A Gaussian Radial Basis Function for Classification of Diabetic Patients: A Case of FMC, Yola

Abstract

Diabetes is a chronic illness that has become a worldwide epidemic. Modern clinical practice shifting towards a patient-centered approach, using computer-based knowledge in finding and understanding health and illness concepts is critical. Thus, several artificial intelligence techniques such as machine learning algorithm, regression tree, and logistic regression are used to build a prediction or classification model to improve effective understanding of this illness. Therefore, this study utilizes the Gaussian Radial Basis Function (RBF) to classify diabetic patients using data obtained from the Federal Medical Center, Yola, Adamawa State, Nigeria. The data consists of 533 patients and their demographic variables and laboratory variables, which include age, mass of a patient, glucose level, pressure, insulin, and class variable. The SPSS version 21 for windows 10.1 was used to analyse the data. The network has 10 units hidden layer with softmax activation function and the output layer has 2 units with identity activation function. Also, the sum of squares is used as the error function. The analysis shows that the suggested Radial Basis Function (RBF) techniques generally achieve 94.3% correct classifications with 92.2 % correct classification for diabetes patients and 96.6% correct classifications for non-diabetes patients. Remarkably, the results indicate the effectiveness of applying RBF as significant method to predict and classify diabetic and non-diabetic patients to help in understanding variables that may influence diabetic treatment