Stacked Ensemble Model with SMOTE for Heart Disease Prediction

This study develops a sophisticated method to predict heart disease using a Stacked Ensemble Model combined with the Synthetic Minority Over-sampling Technique (SMOTE) to address class imbalance in medical datasets. The methodology begins with data pre-processing mputing missing values, encoding categorical variables, and scaling data—followed by exploratory data analysis to understand variable distributions and correlations, and feature engineering to create new predictors. To counteract class imbalance in heart disease data, SMOTE is applied to generate synthetic instances for the minority class during training. The predictive model incorporates various machine learning classifiers (e.g., Logistic Regression, Decision Trees, SVM, XGBoost, etc.) as base models in a stacked ensemble, where a meta-model uses their predictions as inputs. This approach is evaluated through metrics like Accuracy, Precision, Recall, F1 Score, and AUC-ROC Curve, demonstrating improved accuracy, robustness, and potential for early diagnosis in heart disease management.

ABSTRACT: Heart Disease, Stacked Ensemble Model, SMOTE, Class Imbalance, Machine Learning, Predictive Modeling, Feature Engineering, AUC-ROC Curve.