Predicción de riesgo académico usando notas, asistencia a clases y clics en el LMS
Loading...
Publication date
Advisors
Department
Research group
Center
Abstract
Pocos estudios con Analítica del Aprendizaje han intentado predecir los resultados de
conjunto de un año académico. Esta investigación desarrolló un modelo predictivo del riesgo de
suspender el primer año en un grado de negocios (i.e., obtener menos créditos de los necesarios
para aprobar), utilizando Regresión Logística con datos de dos cohortes de estudiantes (n=1046).
El modelo utiliza la tasa de asistencia, calificaciones de evaluación continua de tres asignaturas y los
clics en la LMS, del primer semestre del año. Se probó con 74 estudiantes de una cohorte diferente,
obteniendo una precisión alta, pero una sensibilidad baja. Los resultados sugieren que la asistencia
a clases, las habilidades de comunicación y la competencia numérica son transversales al éxito
académico. El artículo revela una relación no lineal entre la actividad en la LMS y la media
académica y propone un método para tratarlo
few studies with Learning Analytics have attempted to predict the overall outcomes of an academic year. This research developed a predictive model of the risk of failing the first year in a business degree (i.e.; earning fewer credits than needed to pass), using Logistic Regression with data from two cohorts of students (n=1046). The model uses attendance rate, continuous assessment grades for three subjects, and clicks on the LMS, from the first semester of the year. It was tested with 74 students from a different cohort, obtaining high precision but low sensitivity. The results suggest that class attendance, communication skills, and numerical competence are crosscutting to academic success. The paper reveals a nonlinear relationship between LMS activity and academic mean, and proposes a method to address it
few studies with Learning Analytics have attempted to predict the overall outcomes of an academic year. This research developed a predictive model of the risk of failing the first year in a business degree (i.e.; earning fewer credits than needed to pass), using Logistic Regression with data from two cohorts of students (n=1046). The model uses attendance rate, continuous assessment grades for three subjects, and clicks on the LMS, from the first semester of the year. It was tested with 74 students from a different cohort, obtaining high precision but low sensitivity. The results suggest that class attendance, communication skills, and numerical competence are crosscutting to academic success. The paper reveals a nonlinear relationship between LMS activity and academic mean, and proposes a method to address it







