BKT-LSTM: Efficient Student Modeling for knowledge tracing and student performance prediction

Recently, we have seen a rapid rise in usage of online educational platforms.\nThe personalized education became crucially important in future learning\nenvironments. Knowledge tracing (KT) refers to the detection of students'\nknowledge states and predict future performance given their past outcomes for\nproviding adaptive solution to Intelligent Tutoring Systems (ITS). Bayesian\nKnowledge Tracing (BKT) is a model to capture mastery level of each skill with\npsychologically meaningful parameters and widely used in successful tutoring\nsystems. However, it is unable to detect learning transfer across skills\nbecause each skill model is learned independently and shows lower efficiency in\nstudent performance prediction. While recent KT models based on deep neural\nnetworks shows impressive predictive power but it came with a price. Ten of\nthousands of parameters in neural networks are unable to provide\npsychologically meaningful interpretation that reflect to cognitive theory. In\nthis paper, we proposed an efficient student model called BKT-LSTM. It contains\nthree meaningful components: individual \\textit{skill mastery} assessed by BKT,\n\\textit{ability profile} (learning transfer across skills) detected by k-means\nclustering and \\textit{problem difficulty}. All these components are taken into\naccount in student's future performance prediction by leveraging predictive\npower of LSTM. BKT-LSTM outperforms state-of-the-art student models in\nstudent's performance prediction by considering these meaningful features\ninstead of using binary values of student's past interaction in DKT. We also\nconduct ablation studies on each of BKT-LSTM model components to examine their\nvalue and each component shows significant contribution in student's\nperformance prediction. Thus, it has potential for providing adaptive and\npersonalized instruction in real-world educational systems.\n