Identification of diabetic patients through clinical and para-clinical features in Mexico: an approach using deep neural networks