Implementing federated learning for privacy-preserving emotion detection in educational environments.

Area of Science:

• Educational Technology
• Artificial Intelligence
• Data Privacy

Background:

• Emotion detection is vital for student success but faces privacy and scalability issues with traditional models.
• Current systems often transfer sensitive student data, compromising confidentiality and limiting widespread adoption.

Purpose of the Study:

• To develop and evaluate a federated learning-based emotion detection model for educational settings.
• To address data privacy and scalability limitations inherent in centralized emotion detection systems.

Main Methods:

• A federated learning model was developed to process emotional data locally on student devices.
• The model was integrated into the Moodle platform and evaluated using advanced anonymization and preprocessing techniques.

Main Results:

• The model achieved high performance metrics: 87% precision, 85% recall, and 86% F1-score.
• Performance remained robust under challenging conditions like low lighting and ambient noise.
• Observed a 15% increase in academic participation and a 12% improvement in average academic performance.

Conclusions:

• Federated learning offers a privacy-preserving, scalable, and effective solution for emotion detection in education.
• The proposed system positively impacts educational dynamics, improving student engagement and academic outcomes.