Abstract
In this paper, we provide evidence that analysis of student interaction with web-based pedagogical tools in a large partly-flipped CS1 class can be used for early detection of at-risk students. We used student interaction data to estimate five learning style features: engagement, learning speed, confidence, drive, and persistence. We found a positive correlation between final course grades and two of the features of student learning style: engagement and learning speed. Drive and persistence were not good predictors of student success but did provide lower bounds on overall course performance. Confidence was a poor predictor of overall course performance. This kind of analysis provided a more timely and nuanced view of student learning styles than can be obtained from traditional summative assessments and this data could allow for early detection of at-risk students in an introductory programming course.