Publication

Abstract : Cardiovascular heart disease stands as a significant world health concern, holding the ominous title of the foremost significance of mortality and disability on a global scale. Information was obtained from the World Health Organization; cardiovascular heart disease claims a staggering nearly 18 million lives annually, representing thirteen percent of all world fatalities. Cardiovascular heart disease emerges as a prominent contributor to death across the globe. Machine learning and deep learning algorithms have demonstrated remarkable potential in forecasting the likelihood of heart diseases. Meta-learning, called learning-to-learn, is a variant of machine learning that empowers systems to refine their learning process and will enhance the classification accuracy if the dataset contains minimum size. It encompasses a suite of techniques facilitating iterative improvement in learning. In the present study, we endeavor to craft a meta-learning-based classification model tailored for the discernment of heart diseases, predicated upon a meticulously curated dataset comprising seventy-six distinct features. The primary focus resides in predicting the presence of heart disease as the principal attribute. Subsequently, the outcomes derived from our proposed methodology will be juxtaposed with conventional and state-of-the-art contemporary techniques. Moreover, we shall undertake a comparative analysis between the results obtained with and without applying the synthetic minority oversampling technique (SMOTE), a strategic approach to rectify imbalances within target classes. Finally, the proposed meta-learning approach outstrips conventional and other state-of-the-art models, furnishing a more precise prognosis of heart disease risk. The proposed method findings suggest the viability of utilizing the meta-learning approach to enhance predictive accuracy. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2026.