DeepLMS: a deep learning predictive model for supporting online learning in the Covid-19 era.

Area of Science:

• Educational Technology
• Artificial Intelligence
• Machine Learning

Background:

• The COVID-19 pandemic necessitated a rapid shift to online learning environments (OLEs).
• Learning Management Systems (LMSs) and video conferencing platforms became crucial for remote education.
• Predictive models for learner behavior in OLEs are needed to support educators and learners.

Purpose of the Study:

• To introduce DeepLMS, a novel predictive model for forecasting the Quality of Interaction (QoI) with LMS.
• To demonstrate the application of Deep Learning techniques, specifically Long Short-Term Memory (LSTM) networks, for analyzing LMS user interaction data.
• To provide a tool that enhances online learning engagement and offers educators new evaluation methods.

Main Methods:

• Utilized Long Short-Term Memory (LSTM) networks, a type of Deep Learning model.
• Developed and tested the DeepLMS model on user interaction data from LMS databases.
• Evaluated the model's performance using Root Mean Square Error (RMSE) and correlation coefficients.

Main Results:

• DeepLMS achieved an average testing RMSE of [Formula: see text].
• The model demonstrated an average correlation coefficient of [Formula: see text] between predicted and actual QoI values.
• Performance was validated on datasets from pre- and during the COVID-19 pandemic.

Conclusions:

• Deep Learning techniques, via DeepLMS, can effectively predict learner interaction quality in OLEs.
• DeepLMS offers personalized QoI forecasting to scaffold user engagement.
• The model provides educators with supplementary evaluation insights into learner motivation and participation.