Extracting Features to Classify Students Based on their Academic Performance

In today's educational climate, developing tools to support students and learning in a traditional or online context is a crucial responsibility. The first stages in employing machine learning techniques to enable such technology centered on forecasting a student's success in terms of marks earned. The disadvantage of these methods is that they are not as effective at predicting low-achieving students. The goal of our efforts is twofold. To begin, we investigate whether badly performing students may be more accurately predicted by recasting the task as a binary classification problem. Second, in order to learn more about the reasons that contribute to bad performance, we created a set of human-interpretable attributes that quantify these aspects. We conduct a study based on these characteristics to identify distinct student groups of interest while also determining their value.