Investigating performance in a blended SPOC

摘要

In this paper, we describe how to investigate performance in a blended SPOC (small private online course). For the quantitative research, we build predictive models of students' performance each week in a SPOC. We document a performance prediction methodology from raw logging data extracted from Open edX platform to model analysis. These logging data were collected from Computer Structure Lab Course offering at Beihang University. We show how to extract meaningful information from the learning related educational data we gathered. 28 predictive features extracted for 377 students, and our model achieved an AUC (area under curve) in the range of 0.62-0.83 when predicting one week in advance. An early warning system is established to identify at-risk students in the SPOC, especially for the blended lab course. Furthermore, we could use the most important features to form the assessment for each student during the semester.