Performance comparison of featured neural network with gradient descent and levenberg-marquart algorithm trained neural networks for prediction of blood glucose values with continuous glucose monitoring sensor data

Abstract

Continuous Glucose Monitoring Systems are used to track the time course of blood glucose for Diabetes people. Prediction of Hypo/Hyper glycemic occurrences are the main task in the management of Diabetes. Our work involved the development of a feed forward back propagation neural network that is trained with the features of incoming continuous glucose monitoring sensor data and prediction of future blood glucose values with the special activation functions. This paper had presented the comparison of Featured neural network with that of Gradient Descent and Levenberg Marquardt back propagation algorithms.