Hybrid Cascade Forward Neural Network with Elman Neural Network for Disease Prediction

Abstract

In this paper, Hybrid Cascade Forward Neural Network with Elman Neural Network (HECFNN) is employed to classify six benchmark medical data sets, viz. Wisconsin Breast Cancer (WBC), Pima Indian Diabetes (PID), Liver Disorder Disease (LDD), Heart Disease (HD), Thyroid Disease (TD) and Cardiotocography (CTG). Three famous performance metrics in medical applications including accuracy, sensitivity and specificity are computed. The results of HECFNN are analyzed and compared with the well-known Elman Neural Network (ENN) and Cascade Forward Neural Network (CFNN). The experimental outcome shows that the HECFNN results outperform those of the CFNN and ENN. Performance outcome for WBC was 97.94% accuracy (ACC), 98.88% specificity (SPE) and 98.84% sensitivity (SEN), while results of PID were 85.10%, 75.61% and 88.39% for ACC, SPE and SEN, respectively, LDD results were ACC 93.80%, SPE 90.09% and SEN 87.50%, HD prediction for ACC, SPE and SEN were 94.01%, 96.71% and 90.40%, respectively, TD obtained 96.10% for ACC, 96.45% for SPE and 96.66% for SEN and CTG achieved 100.00% for SEN, 99.25% for ACC and 98.00% for SPE. In addition, the obtained accuracy of the HECFNN for every benchmark is also compared with those different methods published in the literature review. The results demonstrate that HECFNN produce higher accuracy compared with other well-known methods. In general, the HECFNN experimental results positively demonstrate that the HECFNN is effective and useful in undertaking medical data classification tasks.