The problem Project Manna solves

The Problem

The world is racing at an immense pace to have a vaccine ready for the nation and the world as a whole is working parallel to combat the deadly COVID-19. In the initial days after the vaccine is ready, it is unexpected to have its volume sufficient enough to vaccinate the entire population at one go. Strategic and wise decisions need to be made using the existing data to choose the right places and individuals to vaccinate initially.

Our solution

• We propose a model-based prediction system that would help in choosing the optimal distribution strategy for a vaccine.
• Our Model takes the standard and widely accepted SIR model for disease spread and combines input from time series Covid 19 India data and the district wise population statistics data from Census India to form a hybrid model.
• This model enables the prediction of new cases that can arise in any district in the upcoming 15 days.
• Using the risk of any district and the number of vaccinated people there(Assuming ongoing vaccination program), we compute a comprehensive parameter called vaccination priority for the district.
• These real-time priority values enable us to determine what amount of vaccines to be given to each district for a given number of vaccines we have for the country.
• With a real-time visualization app, any public/private organization can use this to determine the best priority order for effective delivery of vaccines

This is a race against time and our model is based on predicting the possible future scenario to be to make vaccine accessible to places that would at most risk in the upcoming days.

Challenges we ran into

• We are focusing our project primarily on India. With such a wide demographic, gathering data that represents the entire nation. There were different data sources available but bring all data into the same format and cleaning up data was difficult.
• The right prediction model for subsiding of COVID after vaccination, as it is a purely mathematical model and there is no sure way to verify the model without actual vaccination.
• To have the prediction in real-time so that the organization can take into all possible turbulences in future covid-19 situations be it spike in cases, or ineffectiveness of vaccine in initial stages.
• To integrate a real-time visualization tool using d3.js which can help cluster nearby districts with higher risk.