Comparative Study of Decision Tree, AdaBoost, Random Forest, Naïve Bayes, KNN, and Perceptron for Heart Disease Prediction
Mohammad Maydanchi, Armin Ziaei, Mina Basiri, Alireza Norouzi Azad, Shaheen Pouya, Mehrbod Ziaei, Fatemeh Haji, Saman Sargolzaei
Abstract
Globally, cardiovascular diseases (CVD) are estimated to account for more than 32% of all deaths. Consequently, CVD has become a global health problem, and timely diagnosis is essential (WHO, 2021). Screening for risk factors accelerates the diagnosis and management of CVD, resulting in a more effective and rapid response, reducing the risk of death. This article compares six classification models, AdaBoost, Random Forest, Decision Tree, KNN, Naive Bayes, and Perceptron, to predict CVD symptoms. Based on CDC data collected from Kaggle, classification models were compared with the approach of examining effective factors to predict heart disease. Since the data set was imbalanced, the study performance was measured by AUC and F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> -score in Class 1, which is the critical class in this dataset. AdaBoost is found to have the highest AUC and F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> -score, respectively, of 0.828 and 0.37, while Decision Tree has the lowest AUC of 0.595 and F <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</inf> -score of 0.25.