LF-LKT: A Logistic Regression Knowledge Tracing Model Integrating Learning and Forgetting

Abstract

In the process of learning, learning behavior and forgetting behavior are interwoven, and students' forgetting behavior has great influence on Knowledge Tracing (KT). In order to accurately model learning and forgetting behaviors, this paper proposes a learning-forgetting logistic knowledge tracing (LF-LKT) model that integrated students’ forgetting factors. Three factors affecting knowledge forgetting, including the time interval of KC, KC presentation sequence (that is, the recency effect), learning opportunities of KC, and students’ response of KC are considered in this paper. We also give different level of weights to time interval. The proposed model is compared with other four models on three dataset. Results shows LF-LKT improves the predictive performance as compared to AFM, PFA . Moreover, the ablation study was conducted to investigate the influence of different factors and the result suggest the combination of three factors results in performance improvement.