Breast Cancer Predictor

This project was created for High Tech Hacks 2021

Inspiration

Our inspiration was Dhruv's aunt is a breast cancer doctor, and she has to deal with and diagnose patients all the time. We wanted to make this process easier and more efficient for her, as that means saving lots of lives.

What it does

This program takes in many different measurements, and then it puts it into a machine learning algorithm (decision tree), which then makes predictions of the diagnosis.

How we built it

We coded the original demo in Google Colab. We originally used a logistic regression model, but we dropped that in favor of the decision trees because those metrics were way higher. Then we transformed it into a Jupyter Lab. We downloaded that as an html notebook. We put the notebook into repl.it. And we published the website from that.

Challenges we ran into

We did not know how to connect the back-end machine learning model to a front-end place where it could be seen. We had to troubleshoot this error in many different places, and then eventually we found that we could download it as a html notebook, and went from there.

Accomplishments that we're proud of

We are proud of how high we got the accuracy metrics. Having metrics as high as that (~90% accurate and ~97% precision), could really save lives. This is a great accomplishment to have made.

What we learned

We learned how to use Jupyter Lab. None of us were familiar with that software, so we had to collectively learn how to use it. Also we had to learn the Sklearn decision tree framework. We had never used it before, so we had to use lots of documentation to implement it, and get it's accuracy up.

What's next for Breast Cancer Predictor

We will probably add to the dataset to get the amount of data it trains on up. We will then add to the model more, and probably test other models. This will allow us to pick out the best performing model. Another possible improvement is to make it possible for the user to add the data themselves, and it will make diagnosis for them.