Assessment of Machine Learning Classifiers for Heart Diseases Discovery

Heart disease (HD) is one of the utmost serious illnesses that afflict humanity. The ability to anticipate cardiac illness permits physicians to deliver better knowledgeable choices about their patient's wellbeing. Utilizing machine learning (ML) to minimize and realize the symptoms of cardiac illness is a worthwhile decision. Therefore, this study aims to analyze the effectiveness of some supervised ML procedures for detecting heart disease in respect to their accuracy, precision, f1-score, sensitivity, specificity, and false-positive rate (FPR). The outcomes, which were obtained using python programming language were compared. The data employed in this investigation came from an open database of the National Health Service (NHS) heart disease which originated in 2013. Through the machine learning (ML) technique, a dimensionality reduction technique and five classifiers were employed and a performance evaluation between the three classifiers- principal component analysis (PCA), decision tree (DT), random forest (RF), and support vector machine (SVM). The NHS database contains 299 observations. The system was evaluated using confusion matrix measures like accuracy, precision, f1-score, sensitivity (TPR), specificity, and FPR. It is concluded that ML techniques reinforce the true positive rate (TPR) of traditional regression approaches with a TPR of 98.71% and f-measure value of 68.12%. The true positives rate which is the same as the sensitivity was used to evaluate the accuracy of the classifiers and it was deduced that the PCA + DT outperformed that of the other two with a sensitivity of 98.71% and since the value is on the high side, this implies that the classifier will be able to accurately detect a patient with HD in his or her body.