Prototipo para la predicción del riesgo de enfermedades cardiometabólicas basado en técnicas de IA

Abstract

Abstract: Cardiometabolic diseases, especially type 2 diabetes mellitus, represent one of the main challenges for global public health due to their high prevalence, silent progression, and association with sedentary lifestyles and inadequate dietary habits. In this context, artificial intelligence applied to healthcare emerges as an innovative alternative to improve early detection, monitoring, and clinical decision-making through the efficient analysis of large volumes of data. The objective of this degree project is the development of a predictive prototype based on machine learning techniques to estimate the risk of cardiometabolic diseases, with an emphasis on diabetes, using clinical and personal data. The methodological process included data preprocessing, identification of relevant variables, and elimination of redundant information. Subsequently, four different machine learning models were trained and evaluated Decision Tree, LightGBM, Gradient Boosting, and ExtraGB in order to compare their performance. Gradient Boosting was selected as the model to be used in the project due to its superior performance metrics, which include: ROC AUC = 0.96, Accuracy = 0.91, Precision (+) = 0.89, Precision (−) = 0.92, Recall (+) = 0.86, Recall (−) = 0.94, F1-Score (+) = 0.87, and F1-Score (−) = 0.93. The evaluation using classification metrics demonstrated that artificial intelligence is an effective tool for the early diagnosis and prevention of diabetes.