Diagnosis of Cardiac Disease Utilizing Machine Learning Techniques and Dense Neural Networks

In numerous different polls, heart illnesses constantly rank among the leading causes of death. The reason behind is the complexity of these diseases and the high prevalence of incorrect diagnoses. This is a huge obstacle for those who work in the medical field. Since machine learning (ML) has been shown to be incredibly excellent at estimate and decision-making, it is necessary to develop a system that can identify cardiac disease. A heart disease diagnosed in its initial phases not only makes it possible for patients to avoid having one, but it also makes it possible for medical professionals to gain knowledge about the primary risk factors for heart attacks and take precautions against them before they happen to a patient. In this study, we experimented on three different datasets with different ML algorithms to diagnose heart disease. These datasets have various medical and customary information about the patients, which are necessary elements to determine whether a patient has heart disease or not. Hybrid model proposed by combining custom convolution neural network and extreme gradient boosting shows better accuracy in finding heart diseases among all standard machine-learning methods. We also proposed a custom sequential model formed with seven dense layers to diagnose a patient with cardiac disease. This proposed model performed well, with an accuracy of 92.3% when applied to the modified Cleveland dataset.