Medical Prediagnosis of Type 2 Diabetes Mellitus using Machine Learning

Abstract

The article presents the development of a system to pre-diagnose type 2 diabetes mellitus. The pre-diagnosis process consists of three measurement stages; anamnesis, physical examination and a complementary examination. As a result, the probability of suffering from type 2 diabetes mellitus is disclosed through neural networks that use as inputs: age, gender, height, weight, abdominal circumference, family history related to diabetes, pathology and state of pregnancy if applicable. Perceptrons were used to classify the patterns and the results were validated with Clarke's consensus error analysis grid, making it possible to obtain a non-invasive pre-diagnosis system with a probability of success of no less than 90%