Heart Stroke Prediction Model

Inspiration

I lost my maternal grandmother to stroke. The thing about the condition was she was perfectly fit but she had just been complaining about an unbearable pain in her left arm- a very clear indication of stroke. The cherry on the top was- everyone in my family is from a medical background, but still the clear indication was not visible to us. My mom states" we never thought this can happen to our mother. " It was emotion that blinded diagnosis," but if we eliminate emotions- it can save many lives. One of them could have been my grandmother

What it does

My model takes some features:-{gender, age, hypertension, heart_disease, ever married, work type, Residence type, avg_glucose_level, bmi and smoking status} and uses Logistic Regression classifier to predict if stroke occurs or not.

How we built it

We took a dataset from of size 5100 from Kaggle . First I read the dataset using python and then removed outlier values using z-score. After that, I encoded all the values so that the machine can easily read it. After data was encoded I use Logistic regression to classify if the person is likely to get a stroke or not.

Challenges we ran into

This was my first ML project with a dataset of this size, I wasn't getting a accuracy score greater than 0.5 in my first few tries and it seemed like a dead end but then I tried something out of the way- never tried it any other models- removing outliers and my accuracy boomed to a 0.95. Also encoding the data columns was a task.

Accomplishments that we're proud of

Firstly, this was my first 100 line code so that is what I am the most proud of. Other than that, I wrote this code alone and was able to follow the design lifecycle for the first time, it helped a lot. Lastly, with this high accuracy score- I look to propose it to a cardiologist, so they can save lives

What we learned

I learnt how following a step wise process is important to reach the destination. Alongside that I learnt how thinking simple can applying your basics is 99% of the times the solution to a problem you are facing

What's next for Heart Stroke Prediction Model

I want ta build a better user interface for it , so it becomes easier for people to use. Then I want to propose it to a cardiologist in my town. Along with that, I can propose it to my local municipality, so they can equip this into their apps, and can be used by the PHC-public health center to give checkup to people who can not afford cardiologist consultancy.

Built With

• python