Soteria

Inspiration

Our inspiration for this project was personal experience with the medical system and the inefficiencies and outdated ambulance technology and systems that we have dealt with.

What it does

Our project uses artificial intelligence and an optimized routing algorithm to calculate the severity of an accident and route ambulances both to the scene of the accident and to the closest hospital.

How we built it

We built Soteria by using Flask to connect the frontend website to the backend classification model and routing algorithm. We trained the Softmax model to classify injury severity based on a publicly available dataset and designed the routing algorithm by prioritizing patient injury severity as well as ambulance proximity.

Challenges we ran into

The biggest challenge we ran into was finding a good enough data set in order to train our machine-learning model. Many of the initial datasets we found were either not big enough to train a model on or required too much data cleaning which led to many errors.

Accomplishments that we're proud of

Our simulation successfully works with the inputs that it receives from our AI model. Furthermore, the fact that our model saves over 100 more lives than the traditional ambulance routing system when simulated on a sample of 5000 patients is a major accomplishment.

What we learned

Through this hackathon, we learned to merge the front-end and back-end portions of our application using the Flask framework. We also learned to improve upon a basic classification model using hyperparameter tuning and optimization methods such as temperature scaling.

What's next for Soteria

In the future, we plan to expand beyond Chicago by aggregating information about hospitals in other regions across the country. We also will look into finding a more thorough and updated dataset that includes information about more recent crashes.