Hybrid framework with machine learning and fuzzy approaches for medical decision making

Abstract

The advancements in the field of artificial intelligence and medicine have paved the way for the use of computers in diagnosis. It has also become important to reduce the mortality rate by early detection of diseases. Developing a classifier model using artificial intelligent techniques is one of the major research areas of knowledge mining from clinical datasets. In this research work, a hybrid classifier for medical decision making using artificial intelligent techniques from clinical datasets has been developed. The proposed classifier has been evaluated using five clinical datasets obtained from the University of California at Irvine (UCI) machine learning repository. There are four contributions in this research work. The first contribution evaluates twelve different BPNN training algorithms using varying network parameters. The second and third contributions have developed Artificial Neural Network (ANN) classifier for disease diagnosis using clinical datasets. In the fourth contribution a Clinical Decision Support System (CDSS) has been proposed to diagnose Gestational Diabetes Mellitus (GDM). In the first contribution, an analytic study has been performed on BPNN classifier using twelve different back-propagation algorithms with varying network parameters. The network parameters used for BPNN training are selection of initial weights and biases, number of hidden layers, number of neurons per hidden layer, activation function, learning rate and momentum term. The network parameter values have been evaluated using twelve BP algorithms namely, Gradient Descent BP, Gradient Descent with Momentum newlineBP, Gradient Descent with Adaptive Learning Rate BP. newline newline