Öztürk, AlisanBonfert-Taylor, PetraFügenschuh, ArminKrömker, DetlefSchroeder, Ulrik2019-03-282019-03-282018978-3-88579-678-7https://dl.gi.de/handle/20.500.12116/21049Programming classes are difficult by nature and educators are eager to find ways to deal with high dropout rates. Today’s technologies allow us to capture programming-related student data, which can be used to identify students in need of assistance and in getting insights in student learning. In order to assist novice programming students in learning how to program, we developed a web-based programming environment, which is used by students throughout the whole course. While it also provides students with enhanced error messages, all data of students’ interactions are captured. Through this data, we identified two metrics, related to small programming assignments, which highly correlate with student performance. These metrics along other features further enabled us to implement machine learning algorithms that could accurately predict dropout-prone students, early on in the course. Overall, methods of educational data mining can be utilized to assist both, students and educators in introductory programming courses.enE-LearningMachine LearningIntroductory ProgrammingPredictionEnhanced Error MessagesText/Conference Paper1617-5468