Automated Test Set Quiz Maker Optimizing Solving Time and Parameters of Bayesian Knowledge Tracing Model Extracted from Learning Log

Abstract

Creating a set of quizzes for the students' test is almost an irreplaceable task to teachers. In practice, a teacher could use a learning analytics dashboard while creating a test to control the quiz difficulty and the amount of time it takes to solve. This paper draws inspiration from this practical example, and we propose an automated test set quiz maker by optimizing the time and learning parameters that have been estimated from the analysis of learning logs. First, we estimate the Bayesian Knowledge Tracing (BKT) model parameters: in particular the guess and slip probability from the quiz answer log history. The system automatically generates a test set of quizzes if the user inputs the desired amount of time it should take to solve the test by optimizing the selection of quizzes based on BKT parameters and estimated solving time. This function is expected to reduce the burden of preparing examination questions for teachers, and it can be used as a trial test before the exam for students.