Models to provide guidance in flipped classes using online activity

Abstract

The flipped classroom gives students the flexibility to organize their learning, while teachers can monitor their progress analyzing their online activity. In massive courses where there are a variety of activities, automated analysis techniques are required in order to process the large volume of information that is generated, to help teachers take timely and appropriate actions. In these scenarios, it is convenient to classify students into a small number of groups that can receive dedicated support. Using only online activity to group students has proven to be insufficient to characterize relevant groups, because of that this study proposes to understand differences in online activity using differences in course status and learning experience, using data from a programming course (n = 409). The model built shows that learning experience can be categorized in three groups, each with different academic performance and distinct online activity. The relationship between groups and online activity allowed us to build classifiers to detect students who are at risk of failing the course (AUC = 0.84) or need special support (AUC = 0.73), providing teachers with a useful mechanism for predicting and improving student outcomes.