The number of heart disease patients has significantly increased in recent years, and heart illness is linked to a high death rate. Furthermore, as technology advanced, several sophisticated devices were created to assist patients in measuring their health at home and estimating their risk of developing heart disease. Using six machine learning models, the study seeks to determine how accurate self-measured physical health indicators are at predicting heart disease when compared to all indicators assessed by medical professionals. Logistics Regression, K Nearest Neighbors, Support Vector Model, Decision Tree, Random Forest, and Gradient Boosting Classifier were among the six models employed to forecast heart disease. Twelve different test findings and 1189 patients' heart disease risks are included in the database utilized for the study. While the metrics contains six outcomes that could be tested, the all metrics contained all twelve test results. The accuracy score and false negative rate were calculated for each of the five models that were built for the all metrics.The findings demonstrated that in all five models, all metrics had greater accuracy scores than existing metrics. For five machine learning models, all metrics had false negative rates that were either lower or equivalent to that of existing metrics. The results showed that all physical indicators were more accurate in predicting patients' risk of heart disease than metrics measured physical health indicators. Therefore, all physical health indicators are preferable for assessing the risk for cardiac illnesses in the absence of future development of indicators.