Abstract
Educational games have become hugely popular, and educational data mining has been used to predict student performance in the context of these games. However, models built on student behavior in educational games rarely differentiate between the types of problem solving that students employ and fail to address how efficacious student problem solutions are in game environments. Furthermore, few papers assess how the features selected for classification models inform an understanding of how student behaviors predict student performance. In this paper, we discuss the creation and consideration of two models that predict if a student will develop an elegant problem solution (the Gold model), or a non-optimal but workable solution (the Silver model), in the context of an educational game. A pre-determined set of features were systematically tested and fit into one or both of these models. The two models were then examined to understand how the selected features elucidate our understanding of student problem solving at varying levels of sophistication. Results suggest that while gaming the system and lack of persistence indicate non-optimal completion of a problem, gaining experience with a problem predicts more elegant problem solving. Results also suggest that general student behaviors are better predictors of student performance than level-specific behaviors.
Original language | English (US) |
---|---|
Pages | 448-453 |
Number of pages | 6 |
State | Published - 2016 |
Event | 9th International Conference on Educational Data Mining, EDM 2016 - Raleigh, United States Duration: Jun 29 2016 → Jul 2 2016 |
Conference
Conference | 9th International Conference on Educational Data Mining, EDM 2016 |
---|---|
Country/Territory | United States |
City | Raleigh |
Period | 6/29/16 → 7/2/16 |
Keywords
- Classifiers
- Educational games
- Problem solving
ASJC Scopus subject areas
- Computer Science Applications
- Information Systems