Using learning analytics to support students' engineering design: the angle of prediction

Engineering design plays an important role in education. However, due to its open nature and complexity, providing timely support to students has been challenging using the traditional assessment methods. This study takes an initial step to employ learning analytics to build performance prediction models to help struggling students. It allows instructors to offer in-time intervention and support for these at-risk students. Specifically, we develop a task model to characterize the engineering design process so that the data features can be associated with the abstract engineering design phases. A two-stage feature selection method is proposed to address the data sparsity and high dimensionality problems. Then, instead of relying on probability-based algorithms such as Bayesian Networks to represent the task model, this study used the Radial Basis Function based Support Vector Machines for prediction to identify the struggling students. Next, we employ an extra-tree classification method to rank the importance of these features. Teachers can integrate the feature importance ranking with the abstract task model to diagnose students' problems for scaffolding design. The results show that the proposed approach can outperform the baseline models as well as providing actionable insights for teachers to provide personalized and timely feedback to students. Implications of this study for research and practice are then discussed.