Comparative Evaluation of Machine Learning Models and Deep Neural Networks for Predicting Cardiac Risk

Heart disease keeps being a major killer all over the world. Early detection really matters for saving lives. This study looks at predicting heart disease risk with deep learning methods. We used the Heart Failure Prediction Dataset from Kaggle. It has 918 patient records. Each one comes with 11 clinical features. First, we did data cleaning. Then encoding and scaling. After that we built a Deep Neural Network using TensorFlow and Keras. It had several hidden layers. ReLU activations too. And dropout layers to cut down on overfitting. We checked the model with accuracy. Precision. Recall. And F1-score. Then compared it to older machine learning models. Like Logistic Regression. K-Nearest Neighbours or KNN. Support Vector Machine called SVM. And Decision Tree. Out of those SVM and Logistic Regression did the best. They hit precision and recall around 0.83. The DNN though had solid recall. It picked up at-risk patients pretty well.