Enhanced Heart Disease Prediction Through Meta-Features and Optimized Diagnostic Techniques

Abstract

One of the main causes of death worldwide is still heart disease, for which prompt and precise diagnostic methods are essential for efficient treatment and intervention. In this work, we present an integrated method for heart disease prediction that makes use of meta-features and best diagnostic methods. Through the combination of the advantages of advanced diagnostic techniques and meta-learning, our approach seeks to improve prediction accuracy. We base our method on the use of meta-features, which are higher-order statistical descriptors that are taken from the dataset. Through the capture of subtle relationships and patterns that conventional features might miss, these meta-features provide a comprehensive picture of the features of the dataset. Our goal in including meta-features into our prediction models is to improve the diagnostic framework's generalization and discriminatory power. Moreover, our method includes optimum diagnostic methods designed to fit the particular features of datasets on heart disease. Class imbalance is addressed and minority class representation is improved by using the adaptive synthetic sampling method ADASYN. We also substitute robust classification with support vector machines (SVM), ensemble learning with random forest, and potent meta-learner XGBoost, all of which are tuned to maximize predictive performance for traditional classifiers. We use the Davide Chicco and Giuseppe Jurman dataset, a commonly used benchmark dataset in heart disease research, for experiential analyses to assess the effectiveness of our integrated approach. By means of thorough testing in various case scenarios with different data split ratios, we evaluate the accuracy, precision, recall, and F1-score of our method. We show that the integrated strategy that we have suggested works well for heart disease prediction. Our models continuously perform better than baseline approaches in a variety of case scenarios, demonstrating the promise of meta-feature integration and optimized diagnostic approaches in enhancing robustness and predictive accuracy.  The work presented highlights the need of combining meta-features with improved diagnostic methods in heart disease prediction and provides a viable way to progress the state-of-the-art in cardiovascular health management and diagnosis.