Heart Disease Prediction & Disease Type Classification Using Machine Learningرسالة ماجستير

Abstract

Heart diseases are considered a significant cause of death globally; according to World Health Organization (WHO) statistics, heart diseases are the highest actual cause of death, with 16% ahead of all other causes. Therefore, accurate heart disease diagnosis is essential in determining the necessary treatment plan without negatively affecting the patient's health. There is no doubt about the physician's experience in diagnosing Heart disease. Still, sometimes there are mistakes in predicting and classifying Heart disease that negatively affects the patient’s health due to the difficulty distinguishing between these diseases through medical tests. Machine learning and deep learning techniques have proven successful in many areas, particularly medicine. This encourages utilizing Machine Learning capabilities to predict and classify heart diseases accurately. This study proposes an accurate and efficient machine learning system; the proposed method includes an initial stage to indicate the heart disease presence, followed by an intelligent classification to determine heart disease type, knowing that disease classification depends on the results in the initial stage. However, these predictions mainly depend on biomarkers results provided through laboratory tests. In principle, blood pressure, sugar level, cholesterol, and heartbeat average are the primary indicators to elaborate on heart health's general status. The basic principle of this paper is to identify and build an accurate system for heart disease prediction and classification; The used machine learning techniques will be subjected to the necessary examination to obtain relevant and appropriate results. These algorithms' methods include but are not limited to Logistic Regression (LR), Support Vector Machine (SVM) and K-Nearest Neighbors (KNN) network, and Deep Neural Network(DNN). Moreover, cross-validation will fundamentally outperform best practices to tune the hyper-parameter. Therefore, this study framework suggests a fair scientific comparison concerning the accuracy and execution time, in addition to the purpose of an exemplary model of heart disease prediction and classification trends. Efficiency and accuracy evidence declared an accuracy of 0.94 prediction model using the Gradient Boosting vi algorithm and an accuracy of 0.96 using the Decision Tree model for the classifications process. Machine learning models can help expand the benefits of medical informatics for future work by classifying diseases and predicting their progression. This will enable physicians to adjust treatments and take preventative measures earlier to reduce the risk of heart disease, maintain community health, and decrease mortality risk.