go to test_task_solution.ipynb and run code cell by cell
- all custom data flaws correction modules are developed and stored in flaws_correction python package - this is the one that should be used later on in production if needed
The Covid-19 virus was found first in 2019, then spread worldwide in 2020. Initially, most countries weren’t ready for such an epidemic, so even the data gathered contained many flaws in it.
On the other hand, having the ability to predict the number of infected citizens on a daily basis can have a huge impact on decision-making and could save lives.
~7Gb if cloning the whole repository.
Load the first 6 months of the data.
- You’re given raw data that contains many flaws. Please find 5 different flaws in the data and fix them. Try to find the most important flaws that require to be fixed in order to enable forecasting (task 2).
- Please describe (no code is needed) what will be the process required for forecasting the future number of active patients per country.
- Understanding the quality of your coding skills. We do not expect this code to be production-ready, but production oriented. It means we expect organized code that will require minimal refactoring to be used as part of a daily forecasting production code.
- Understanding the way you think, analyze, and tackle data flaws.
- Examine your creativity and ML knowledge.