View in EDS

SWDAKT: Knowledge tracing using sliding window-based dynamic ability perception.

Saved in:
Bibliographic Details
Title: SWDAKT: Knowledge tracing using sliding window-based dynamic ability perception.
Authors: Wang, Jinwei1,2 (AUTHOR) wangjinwei@tjnu.edu.cn, Lu, Jiajie1,2 (AUTHOR) lujiajie0054@stu.tjnu.edu.cn, Su, Zilong1,2 (AUTHOR) 2411090053@stu.tjnu.edu.cn
Source: Expert Systems with Applications. Mar2026, Vol. 303, pN.PAG-N.PAG. 1p.
Subjects: Learning ability testing, Cognitive ability, Long short-term memory
Abstract: • Sliding window modeling for fine-grained dynamic student-ability estimation. • Fusion of correct/incorrect difference and cognitive fluency as a novel ability feature. • Quantile-guided piecewise mapping of discrete ability labels to continuous values. • Feature-level fusion of ability-change temporality and knowledge-mastery dynamics. • Superior performance over existing models on three mainstream KT datasets. Knowledge tracing (KT) is a technique that models and predicts students' knowledge mastery by analyzing their behavioral data during learning. The introduction of deep learning models has significantly advanced KT. To better reflect students' cognitive processes, some KT models, such as dynamic student classification, classify students on the basis of their learning abilities and update the classification at each time interval to construct personalized models. However, these models generally face several limitations. First, the assessment of student ability within each time interval is static and does not reflect dynamic changes. Second, the classification of student ability does not account for cognitive fluency. Third, student ability clustering results in discrete categories with insufficient differentiation. To address these limitations, we propose a new KT model called knowledge tracing using sliding window-based dynamic ability perception (SWDAKT). Specifically, we first employ a sliding window approach to finely and dynamically model student ability within each time interval. Second, drawing on cognitive theory, we propose a two-dimensional ability feature that considers both the correct/incorrect rate difference and cognitive fluency. Third, we design quantile-guided piecewise linear mapping to convert discrete student ability labels into continuous ability values. We further conduct a comparative analysis using two different clustering methods. Finally, we integrate temporal features of ability changes extracted by LSTM with dynamic representations of knowledge mastery obtained via a dynamic key-value memory network. The experimental results demonstrate that our SWDAKT model significantly outperforms existing state-of-the-art models on three mainstream KT datasets, thereby verifying its effectiveness and superiority. [ABSTRACT FROM AUTHOR]
Copyright of Expert Systems with Applications is the property of Pergamon Press - An Imprint of Elsevier Science and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Description
Abstract:• Sliding window modeling for fine-grained dynamic student-ability estimation. • Fusion of correct/incorrect difference and cognitive fluency as a novel ability feature. • Quantile-guided piecewise mapping of discrete ability labels to continuous values. • Feature-level fusion of ability-change temporality and knowledge-mastery dynamics. • Superior performance over existing models on three mainstream KT datasets. Knowledge tracing (KT) is a technique that models and predicts students' knowledge mastery by analyzing their behavioral data during learning. The introduction of deep learning models has significantly advanced KT. To better reflect students' cognitive processes, some KT models, such as dynamic student classification, classify students on the basis of their learning abilities and update the classification at each time interval to construct personalized models. However, these models generally face several limitations. First, the assessment of student ability within each time interval is static and does not reflect dynamic changes. Second, the classification of student ability does not account for cognitive fluency. Third, student ability clustering results in discrete categories with insufficient differentiation. To address these limitations, we propose a new KT model called knowledge tracing using sliding window-based dynamic ability perception (SWDAKT). Specifically, we first employ a sliding window approach to finely and dynamically model student ability within each time interval. Second, drawing on cognitive theory, we propose a two-dimensional ability feature that considers both the correct/incorrect rate difference and cognitive fluency. Third, we design quantile-guided piecewise linear mapping to convert discrete student ability labels into continuous ability values. We further conduct a comparative analysis using two different clustering methods. Finally, we integrate temporal features of ability changes extracted by LSTM with dynamic representations of knowledge mastery obtained via a dynamic key-value memory network. The experimental results demonstrate that our SWDAKT model significantly outperforms existing state-of-the-art models on three mainstream KT datasets, thereby verifying its effectiveness and superiority. [ABSTRACT FROM AUTHOR]
ISSN:09574174
DOI:10.1016/j.eswa.2025.130735