Ensembling Predictions of Student Post-Test Scores for an Intelligent Tutoring System

Over the last few decades, there have been a rich variety of approaches towards modeling student knowledge and skill within interactive learning environments. There have recently been several empirical comparisons as to which types of student models are better at predicting future performance, both within and outside of the interactive learning environment. A recent paper (Baker et al., in press) considers whether ensembling can produce better prediction than individual models, when ensembling is performed at the level of predictions of performance within the tutor. However, better performance was not achieved for predicting the post-test. In this paper, we investigate ensembling at the post-test level, to see if this approach can produce better prediction of post-test scores within the context of a Cognitive Tutor for Genetics. We find no improvement for ensembling over the best individual models and we consider possible explanations for this finding, including the limited size of the data set.