Using data analytics to examine undergraduate computer science student course performance

Four studies were conducted using different data analytic techniques to give insight into how undergraduate students perform in their required courses in the computer science sequence at the University of Illinois at Urbana-Champaign. Undergraduate student data from students who majored in computer science were collected for the length of time they remained in the computer science department. Principal Components Analysis was used to determine the general patterns behind student grades as they progressed in the major, looking at whether student grades were an indication of what kind of students they were, or how much prior experience they had, or the courses that they took when they reached Illinois. Decision trees were used to determine the factors that can be used to make predictions about whether students would fail a difficult upper level course, so that vulnerable students might be detected ahead of time to be provided with extra help and resources. Current students already in the computer science program were examined to determine if there should be a threshold technical GPA for being allowed to take higher level computer science courses, by looking at the factors behind a student repeating upper level computer science courses. Finally, Support Vector Machine was used to examine and critique the current rules for letting undergraduates transfer into the computer science major, to determine whether the current standards make sense with the pattern of student performance in previous courses and whether those standards’ threshold are accurate enough for predicting their success should they transfer