Intelligent optimization of personalized learning path based on transformer and reinforcement learning.

Area of Science:

• Artificial Intelligence in Education
• Machine Learning for Educational Data Mining
• Educational Technology

Background:

• Smart education platforms generate vast learning data, but static recommendations fail to address individual learner differences.
• Existing learning path methods lack dynamic state awareness and long-term path optimization capabilities.
• Personalized learning requires adaptive strategies that consider evolving learner states and cognitive progress.

Purpose of the Study:

• To develop an optimized personalized learning path (PLP) model using Transformer and Reinforcement Learning (RL).
• To enhance learning resource recommendation accuracy, path continuity, and learning completion rates.
• To generate learning sequences aligned with individual cognitive processes via dynamic state awareness and educational constraints.

Main Methods:

• Unified framework encoding learning behaviors, knowledge, feedback, resource access, and time intervals.
• Transformer model for capturing long-range dependencies and dynamic state features.
• Reinforcement Learning for sequential policy decision-making and optimization within educational constraints (prerequisites, difficulty, load).

Main Results:

• TFRL-Path model demonstrated superior performance across three public datasets (EdNet, ASSISTments, Junyi).
• Achieved a hit ratio of 0.846 on EdNet, outperforming UniKT by 0.025.
• Showcased improved mean average precision (0.737 on ASSISTments), completion rate (0.801 on ASSISTments), prerequisite satisfaction (0.872), and reduced repetition (0.121).

Conclusions:

• The integrated Transformer and Reinforcement Learning approach significantly improves personalized learning path recommendation quality.
• Dynamic state awareness, sequential policy optimization, and educational constraints are crucial for effective adaptive learning systems.
• TFRL-Path offers a promising solution for creating more effective and engaging personalized learning experiences.