Logistic Regression Versus Neural Networks: The Best Accuracy in Prediction of Diabetes Disease

Abstract

To derive actionable insights from vast amount of data in an intelligent fashion some techniques are used called machine learning techniques. These techniques support for predicting disease with correct case of training and testing. To classify the medical data logistic regression and artificial neural networks are the models to be selected. Today in world a major health problem is Diabetes Mellitus for which many classification algorithms have been applied for its diagnoses and treatment. To detect diabetes disease in early stage it needs greatest support of machine learning, since it cannot be cured and also brings great complication to our health system. In this paper, we establish a general framework for explaining the functioning of Artificial Neural Networks (ANNs) in binomial classification and implement and evaluate the variants of Back propagation algorithm (Standard Back Propagation, Resilient – Back propagation, Variable Learning Rate, Powell-Beale Conjugate Gradient, Levenberg Marquardt, Quasi-Newton Algorithm and Scaled Conjugate Gradient) using Pima Indians Diabetes Data set from UCI repository of machine learning databases. We also compare Artificial Neural Networks (ANNs) with one of the conventional techniques, namely logistic regression (LR) to predict diabetic disease decisions.

Keywords

Artificial Neural Networks (ANNs), logistic regression (LR), Quasi-Newton Algorithm, Resilient - Back propagation, Standard Back Propagation.

Full Text: