Browsing by Subject "Machine Learning"

Now showing 1 - 6 of 6
  • Thumbnail Image
    Publication
    Open Access
    Determinantes del bajo rendimiento del alumnado español en comprensión lectora en PISA: un enfoque de machine learning
    (Universidad de Murcia. Servicio de Publicaciones, 2026) Arroyo Resino, Delia; Navarro Asencio, Enrique; Castro Morena, María; Constante-Amores, Alexander; Sin departamento asociado
    Low performance in reading comprehension is one of the major challenges in the Spanish educational system. Therefore, the objective of this study is to investigate the contextual deter-minants associated with this low reading performance. The sample consists of a total of 35,943 Spanish students and 1,089 educational centers that participated in the 2018 PISA assessment. The criterion variable is reading competence, which has been dichotomized (0 = medium and high performance, 1 = low performance). A total of 721 predictors were selected as independent variables. For data analysis, Random Forest machine learning algorithm was applied, and a multilevel binary logistic regression was conducted. The 30 most important variables related to students and school center explain 46% and 24% of the criterion variable, respectively. The final model (comprising both predictors) explains 47%. Among the main conclusions, the sig-nificance of educational process variables and non-cognitive and meta-cognitive constructs in low reading performance stands out. Therefore, the importance of addressing this educational phenomenon from a perspective less linked to socio-economic determinants and more focused on pedagogical aspects is emphasized.
  • Thumbnail Image
    Publication
    Open Access
    Discurso influenciado: aprendizaje automático y discurso de odio
    (Universidad de Murcia. Servicio de Publicaciones, 2023) Jaimes, Federico Javier
    Este trabajo tematiza la cuestión de los programas informáticos que discriminan, desde la filosofía del lenguaje. En esta disciplina, la bibliografía sobre discurso de odio ha centrado su análisis en los efectos que este produce en los grupos oprimidos. La idea central del presente trabajo será presentar una nueva noción, el discurso influenciado, que permita explicar lo que el grupo opresor es llevado a afirmar en base a la opresión sistemática. Así, el discurso influenciado permitirá tanto explicar la reproducción social de los discursos de odio como enmarcar teóricamente las afirmaciones discriminatorias realizadas por los programas informáticos previamente mencionados.
  • Thumbnail Image
    Publication
    Open Access
    ¿Es la inteligencia artificial doxástica un igual epistémico?
    (Universidad de Murcia. Servicio de Publicaciones, 2024) Murcia Carbonell, Alberto
    La inteligencia artificial doxástica (IAD) es un tipo de inteligencia artificial que reproduce actitudes doxásticas. Si la IAD cum-ple con las condiciones de paridad epistémica que se le exige a un humano, ¿podría ser también un igual epistémico? Dos iguales epistémicos sos-ienen propiedades cognitivas simétricas como la inteligencia, el razonamiento o la ausencia de sesgos. Para evaluar si alguien es un igual se ten-drán en cuenta estas condiciones: (1) la igualdad probatoria, (2) la igualdad cognitiva y (3) revelación completa. La IAD cumple tanto (1) como (2), pero es en (3) cuando se descubre que no puede ser un igual epistémico. Ésta responde con la opinión popular más aceptada estadísticamente, es incapaz de sostener y defender sus propias afirmaciones y éstas no son un genuino acto de habla. Es una máquina doxástica incapaz de señalar cuáles son las razones que guían su respuesta.
  • Thumbnail Image
    Publication
    Open Access
    LearningML: A Tool to Foster Computational Thinking Skills Through Practical Artificial Intelligence Projects
    (Universidad de Murcia, Servicio de Publicaciones, 2020) Rodríguez-García, Juan David; Moreno-León, Jesús; Román-González, Marcos; Robles, Gregorio
    The use of Artificial Intelligence (AI) offers new and thriving opportunities, but introduces also new risks and ethical issues that should be dealt with. We argue that the introduction of AI contents at schools through practical, hands-on, projects is the way to go to educate conscientious and critical citizens of the future, to awaken vocations among youth people, as well as to foster students’ computational thinking skills. However, most existing programming platforms for education lack some of the required educational features to develop a complete understanding of AI. In this paper we present LearningML, a new platform aimed at learning supervised Machine Learning (ML), one of the most successful AI techniques that is in the basis of almost every current AI application. This work describes the main functionalities of the tool and discusses some decisions taken during its design. For its conception, we have taken into account lessons learned from the research literature on introducing AI in school and from the analysis of other educational tools built with the aim to allow learners to use ML. We offer as well some promising results obtained after a preliminary testing pilot workshop. Finally, the next steps in the development of LearningML are presented, focused on the face and instructional validation of the tool.
  • Thumbnail Image
    Publication
    Restricted
    ModelSet : a dataset for machine learning in model-driven engineering
    (Springer, 2022) Hernández López, José Antonio; Cánovas Izquierdo, Javier Luis; Sánchez Cuadrado, Jesús; Informática y Sistemas
  • Thumbnail Image
    Publication
    Open Access
    Survival risk prediction in hematopoietic stem cell transplantation for multiple myeloma
    (De Gruyter, 2025-06-03) Belmonte, José María; Blanquer Blanquer, Miguel; Bernabé García, Gregorio; Jiménez Barrionuevo, Fernando; García Carrasco, José Manuel; Ingeniería y Tecnología de Computadores
    This paper investigates the application of Survival Analysis (SA) techniques to forecast outcomes after autologous Hematopoietic Stem Cell Transplantation (aHSCT) for Multiple Myeloma (MM). By leveraging six SA models, we examine their predictive capabilities, measured through the Concordance Index (C-index) metric. Beyond evaluating model performance, we analyze feature importance using permutation and SHAP methods, highlighting key clinical factors such as treatment history, disease stage, and prior disease progression or relapse as critical predictors of survival. The findings suggest that while all models performed well based on the C-index, a detailed examination revealed variations in how each model processed data. Specifically, the Coxnet and Random Survival Forest models exhibited a more thorough use of clinical variables, whereas the gradient boosting models appeared to rely on a narrower range of features, potentially limiting their ability to differentiate between patients with comparable profiles. Risk predictions categorized patients into low, moderate, and high-risk levels. For lower-risk patients, the procedure showed positive outcomes, while higher-risk individuals were predicted to have limited survival benefits, recommending alternative treatments. Lastly, we propose future research to expand these models into time-to-event estimations, offering additional support for decision-making by predicting patient life expectancy post-transplant, considering their pre-transplant clinical attributes.