A Survey on Optimal Learning Models to Monitor Student Progress in Knowledge Tracing

Abstract

Understudies' ability to study in web-based learning frameworks has improved as a result of Knowledge tracing (KT), which refers to following understudies' shifting information state as they learn. Due to its fundamental significance in education, KT has recently attracted considerable examination attention. Nevertheless, the bulk of contemporary KT tactics aim for high accuracy in understudy execution expectations while ignoring the consistency between understudies' shifting information states and their learning styles. In this article, we explore an alternative worldview for the KT job and offer a creative model called Learning process consistent Knowledge Tracing (LPKT) and LPKT-S that checks learners' knowledge levels by plainly demonstrating their preferred method of learning. In specifically, the fundamental learning cell is initially formalised as the tuple practise answer time reply. Then, using the difference between the present and past learning cells, their duration, and the related information state of the students, we carefully measure the learning gain as well as its diversity. In order to determine the amount of information that students can absorb, we also design a learning door. In addition, we design an ignoring door to show how understudies' information deteriorates over time depending on their prior information condition, current learning gains, and time span. Broad testing findings on an open dataset demonstrate how LPKT could obtain more accurate information state as experience grows. Furthermore, LPKT also outperforms modern KT techniques in terms of expected understudy performance. Our work illustrates a potential future test bearing for KT that has a high level of precision and interpretability.