Machine Learning Model to Classify Patients with Complicated and Uncomplicated Type 2 Diabetes Mellitus in the New Civil Hospital of Guadalajara “Juan I. Menchaca”

Diabetes Mellitus (DM) is a chronic disease worldwide. By 2030 are estimated to be 643 million people with DM, and by 2045 is projected to be 783 million, according to International Diabetes Federation. Machine Learning (ML) can be used as a smart preventive medicine tool for clinical records in hospitals, clinical laboratories, medical personnel, and patients. Implementing ML in current healthcare systems could translate into early diagnosis of DM. This work aimed to implement a classification algorithm for complicated Type 2 DM(T2DM), uncomplicated T2DM, and healthy Mexican participants. For this work, we enrolled 82 subjects from New Hospital Civil Juan I. Menchaca of Guadalajara, divided into 26 complicated T2DM, 26 uncomplicated T2DM, and 30 healthy subjects. ML algorithms used were decision tree (DT), Random Forest, AdaBoost, Bagging Classifier, and Support Vector Machine (SVM). The models use a dataset of 24 different clinical, biological, and molecular variables to discriminate between the 3 groups. The average accuracy was 78% from the C4.5 DT classifier, and we performed an AUC-ROC curve with value = 088. ML models can serve for early diagnosis of T2DM in healthcare systems, implementing this in preventive medicine clinics, developing an APP for smart mobile for personal care, and improving pharmaceutical approaches for treating T2DM.