Machine Learning Approaches for Early Identifictaions of Heart Disease in Healthcare Applications

Abstract

Early detection and accurate diagnosis are crucial for improving patient outcomes and prescribing appropriate therapy for heart disease, which is still one of the leading causes of death globally. It can be extremely costly, time-consuming, and intrusive to use conventional diagnostic methods in settings with limited resources. Automated, data-driven prediction of cardiovascular problems has shown promising potential thanks to new ML and DL advances. This research presents a strong artificial neural network (ANN) model for the prediction of cardiovascular disease using a large-scale Kaggle dataset on the subject. The dataset includes more than 300,000 samples with 17 health-related variables. Complete data pre-processing, including missing value filling, duplicate removal, feature selection via the Chi-Square test, and min-max normalization, is required by the method. Two sets of data, one for training and one for testing, are used to teach an artificial neural network (ANN) to correctly identify cases of cardiac diseases. The experimental findings indicate that the proposed ANN is more effective, with an accuracy (ACC) of 98.75%, precision (PRE), recall (REC) and F1-score (F1) of 99, which is better than the baseline models including MLP, DNN, and Random Forest. Remarkably, the model is stable, can be generalized and is reliable, which is evidenced by ROC and loss curve analyses. The results point out the potential of the proposed method as a high-precision, cost-effective and clinically appropriate tool of early heart disease detection, which may help achieve timely interventions and better healthcare decisions.