Predictive model for Type 2 Diabetes Risk using Artificial Intelligence and multivariate data

Type 2 Diabetes Mellitus (T2DM) is a global public health concern due to its growing prevalence and silent progression. This study presents the development and validation of a predictive model based on artificial neural networks (ANN) using multivariate, non-invasive data. The model integrates three data sources: near-infrared spectroscopy (NIRS) for estimated glucose levels, plantar pressure sensors (FSR) for neuropathy indicators, and a digital lifestyle questionnaire (IMEVID). After preprocessing and training using a stratified dataset of adults without previous diagnosis, the ANN achieved 82 % accuracy, with high sensitivity and a Receiver Operating Characteristic Area Under the Curve (ROC-AUC) of 0.89. Compared to traditional methods like logistic regression, the proposed model demonstrated superior performance in classifying individuals at high or low risk. These results demonstrate the potential of non-invasive, AI-based screening tools to identify early-stage metabolic dysfunction, enabling timely preventive interventions. This approach may be especially useful in community-based or resource-limited settings to support public health decision-making.