An adaptable and personalized framework for top-N course recommendations in online learning

Area of Science:

• Educational Technology
• Artificial Intelligence
• Machine Learning

Background:

• Massive Open Online Courses (MOOCs) offer vast learning opportunities but can overwhelm learners due to information overload.
• Personalized Recommender Systems (RSs) are crucial for navigating educational content effectively.
• Existing machine learning and reinforcement learning (RL) methods have limitations for multi-task recommendations in e-learning.

Purpose of the Study:

• To develop an adaptive recommender system for MOOCs that addresses the limitations of traditional methods.
• To personalize the learning experience by considering learner-specific factors.
• To improve resource acquisition and learner engagement in online learning environments.

Main Methods:

• Integration of a Deep Reinforcement Learning (DRL) framework with a multi-agent approach.
• Development of a DRL-based Actor-Critic model (DRR) for sequential decision-making in recommendations.
• Incorporation of learner sentiments, learning style, preferences, competency, and adaptive difficulty levels into the personalization process.

Main Results:

• The proposed DRR model demonstrated superior performance over baseline models in extensive experiments on a MOOC dataset.
• The system effectively provides top-N course recommendations and personalized learning paths.
• Validation on the 100K Coursera course review dataset confirmed the model's effectiveness for long-term recommendations.

Conclusions:

• The DRR model offers a robust solution for personalized recommendations in MOOCs.
• This research contributes to advancing e-learning technology by guiding the design of effective course recommender systems.
• The findings facilitate more relevant and personalized recommendations, enhancing the overall online learning experience for students.