Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/152178
Title: Aplicación de Técnicas de Explicabilidad (XAI) y Cuantificación de Incertidumbre (UQ) en la Predicción del Síndrome Metabólico mediante Aprendizaje Automático
Author: Belmonte Marín, Xema
Director: Moreno de Castro, María
Tutor: Sánchez Castaño, Friman
Abstract: This final degree project focuses on the application of supervised machine learning techniques, employing explainability and uncertainty quantification methods, specifically for binary classification to predict metabolic syndrome, a medical condition that increases the risk of cardiovascular disease and diabetes. The work begins with an exploratory data analysis of a dataset composed of demographic and clinical measurements from over two thousand individuals labeled with either the absence or diagnosis of metabolic syndrome. Various models, such as random forest and gradient boosting machines, are trained and evaluated, optimizing hyperparameters and decision thresholds. The selected model is a random forest, chosen not only for its good performance but also for its ability to minimize false negatives, which is crucial in clinical diagnosis. Explainability techniques (XAI), such as SHAP, LIME, and counterfactual explanations, are applied to interpret predictions and explain the model's decisions, identifying the most relevant features for diagnosing metabolic syndrome. Furthermore, uncertainty quantification is addressed using Conformal Prediction, which provides prediction intervals with coverage guarantees and calibrated probabilities. These approaches ensured transparency and reliability, essential in a medical environment regulated by frameworks such as the European Union AI Act.
Keywords: supervised machine learning
explainable artificial intelligence
uncertainty quantification
Document type: info:eu-repo/semantics/bachelorThesis
Issue Date: Jan-2025
Publication license: http://creativecommons.org/licenses/by-nc-nd/3.0/es/  
Appears in Collections:Bachelor thesis, research projects, etc.

Files in This Item:
File Description SizeFormat 
jbelmontemariTFG0125.pdfMemoria7,09 MBAdobe PDFThumbnail
View/Open

jbelmontemariTFG0125presentacion.mp4

155,16 MBMP4View/Open
Share:
Export:
View statistics

Items in repository are protected by copyright, with all rights reserved, unless otherwise indicated.