Using Students' Programming Behavior to Predict Success in an Introductory Mathematics Course

Abstract

Computer science students starting their studies at our university often fail their first mandatory mathematics course, as they are not required to have a strong background in mathematics. Failing can also be partly explained by the need to adjust to a new environment and new working practices. Here, we are looking for indicators in students’ working practices that could be used to point out students that are at risk of failing some of their courses, and could benefit from an intervention. We present initial results on how freshman students’ programming behavior in an introductory programming course can be used to predict their success in a concurrently organized introductory mathematics course. A plugin in students’ programming environment gathers snapshots (time, code changes) from students actual programming process. Gathered snapshots are transformed to data items that contain features indicating e.g. deadlinedriven mentality or eagerness. Our results using Bayesian networks indicate that we can identify students with a high likelihood of failing their mathematics course already at a very early phase of their studies using only data that represents their programming behavior.