Predicting Student Failure in an Introductory Programming Course with Multiple Back-Propagation

Abstract

One of the most challenging tasks in computer science and similar courses consists of both teaching and learning computer programming. Usually this requires a great deal of work, dedication, and motivation from both teachers and students. Accordingly, ever since the first programming languages emerged, the problems inherent to programming teaching and learning have been studied and investigated. The theme is very serious, not only for the important concepts underlying computer science courses but also for reducing the lack of motivation, failure, and abandonment that result from students frustration. Therefore, early identification of potential problems and immediate response is a fundamental aspect to avoid student’s failure and reduce dropout rates. In this paper, we propose a machine-learning (neural network) predictive model of student failure based on the student profile, which is built throughout programming classes by continuously monitoring and evaluating student activities. The resulting model allows teachers to early identify students that are more likely to fail, allowing them to devote more time to those students and try novel strategies to improve their programming skills.