Heart disease prediction is crucial for effective clinical management, and therefore to address the limitations of conventional approaches, researchers are actively building predictive models using artificial neural networks and machine learning techniques. These models can analyze vast medical data, offering comparable accuracy to traditional methods while expediting diagnostic processes. By leveraging extensive medical datasets, machine learning algorithms can automate intricate data analysis while continually refining their predictive accuracy. They show promise in navigating the complexities of heart disease diagnosis and improving patient outcomes.
This study investigates the development of such a deep neural network model, employing a classification approach within a machine learning framework to identify heart disease. The classification goal of the study is to predict whether the patient presents heart disease or not, leveraging PyTorch capabilities to set up a Deep Neural Network.
High precision is desirable as false positive predictions would be costly or undesirable, such as unnecessary medical interventions or treatments. High recall is also important as false negatives (missed diagnosis) are costly or unacceptable, and high recall ensures that the model identifies as many cases of heart disease as possible.
Given the importance of both precision and recall in the context of predicting heart disease, the f1 score, the harmonic mean of precision and recall, will be investigated as it provides a single value that combines both measures.