Current Approaches in Diabetes Mellitus Prediction: Applications of Machine Learning and Emerging Biomarkers

The purpose of the present chapter was to compare performance and accuracy of three different approaches to diabetes mellitus risk prediction. It was shown, in particular, that a multilayer perceptron, logistic regression, and random forest classifier can be successfully employed for prediction of the T2DM risk using the electronic medical record (EMR) system data with a relatively large number of individuals, albeit a limited dimension of predictor parameters space. Further improvement of the models should be addressed through the following avenues: application of machine and deep learning models for analysis of greater set of factors and derived parameters from the EMR system, including biometric, socioeconomic, geographical, ethnical, and genetic parameters.