GRID SEARCH BASED HYPERPARAMETER OPTIMIZATION FOR PREDICTING CORONARY ARTERY DISEASE USING MACHINE LEARNING TECHNIQUES

Abstract

Abstract— Heart disease is a major worldwide health concern that raises rates of morbidity and death. To successfully prevent and manage Coronary Artery Disease (CAD), early and precise prediction is essential. Still, it's a difficult undertaking to do. In order to forecast cardiac illness, this paper suggests a machine learning model that makes use of a number of preprocessing stages, hyperparameter optimization strategies as well. Making more accurate judgments for cardiac problems is facilitated by the prediction of this type of cardiac sickness. Hence, one practical method for predicting the illness as soon as feasible is to employ Grid Search Optimization (GSO) Machine Learning (ML) models. The GSO is used to find the optimal values of the hyper-parameters. The most advanced technique involves choosing the feature and adjusting the hyperparameter in tandem with the model search to reduce the false-negative rate. Feature selection for multivariate analysis based on statistical and correlation matrix analysis. Extensive comparative results are generated for the Grid Search models using retrieval, F1 score, and accuracy metrics. The performance results of the machine learning techniques were obtained. The Random Forest (RF) with tuned hyperparameters achieved the highest accuracy of 90.14% for predicting the disease. Z-Alizadeh sani dataset was used for this research. The dataset was pre-processed to making it suitable for a Machine Learning (ML) model. The Synthetic Minority Oversampling Technique (SMOTE) as well as SelectKBest techniques were used to improve a model's effectiveness and precision. SMOTE often results in a higher recall at the expense of decreased accuracy. The application of grid search to improve performance in training and testing was innovative in this study. The outcomes of the experiments were used to determine the optimal parameters for heart disease prediction. The methodology employed in this work greatly surpassed the findings of previous research that focused on the prediction of heart disease, according to comparative analyses that were also conducted.