Optimized Risk Assessment Model for Predicting Cardiac Disorders Using AI

Abstract

Cardiovascular disease (CVD) remains the leading cause of mortality worldwide, with the World Health Organization reporting over 19.1 million deaths attributed to CVD in 2022—accounting for approximately 33% of global fatalities. Early diagnosis and effective prediction play a crucial role in reducing such numbers, especially through the integration of artificial intelligence (AI) techniques. Electrocardiogram (ECG) data, widely used in clinical settings, offers valuable insight for CVD detection, yet the challenge lies in selecting the most relevant features from this complex data. Addressing this challenge, various feature selection (FS) techniques have been implemented to optimize classification accuracy. These techniques include Analysis of Variance (ANOVA) FS with Particle Swarm Optimization (PSO), Minimum Redundancy Maximum Relevance (MRMR) FS with PSO, Least Absolute Shrinkage and Selection Operator (LASSO) FS with PSO, Feature Correlation Based Technique (FCBT) FS with PSO, and ReliefF FS with PSO. These FS strategies were combined with machine learning algorithms to identify meaningful cardiac-related patterns and remove redundant or noisy features. The models were trained and validated using benchmark datasets such as the Heart Health Data Database (HHDD) and Behavioral Risk Factor Surveillance System (BRFSS), ensuring diverse data representation for improved generalizability. Among the classifiers evaluated, a Voting Classifier—an ensemble learning method that combines multiple base models—exhibited the highest predictive accuracy, reliability, and robustness across all feature selection methods and datasets. This approach leverages the strengths of individual classifiers to provide a more stable and accurate prediction of cardiac disorders. The outcomes clearly emphasize the effectiveness of combining optimized feature selection strategies with ensemble-based classification in enhancing AI-powered diagnosis. This fusion of statistical selection and advanced machine learning enables more precise and scalable solutions for cardiac risk prediction, contributing significantly to improved clinical decision support systems and better patient outcomes.

Ähnliche Arbeiten