Early Prediction of Heart Failure via Supervised Machine Learning Models: A Performance Analysis

Abstract

Heart disease is one of the most important illnesses face today, and most of its victims die. Diagnosing and treating heart disease is no easy task. Being precise and efficient is essential for this difficult diagnostic procedure. Prevention of mortality was achieved with the early diagnosis of cardiac illness. The rising incidence of cardiac problems has made the prediction of their onset one of the most difficult medical jobs to date. The Cleveland Heart sickness dataset is utilized in this study to develop a Convolutional Neural Network (CNN) based model for the prediction of cardiovascular sickness. The dataset is split into two parts, with a ratio of 80:20, after undergoing essential preprocessing operations such as handling missing values, feature selection by correlation analysis, Min-Max normalization, and categorical encoding.6 Convolutional neural networks (CNNs) are designed to automatically extract hierarchical information for correct classification using their pooling, fully connected, and convolutional layers. With an AUC close to 1.0, a recall of 99.97%, a precision of 99.98%, and an F1-score of 99.96%, the CNN model demonstrates outstanding discriminative power in experimental assessments. Compared to more conventional models like Multi-Layer Perceptron, Logistic Regression, and Random Forest. Findings demonstrate CNN's dependability and robustness in identifying patterns of cardiac disease, lowering the rate of incorrect predictions, and bolstering clinical diagnosis; hence, CNN is a scalable option for practical healthcare applications.