An Approach for Dealing with Sequential Data in Intelligent Tutoring Systems

Tóm tắt

The use of educational data to gain deeper
insights into learners’ interaction histories with Intelligent
Tutoring Systems (ITS) is receiving increasing attention,
especially in the context of online learning and the growing
demand for digital transformation in education. Predicting
learners’ academic performance through the analysis and
evaluation of their recorded activities in ITS plays a critical
role in supporting educational administrators and instructors.
An understanding of learners’ abilities helps refine teaching
methods and optimize learning environments, ultimately en
hancing educational quality. Our research improves upon our
previous work, which utilized only LSTM, by incorporating
an ensemble model- CL-PSP, combining CNN and LSTM
networks. Specifically, the study focuses on predicting learn
ers’ CFA capability- the likelihood of learners answering
correctly on their first attempt. Knowledge evolves over time,
observed in users’ interaction preferences within session
based recommendation systems. CL-PSP leverages critical
factor in shaping learners’ academic performance in two
educational datasets, KDD Cup 2010 and Assistment 2017.
Several enhancements, including a revised ensemble architec
ture, improved error measurement, and refinements in data
preprocessing. Results demonstrate that the proposed model
significantly outperforms existing models, achieving superior
performance with a lower Root Mean Square Error (RMSE).
On the KDD Cup 2010 dataset, the model achieves a mini
mum RMSE of 0.375, while notable improvements are also
observed on the Assistment 2017 dataset, further underscoring
the model’s effectiveness and robustness. The experimental
results underscore the feasibility and considerable potential of
utilizing session-based data in ITS to enhance both learning
performance and educational quality.