LFCKT: A Learning and Forgetting Convolutional Knowledge Tracking Model

Abstract

Personalized exercise recommendation is a key research direction of personalized learning. In personalized exercise recommendation, we recommend suitable exercises for students according to their knowledge mastery status to improve their learning efficiency. Therefore, the accuracy of predicting students’ knowledge state in personalized exercise recommendation affects the goodness of the exercise recommendation. In the process of students’ learning, learning behavior and forgetting behavior are intertwined, and students’ forgetting behavior has a great influence on the knowledge state. In order to accurately model students’ learning and forgetting, we propose a Learning and Forgetting Convolutional Knowledge Tracking model (LFCKT) that takes into account both learning and forgetting behaviors. The model takes into account three factors that affect knowledge forgetting, including the interval time of target knowledge interaction, the count of past target knowledge interaction and student’s state of knowledge. LFCKT model uses students’ answer results as indirect feedback of knowledge mastery in the process of knowledge tracking, and integrates individual personalized learning behavior and individual forgetting behavior. Through experiments on the real online education public dataset, LFCKT can better track students’ knowledge mastery status and has better predictive performance than current knowledge tracking models.