| Abstrakt: |
Heart disease, a leading cause of mortality globally, is increasingly impacting populations worldwide. Effective prediction methods are essential to mitigate this growing health crisis. This study proposes a novel prediction framework, employing a Bipolar Sigmoid-Recurrent Neural Network (BS-RNN), to efficiently classify the heart disease database. Initially, patient health data and body function details are collected and balanced using Apache Kafka before being stored in a cloud database. This balanced data is then pre-processed and subjected to risk analysis. Subsequently, risk and non-risk factors are clustered using the Gibbs Entropy-K-Means Algorithm (GE-KMA), from which features are extracted. The correlation between the extracted features and those trained on the UCI database is assessed using a Pre-Policy Medical Check-Up (PPMC). Subsequently, the Karush-Kuhn-Tucker-Arithmetic Optimization Algorithm (KKT-AOA) is employed to select the optimal correlated features. These features are then input into the BS-RNN classifier for heart disease prediction. In addition to prediction, the framework measures disease severity based on the extracted features. The performance of the proposed model was found to surpass existing techniques, achieving an accuracy of 98.95%, an F-measure of 96.01%, and a specificity of 96.93%. The proposed clustering algorithm also demonstrated efficiency, forming clusters in 6208 ms with a superior prediction rate. These results underscore the potential of the proposed disease prediction framework to outperform existing methods in heart disease prediction. [ABSTRACT FROM AUTHOR] |
