Modeling student learning outcomes in studying programming language course
Shanshan Wang, Yong Han, Wenjun Wu, Zhenghui Hu · 2017
Learning outcome assessment is of great significance in the field of traditional on-campus teaching especially on the courses of programming languages. In this work, we take advantage of the data offered by our programming assignment judge system and propose a new IRT-BKT model for estimation of learning outcome. This new framework: Item Response Theory (IRT) model that estimates students' initial knowledge status, and joins it with the discrimination and difficulty of each skill estimated to evaluate the probability of knowing a skill before training it. We then estimate parameters learn, guess, and slip probabilities of Bayesian Knowledge Tracing (BKT) Model. Using real data, we show that IRT-BKT model outperforms Item Response Theory and Bayesian Knowledge Tracing in terms of prediction accuracy.