Polynomial FLANN Classifier for Fetal Cardiotocography Monitoring

Abstract

An efficient adaptive classifier for fetal electronic monitoring based on a modified structure of neural networks is presented. It employs polynomial series as a functional expansion. Training of the Polynomial Neural Network (PNN) classifier is performed using a NewtonLeast Mean Square (NLMS) adaptive algorithm, which requires few iterations and epochs. The convergence is achieved using the PNN classifier in a very short training time. The performance of the proposed classifier has shown a very high overall classification accuracy of 99.74% in comparison with those of the other excising machine learning classifiers. A performance comparison between the proposed PNN classifier and other Functional Link Artificial Neural Network (FLANN) classifiers such as Legendre Neural Network (LNN) and Volterra Neural Network (VNN) based classifiers in electronic fetal monitoring is provided. The simulation results reveal that the PNN classifier outperforms both the LNN and VNN classifiers in terms of mean square error, overall classification accuracy, computational time and computational complexity.