Steps towards a predictive semantic resource recommendation system for university students

Using information from the University of Los Andes, we created student performance prediction models for each compulsory course of three different programs: Industrial Engineering, Economy, and Systems and Computing Engineering. We created 3 feature sets using a wide range of academic metrics. Using 6 classification models we predicted whether a student would get a grade above the course average or below the course average, achieving a balanced accuracy of up to 77.8%. Using 7 regression algorithms we predicted a student's final grade (numerical value between 0 and 5) on a given course, achieving a minimum MAE of 0.234 and a R squared score of up to 46.4%. Results from the machine learning models indicated that the Random Forest algorithm was the best when predicting a course's grade using previous course grades. Using the Random Forest models, we analyzed the feature importance of certain courses when predicting other courses...

• Autores: Cueto Ramírez, Felipe
• Fecha de publicación: 2020
• Institución: Universidad de los Andes
• Repositorio: Séneca: repositorio Uniandes
• Idioma: eng
• OAI Identifier: oai:repositorio.uniandes.edu.co:1992/50979
• Acceso en línea: http://hdl.handle.net/1992/50979
• Palabra clave: Rendimiento académico; Predicciones educativas; Aprendizaje automático (Inteligencia artificial); Universidad de Los Andes (Colombia); Ingeniería
• Rights: openAccess
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/