eversvd,+01 (4)

Machine learning is increasingly applied to medical diagnosis to improve accuracy and reduce invasiveness, particularly for cardiac arrest, a leading cause of death. This study develops and implements machine learning algorithms, specifically a feed-forward artificial neural network, to predict cardiac arrest using a heart disease dataset with 11 clinical and demographic features, including age, sex, chest pain type, blood pressure, cholesterol, ECG findings, and exercise-induced angina. The data were explored using univariate, bivariate, multivariate analyses, pair plots, and a correlation heatmap to identify key predictors and their relationships with the target variable (cardiac arrest). Results show that asymptomatic (ASY) chest pain type is most strongly associated with cardiac arrest (54%), followed by non-anginal pain (NAP, 22%) and atypical angina (ATA, 19%), and that males have a higher incidence of cardiac arrest than females. The trained model was deployed as a web application using Streamlit, enabling users to input patient parameters and receive cardiac arrest risk predictions. The system not only predicts the likelihood of cardiac arrest but also suggests possible treatments, medications, and exercise regimens, offering a non-invasive, accessible decision-support tool for early detection and management.