This repository is for the Group Project developed as a part of the Advanced Machine Learning Course pursued in the Masters program in Business Analytics from The University of Texas at Austin, McCombs School of Business.

This project was built by Nittala Venkata Sai Aditya, Disha Gandhi, Sai Bhargav Tetali, Vishwak Venkatesh and Soumith Reddy Palreddy

The goal of this project was to be in a position to classify a new heart beat sound as abnormal or normal. In the case of an abnormal heart sound, the person should be able to get a quick diagnosis. This allowed us to work heavily on audio data and learn more about librosa and audio signal features. We implemented a lot of machine learning algorithms from Random Forest to CNN and made our observations on the accuracies given by them.

The dataset used can be found here: http://www.peterjbentley.com/heartchallenge/index.html. It is also present inside this repo. The major challenge this dataset gave us was trating heavy class imbalance in training data. Hence we have recorded the result in both scenarios and played a lot with upsampling to deal with the same.

You can find more information about the project on Medium : https://medium.com/@vishwakvv29/heartbeat-audio-classification-using-machine-learning-305e568efecc