SAINT+: Integrating Temporal Features for EdNet Correctness Prediction

We introduce an addition to SAINT in SAINT+, where temporal features are added to the list of inputs used to train our model.

Definition:

SAINT(Separated Self-AttentIve Neural Knowledge Tracing) is one of our knowledge tracing model that captures complex relations between educational exercises and student responses to assess a student’s knowledge state. Learn more about it here

Empirical evaluations show a 1.25% increase in AUC compared to SAINT, and a maximum 3.61% increase in AUC compared to other deep learning-based knowledge tracing models such as DKT, DKVMN, and SAKT.