Stroke Predictor

Inspiration

The inspiration for this project came from my grandmother who was diagnosed with stroke a couple of years ago. After her stroke, she was no longer herself and it broke my heart. I always wondered if there was a way to have seen this coming or to prevent it. I later took a machine learning class and created a skin cancer predictor where I learned that it is possible to predict certain diseases and syndromes. This hackathon was a great opportunity for me to explore this and create a stroke predictor based on what I learned with those machine learning courses.

What it does

It uses factors from a dataset found on kaggle to see whether a person is predisposed to stroke. These factors include things in a person's personal life such as residence type, work type, relationship status, as well health factors such as heart disease, average glucose levels, blood pressure, etc.

How we built it

I used a machine learning model: K Nearest Neighbors which uses predictions from the training set of data and plots points with a high correlation to stroke and compares the other points and factors to it. This creates a pattern that can be used on the test data.

Challenges we ran into

Some challenges I ran into were syntax errors and figuring out how to convert string values to binary values. I also did not know which machine learning model would be best to use for it because I only know a handful: K Nearest Neighbors, neural networks, and linear and logistic regression.

Accomplishments that we're proud of

I am proud that I was able to build a working program with a 94.2% accuracy rate.

What we learned

I learned the importance of correlation matrices and showing how the data impacts other factors. I also learned that this is not a straight process and sometimes trial and error is necessary to have an accurate program.

What's next for Stroke Predictor

I would like to create a user interface where the user can input real life data and it will show the user whether they are predisposed to a stroke. The user can then bring up any concerns with their doctor and learn what they can do to prevent it.

Built With

• google-colab
• kaggle
• python