LogNNet Neural Network Application for Diabetes Mellitus Diagnosis

Abstract

The paper presents a LogNNet neural network algorithm for diabetes mellitus diagnosing based on a public dataset. The study used 100 thousand records of patient conditions. Model quality was evaluated using the Matthews Correlation Coefficient metric (MCC). The LogNNet neural network model showed high accuracy (MCC=0.733) in diabetes mellitus recognition. A highly positive relationship between HbA1c level and glucose level in the disease diagnosing was found using the LogNNet model. It has been observed that evaluating these variables together is much more effective than their individual effects in diagnosing the disease.