Predicting Students Progression Using Existing University Datasets: A Random Forest Application

This paper proposes the use of data available at Manchester Metropolitan University to assess the variables that can best predict student progression. We combine Virtual Learning Environment and MIS student records data sets and apply the Random Forest (RF) algorithm to ascertain which variables can best predict students’ progression (students satisfactorily completing one year and passing to the next or graduating). RF was deemed useful in this case because of the large amount of data available for analysis. The paper reports on the initial findings for data available in the period 2007-08. Results seem to indicate that variables such as students’ time of day usage, the last time students access the VLE and the number of document hits by staff, are the best predictors of student progression. The paper contributes to VLE evaluation and highlights the usefulness of a technique initially developed in the field of biology in an educational environment.