A Comparative Analysis of Machine Learning-Based Prediction for Heart Disease Detection

Heart failureHeart failure is a complex clinical condition that affects a large proportion of the global population. Heart diseaseHeart disease remains a major health concern. Therefore, early and accurate detection is crucial for successful treatment and timely recovery. This chapter explores machine-learning techniques and clinical assessments to evaluate their performance in detecting heart-related disordersDetecting heart-related disorders. This chapter also investigates the limitations and challenges associated with different detection approaches. Five state-of-the-art machine learning models such as Logistic Regression (LRLogistic Regression (LR)), Random Forest (RFRandom Forest (RF)), Naive Bayes (NBNaive Bayes (NB)), Decision Tree (DTDecision Tree (DT)), Support Vector Machine (SVMSupport Vector Machine (SVM)), and three hybrid model hybrid models (LR, RF), Hybrid model (DT, NB) and Hybrid Model (KNN, RF) explained in this chapter. The dataset used in this chapter was taken from the University of California Irvine (UCI) and consisted of 14 attributes, including Blood Pressure (BPBlood Pressure (BP)), cholesterol (Chol), electrocardiogram (ECGElectrocardiogram (ECG)), and a total number of 1026 data records. The chapter shows that the machine learning model LR achieves accuracy, precision, recall, and F1-score of 85%, 87%, 85%, and 86%, respectively; SVM achieves 85%, 87%, 85%, 86% respectively, and DT achieves 77%, 83%, 73%, and 78%, respectively; NB achieves 86%, 90%, 85%, and 87%, respectively, and RF achieves 86%, 88%, 88%, and 88%, respectively. Hybrid model K-Nearest Neighbor (KNNK-Nearest Neighbor (KNN)) with Random Forest achieve accuracy, precision, recall, F1-score 0.9272, 0.9151, 0.9417, 0.9282, Linear Regression with Random Forest 0.92718, 0.9583, 0.8932, 0.9246 and Decision Tree with Naïve Bayes 0.9126, 0.9293, 0.8932, 0.9109 respectively. The results of this study show that the hybrid models have superior results compared with the other models.