Abstract
The precise diagnosis of heart disease represents a significant obstacle within the medical field, demanding the implementation of advanced diagnostic instruments and methodologies. This article conducts an extensive examination of the efficacy of different machine learning (ML) and deep learning (DL) models in forecasting heart disease using tabular dataset, with a particular focus on a binary classification task. An extensive array of preprocessing techniques is thoroughly examined in order to optimize the predictive models' quality and performance. Our study employs a wide range of ML algorithms, such as Logistic Regression (LR), Naive Bayes (NB), Support Vector Machine (SVM), Decision Tree (DT), Random Forest (RF), K-Nearest Neibors (KNN), AdaBoost (AB), Gradient Boosting Machine (GBM), Light Gradient Boosting Machine (LGBM), CatBoost (CB), Linear Discriminant Analysis (LDA), and Artificial Neural Network (ANN) to assess the predictive performance of these algorithms in the context of heart disease detection. By subjecting the ML models to exhaustive experimentation, this study evaluates the effects of different feature scaling, namely standardization, minmax scaling, and normalization technique on their performance. The assessment takes into account various parameters including accuracy (Acc), precision (Pre), recall (Rec), F1 score (F1), Area Under Curve (AUC), Cohen's Kappa (CK)and Logloss. The results of this research not only illuminate the optimal scaling methods and ML models for forecasting heart disease, but also offer valuable perspectives on the pragmatic ramifications of implementing these models within a healthcare environment. The research endeavors to make a scholarly contribution to the field of cardiology by utilizing predictive analytics to pave the way for improved early detection and diagnosis of heart disease. This is critical information for coordinating treatment and ensuring opportune intervention.