RFC4Heart

Correct Result vs Predicted Result
Training Cost and Validation Cost

How to use

Upload the heart_processed.csv provided by the hackathon organizers in Google Colab.

Inspiration

Heart disease affects over half population globally, yet early detection remains a major challenge.

I was inspired to leverage Random Forest Classifier (RFC), which have demonstrated state-of-the-art performance.

The most critical challenge was handling the outliers in the features.

What it does

RFC4Heart is an automated Heart disease detection system that analyzes heart reports and classifies disease stages with 86% accuracy.

Key highlights:

Handles extreme class imbalance and outliers
Achieves a perfect F1-score of 0.86+ in all test cases
Provides fast inference along with confidence scores for clinical decision support
Designed to be deployment-ready for real-world hospital integration

How we built it

I used Scikit Learn's Random Forest Classifier to train on the heart reports and fine tuning the parameters.

The model was trained on 800+ heart reports using Scikit Learn.

Challenges we ran into

The major challenge was the model high cost convergence that is stuck on 86% with all optimizations.

Accomplishments that we're proud of

The accuracy of 86% is a very good base line for the model that classifies heart disease.

What we learned

Good to know that this was my first Random Forest Classification even my first ML project, and for this project i learn in 5 days.

What's next for RFC4Heart

With a good base line model, I think to try tuning the model and testing different models to check for any advantages and make it a user friendly model by implementing in a website.