Inspiration

In light of the global impacts of COVID-19, we decided that we would want to figure out how we could train a computer to give us data on how the world would be impacted by this disease, in hopes of using this data to best prepare hospitals and health professionals for the future cases of this disease.

What it does

Our project creates a model of what COVID-19 will look in the upcoming weeks, based off LSTM Neural Networks. We trained this data with data merged from Johns Hopkins and the NOAA, which we then visualized with heatmaps in a Vue app, using the Google Maps API.

How we built it

Training the ML was done by tensorflow and scikit in Python, which took a considerable amount of time due to the large amounts of data taken in. We used Recurrent Neural Network algorithms in order to train the computer to factor in current events, population density, temperature, location, and COVID - 19 transmission frequency. The approach was to gather as many features as possible in order to make up for the lack of data, as the COVID-19 pandemic is still ongoing, making it hard to track.

Challenges we ran into

Lack of data made it hard to train the Machine Learning algorithm, so we had to navigate that very carefully. Mining data was also a problem.

Accomplishments that we're proud of

We're proud of implementing the LSTM Neural Network and having sensible results, as well as the unique way in which we implemented the algorithm - we took into account weather, as well as current COVID-19 statistics, in order to assess the conditions in which people would be subjected to during the virus outbreak. We also are happy that the heatmap is able to render and work well for people to actually see the data.

What we learned

From a technical standpoint, we both gained tremendous experience in ML, as well as app building. From our project, we learned that ML can be quite a useful tool in creating models - though we aren't sure how much we should trust our model just yet, it makes quite a bit of sense. In our models, China doesn't grow nearly as fast as other countries, which makes sense given what is going on, and the United States is growing fastest, which also makes sense given our sociopolitical environment.

What's next for On the Future of COVID-19 - ML and Data Visualization

In the future, we hope to factor in more data and features in our ML model. We also hope to create a website that has a dynamical heat map, that is able to show the virus in specific areas, so people can check where they live to see the model in action.

Built With

Share this project:

Updates