Design of an integrated multi-modal machine learning framework for real-time student engagement evaluation and...

Area of Science:

• Educational Technology
• Artificial Intelligence in Education
• Human-Computer Interaction

Background:

• Effective distance learning requires real-time monitoring of student engagement and cognitive load.
• High dropout rates and suboptimal learning outcomes persist in current e-learning environments.
• Integrating diverse data modalities can provide a more holistic view of student learning states.

Purpose of the Study:

• To develop and evaluate a multi-modal framework for accurate real-time prediction of student engagement, dropout risk, and cognitive load.
• To assess the impact of adaptive learning pathways generated by the framework on distance learning outcomes.
• To demonstrate the computational efficiency and practical applicability of the proposed system.

Main Methods:

• A novel multi-modal framework integrating behavioral, physiological, and interaction-based features was developed.
• Five interconnected models were employed, utilizing real-world datasets for training and validation.
• A Reinforcement Learning-based Adaptive Learning Pathway (RL-ALP) component was integrated for personalized learning optimization.

Main Results:

• The framework achieved high accuracy: 94.2% for engagement prediction, 92.3% for dropout risk, and 91.5% for cognitive load estimation.
• Adaptive learning pathways improved distance learning outcomes by 14.3% and reduced dropout rates by approximately 20.3%.
• The framework demonstrated computational efficiency with an inference time of 21.3 ms, 50% faster than previous methods.

Conclusions:

• The proposed multi-modal framework accurately monitors student engagement and cognitive load in real-time.
• Adaptive learning optimizations significantly enhance distance learning performance and student retention.
• The system provides valuable data-driven insights for educators to refine teaching strategies.