Comparative Study of Heart Disease Prediction Algorithm

The prevalence of cardiovascular disease has been steadily rising, driven by factors such as sedentary lifestyles, unhealthy dietary habits, tobacco use, obesity, and an aging population. According to the According to the World Health Organization (WHO) and the (GBD) study, cardiovascular disease is the main cause of death. In addition to comparing the performance of the five machine learning algorithms and two different datasets, this study also aims to investigate the impact of different algorithmic approaches on heart disease prediction. Logistic regression is a widely-used algorithm for binary classification, utilizing a linear function to model the relationship between input features and the target variable. K-nearest neighbors (KNN) algorithm relies on similarity measures to classify instances based on their proximity to known instances. Random forest, a powerful ensemble method, combines multiple decision trees to improve prediction accuracy. Decision tree algorithms employ a hierarchical structure of nodes and branches to make sequential decisions. Support vector machines (SVM) aim to find an optimal hyperplane that separates different classes in the feature space To conduct the evaluation, a dataset from the UCI Machine Learning Repository containing relevant features and corresponding heart disease labels was utilized. Prior to training and testing the algorithms, the dataset underwent exploratory data analysis to understand its characteristics. Preprocessing steps were employed to handle missing values and scale the features appropriately. To ensure unbiased evaluation, the collection of data is subsequently split between train and testing collections with a 70 to 30 ratio. Several factors, such as the F1 score, an area underneath a receiver operating characteristic curve (AUC-ROC), recall, accuracy, and exactness were used to assess the algorithms' performance. Depending on the experimental findings, the random forest method outperformed logistic regression, KNN, decision trees, and SVM in terms of all assessment measures. Its greater F1 score, AUC, recall, accuracy, and precision all point to its improved capacity to distinguish between cases and non-cases of heart disease.