- Version
- Download 1
- File Size 566.29 KB
- File Count 1
- Create Date November 27, 2025
- Last Updated November 27, 2025
ENHANCING DIAGNOSTIC ACCURACY: ANFIS-PSO WITH EXPANDED RULE BASE FOR HEART DISEASE PREDICTION
ABSTRACT
This study introduces a new method for diagnosing heart disease using adaptive neural-flow identification (ANFIS) technology, optimised by particle swarm optimization (PSO). The heart disease, the world's leading killer, requires precise diagnosis. Traditional ANFIS models often suffer from limited rule bases and insufficient initialization. Here we propose an ANFIS-PSO with a 140 rule base, which exceeds the 87 rule base of the reference ANFIS-GA "Jindong Model" which avoids the early convergence of GA and increases the effectiveness of the training. Both models were tested in the Cleveland Heart Disease and Heart Failure Prediction Datasets for generalizability. The metrics included accuracy, sensitivity, specificity, plausibility, F1 score, and volatility. For Cleveland, the accuracy was similar, but the specificity (100 percent vs. 90.32), accuracy (100 percent vs. 96), F1 score (87 percent vs. 83) and MCC score (81 percent vs. 74 percent) were superior, and were confirmed by robust generalizability in the heart failure study. Despite the slight accuracy advantage, the reduction of false positives is crucial for medical applications. Future research will explore further algorithms, data sets and statistical validation.
Keywords: Heart disease diagnosis, ANFIS, PSO, Genetic