COVID-19 Vaccine Distribution Model

Imagine, it’s December 2020. There are long winded queues in front of the pharmacies and vaccine distribution centres. The Australian government only has a very limited number of vaccines. The population is desperate to return to normality and the vaccine is the only key to achieving this. For the government to reach the minimum requirement herd immunity needs to be reached. With a population of 25 million and barely 15 million vaccines. How will the government succeed in effectively distributing the vaccines?

Fortunately, a small team called the Hacker-Knights come to the rescue.

A novel way of achieving herd immunity and a simple prediction tool with intuitive visualisation filters, that will assist the government in taking control of COVID-19 in the most efficient manner.

The idea is to achieve below outcomes with the government data:

  • Forecast for the covid impact for future months with the existing covid factors (e.g: cases on daily basis, confirm, recovered, deaths and infection rate etc.)
  • Predict the minimum number of vaccines required to achieve herd immunity of a particular region. We considered R-nought, a variable factor, as 2.2 (average according to available sources) hypothetically for the limitation of data and time constraint.
  • Find out the optimal way of distributing available vaccines to the most affected regions based on multiple factors like business impacts, available distributed channels, social, economic conditions and also to achieve herd immunity across the region as fast as possible.

Data Story

The datasets used come from various sources including Australian Bureau of Statistics, Github open source data and Australia Population (2020) Worldometer. One of the major challenges has been to extract precisely the most relevant data by combining different unstructured datasets into a single effective aggregated dataset that has been normalised. Having the monthly statewise covid cases, deaths and population we extrapolated the number of vaccines needed for each region and improved the prediction efficiency using factors such as employment and unemployment. An interactive visualisation web dashboard(shown in the video) has been built for users.

This model takes into consideration covid cases that have been confirmed and deaths per state over a time period to calculate the R naught value. These cases have been forecasted upto march 21 using multiple linear regression.

After calculating R naught found to be 2.5 on average. We calculated the number of vaccines needed to reach herd immunity for each state according to its population. Formula for Herd immunity: 1/(1-Rnaught) We have run this algorithm to our forecasted range of Mar 21.
If we get limited number of vaccines, the algorithm calculates the optimal distribution of vaccines to take smaller steps towards herd immunity (no more a pandemic).

Future implementation could use more features(density, social distance, house rents, food habits) to train a Neural Network model for the weighted and variability factor to achieve the real time R-naught more accurately to eradicate COVID-19.

Team

Knights who served for the hack

Nazmul

Naga

Mahesh

Afrida

Quang

Jan

Team DataSets

Datasets that were used

Australia employed and unemployed stats

Australia Demographics data and population rate

Australia Demographics data

Covid 19 cases for Australia

Frequently Asked Questions

For NSW the map displays 4,934,944 vaccines are needed, but how to achieve herd immunity, when we have 4 million vaccines only for the whole of Australia now?

According to the map NSW would be able to achieve herd immunity over cumulative months initially requiring 1,421,249 vaccines(See label with “number of vaccines available”). This helps the Government to rapidly detect the area to respond in an emergency.

We’d like to implement the tool outside of Australia. Whom to contact?

Hackerknights would be extremely happy to assist you. Please contact Nazmul (team leader) through LinkedIn for any questions regarding or expanding the project.

How accurate is Covidvaccineready’s prediction?

Currently its as accurate as any machine learning forecast but the accuracy can be strictly measured once we have the real number of covid cases until March 2021.