Early Heart Disease Predictıon Using LV-PSO And Fuzzy Inference Xception Convolution Neural Network On Phonocardiogram Signals

Prabha Devi D^*Palanisamy C

• Bannari Amman Institute of Technology Department of Computer Science and Engineering, Sathyamangalam, India

Heart disease is a significant cause of death worldwide, and early detection and classification are essential for effective treatment. Phonocardiogram (PCG) signals have shown promising results in diagnosing heart conditions, but accurately classifying these signals remains a challenge due to feature dimensions. The accuracy, recall, coefficient values, and f1score precision are all misclassified when the actual margin increases in data.To resolve this, we proposed a Linear Vectored -Particle Swarm Optimization (LV-PSO) based on Fuzzy Inference Xception Convolution Neural Network for early heart risk prediction. The Sound waves are converted into PCG signals to identify feature variations like Delta, Theta, diastolic, and systolic differences in dataset. To identify the scalar difference and disease behavioural impact, Support Scalar Cardiac Impact Rate (S2CIR) is adopted. Non-linear scaling values are processed with LV-PSO to reduce feature dimensionality. The selected features are trained with fuzzy inference Xception Convolution Neural Network (XCNN) to categorize disease type by class and tested on different test dataset. The results prove the proposed system achieves high accuracy in predicting heart disease, with precision rate of 95.6%, 93.1% recall rate and 95.8% prediction accuracy, outperforming in multivariate class based on the disease type definition compared to other system.