Automatic Triaging for the average cautious citizen

Instead of extra work on medical workers, this little Python project can do a decent job of triaging the average citizen and helping them decide whether they should come to the hospital or not

Comment

Inspiration

I was leaving school for the quarantine. On the ride home, I go through the eastern end of the Rust Belt. I'm mildly allergic to dust, smog etc.,. Fast forward to my 3rd of quarantine and I'm coughing. I didn't know why. I was worried that I might've caught COVID-19. But, I kept a close watch on my health and turns out my cough was down to the smog and dust. I was hesitant to run to the nearest hospital because the last time I went to the hospital with a minor ailment I got slapped with a massive bill from the hospital. Luckily, my family's insurance covered it. So, for people in my situation, I made an automatic triaging software.

What it does

It does categorizing like the triaging process based on number of symptoms, combinations of symptoms.

How I built it

I used all the machine learning frameworks in Python to do a machine learning task called classification. I randomly generated data and simply changed the target variables/decisions to match the recommendations of the WHO. I ran the classic k-nearest neighbors algorithm on some randomly generated data. I also repeatedly generated random data and tried out neural networks of varying input sizes and layers. The network with highest accuracy is the one to be chosen.

Challenges I ran into

Since I don't have access to comprehensive patient data with summaries, I don't know all the ailments of every COVID patient. Since I don't have all the ailments, I may not take all symptoms into consideration. So that's another small exercise in Bayesian Networks to confirm what ailments are dependent on the disease and what ailments are irrelevant. I also didn't have data so I tried to randomly generate it. The Data Generating Process was also buggy. If someone can debug it, the code will be a fully functional prototype that follows current WHO guidelines.

Accomplishments that I'm proud of

I learned a lot about the optimizations in machine learning and took a step up from a junior college student in CS to someone who truly dives into the unknown, applying what I know, learning new things on the road, all to answer questions.

What I learned

What's next for Automatic Triaging for the average cautious citizen

Built With

Share this project:

Updates