COVID-19 Policy Recommender

Inspiration

COVID policy decisions are typically tainted by public opinion and personal desires. Even rich and well established governments have had trouble containing the spread of COVID-19. This AI model is able to make policy decision based on what has been optimal at lowering infection rates.

How I built it

This Neural Network was coded from scratch in java using only a matrix multiplication library. Nine data points are taken to represent a country's given COVID situation. These input values are scaled between -1 and 1 and are then fed through the network.

The loss function is computed using infection rates 10-14 days after a given policy is enacted. This optimizes the network to minimize infection rates through its policy decisions.

Challenges I ran into

I had to spend the majority of time formatting data. It would have liked to give the model more training time.

Accomplishments that I'm proud of

Our neural network was successful at general tests like XOR and MNIST. Our neural network was able to make sensible policy recommendations based on a county's COVID data.

In the future I want to continue running tests to gauge the models accuracy. I also would like to train it for longer and with more data.

Code can be found at: https://github.com/ttenneb/Covid-Policy-Recommender