Knowledge Tracing with Contrastive Learning and Attention-Based Long Short-Term Memory Network

Online learning platforms, in contrast to traditional offline classrooms, have the capacity to collect substantial learner-related data, leveraging this information to provide tailored recommendations for learners. Knowledge tracing (KT) plays a crucial role in this task, aiming to evaluate learners’ evolving knowledge status (KS) based on historical learning records and forecast their future performance. In this paper, we propose CALSKT, a KT model constructed by incorporating contrastive learning (CL) and a long short-term memory (LSTM) network based on attention mechanism (ALSTM). Specifically, ALSTM is employed to integrate information from learners’ previous KS and emphasize the current KS. Subsequently, we construct a CL Framework aimed at extracting self-supervision signals from learners’ original learning history and revealing semantically similar or distinct historical data samples. Within the CL Framework, we design four data augmentation methods to enrich the original data. Extensive experiments are conducted on five datasets, comparing CALSKT with six baselines, and the experimental results demonstrate that CALSKT achieves a fantastic result.