Using bayesian networks for learning analytics in engineering education: A case study on computer science dropout at UCLM

Student dropout in Engineering Education is an important problem which has been studied from different perspectives andusing different techniques. This manuscript describes the methodology used to address this question in the context oflearning analytics, using Bayesian networks because they provide adequate methods for the representation, interpretationand contextualization of data. The proposed approach is illustrated through the case study of the abandonment ofComputer Science (CS) studies at the University of Castilla-La Mancha, which is close to 40%. To that end, severalBayesian networks were obtained from a database containing 363 records representing both academic and social data ofthe studentsenrolled in the CS degreeduring fourcourses.Then, theseprobabilistic modelswere interpretedand evaluated.The results obtained revealed that the great heterogeneity of the data studied did not allow to adjust the model accurately.However, the methodology described here can be taken as a reference for other works where a less heterogeneous databasecould be obtained, aimed at analysing student characteristics from a database.