Cardiovascular disease (CVD), a common deadly disease, is primarily attributed to unhealthy human habits, making it a major global cause of death and a significant public health concern. Timely diagnosis of CVD is crucial, as untreated CVD can result in fatalities. The presence of an accurate decision support system becomes of utmost importance in identifying heart disease at an early stage, particularly in developing nations, specifically in distant and rural areas where the availability of heart specialists is limited. Machine learning (ML) algorithms offer a cost-effective and efficient way to identify heart disease promptly. This study aimed to determine ML classifiers that exhibit the highest accuracy when used for diagnosing CVD. Several machine-learning algorithms, including SVM (Support Vector Machine), DT (Decision Tree), ANN (Artificial Neural Network), LR (Logistic Regression), KNN (K-Nearest Neighbor), Bagging DT, Bagging SVM, Bagging KNN, Bagging Gaussian Process, and AutoML, were applied and compared based on some performance measures like accuracy, precision, recall and F1 score in heart disease prediction. Heart Disease data from UCI Machine Learning Repository was used in this study. Based on the findings, the results indicated that AutoML emerged as the best classifier, exhibiting exceptional performance with perfect accuracy, precision, recall, and Fl score, all achieving a value of 100%. In addition, a risk factor analysis was carried out using Bayesian Belief Network (BBN) to assess the risk of heart disease among individuals already diagnosed with heart disease that is predicted by AutoML. The risk factor analysis showed that women within the age range of 28 to 42 and 73 to 87, with cholesterol levels ranging from 86 to 245 and resting blood pressure between 91 and 174, possess the highest probability of experiencing heart disease within the given dataset.