Hybrid Classification Approach for Heart Disease using Few Shot Inspired Machine Learning Models

Accurate classification of heart disease is a challenging problem in healthcare, with significant implications for early diagnosis. This paper explores both conventional and innovative approaches for heart disease classification using a heart disease dataset. Baseline ML models, including Logistic Regression, Decision Tree, Random Forest, and SVM, are initially applied to establish a baseline performance benchmark. Building on these standard models, a few-shot-inspired prototype-based classifier is introduced, using class mean vectors to classify instances based on proximity to class prototypes, simulating a few-shot learning approach. Additionally, a KNN classifier is utilized to complement the prototype-based results by considering similarity-based classification. A comparative analysis demonstrated that the few-shot-inspired approach, in combination with baseline models, offers valuable insights into classification performance and robustness for healthcare datasets. The proposed method achieved an impressive accuracy of 98.5%. These findings underscore the potential of integrating prototype-based methods with traditional models to enhance classification accuracy in heart disease diagnosis.