Heart disease prediction using deep learning
A college project where we built a deep learning model and web app that predicts heart disease risk from basic health details
Business Challenge
Heart disease is a serious problem, and spotting it early helps patients get care sooner. We wanted a simple tool that estimates risk from details like age, blood pressure and cholesterol. It is a college project, not used in real clinics.
Solution Delivered
The project was built by a neural network in TensorFlow/Keras on the UCI Cleveland Heart Disease dataset. We cleaned and scaled the data, trained the model with Dropout and Early Stopping, and checked it with accuracy, precision, recall, F1 and ROC-AUC and also tested it with many different patient inputs. The model is connected to a web app that gives a prediction from the details a user enters. Source code: https://github.com/pranathi006/Heart-disease-prediction
Outcomes Achieved
The model can predict heart disease risk from 13 clinical details with 91 percent accuracy on the test set. We checked it with accuracy, precision, recall, F1-score, ROC-AUC and a confusion matrix, and tested it with many different patient inputs. The model is connected to a web app that gives a result straight away. This is a college project and is not used in real clinics.
Measurable Business Outcome
Reached 91% accuracy and also ROC-AUC on a held-out test set from the UCI Cleveland dataset. Early Stopping and Dropout helped keep the model from overfitting. The web app gives a prediction instantly from the details a user enters.