What inspired us.
Countries are slowly opening up economies after a long lockdown, but are not very sure about the risks involved due to the dynamic nature of people movement. We have used google mobility dataset along with "Our World in Data COVID-19 Testing dataset" to generate a risk and recovery index of any particular country.
What it does
Source Datasets used as below:
Methodology: It Aggregates source data at a daily country level and consolidates as master data Calculate risk index based on the rate of infection and the total number of infections Train machine learning model to translate the relationship between risk drivers (mobility and/or restriction) and risk index Prediction Scenario:
To prevent the second spike in the recovery phase, one would be able to predict the risk index given the assumed risk drivers therefore to adjust government policies and countrywide communications Risk Index Formula:
rate of infection: the increase/decrease rate of current total cases against total cases from 14 days ago (normal virus incubation period) number of infection: current total cases per million The formula refers to the UK government risk index guidance: https://www.spectator.co.uk/article/how-number-10-should-illustrate-its-covid-alert-formula
Prediction Model Build:
It makes use of a classification model to capture the relationship between risk drivers (mobility and/or restriction) and risk index to future prediction that aids recovery phase planning
How I built it
Risk & Recovery Index generator App is powered by python notebooks running on Azure Databricks using datasets from Our World in Data COVID-19 Testing dataset and Google Covid19 Community Mobility Reports. We used Azure Databricks to run our notebooks attached to a Spark cluster. Pushed notebook analysis outputs to an Azure WebApp services python site. On the web app, a user can input destination location to get Risk Index. User can use another field to get destination flight information.
Challenges I ran into
Webservices do not work as intended in a python flask environment. Runs fine locally.
Accomplishments that I'm proud of
We not only managed to get a brilliant analysis done based on datasets available but made it interactive with a web app running on Azure web services. Risk index calculator is our USP. We created a formula based on R-value calculator.
What I learned
How to configure Azure Databricks and how the spark clusters power the notebooks for machine learning. Notebook integration with a web app using react and flask. Python coding.
What's next for OpenConsulting Team.
We plan to put Risk and Recovery Index to production by creating a Chat App for Travel, Human resources and retail sector powered by Azure Databricks.