English Learning Knowledge Point Recommendation Algorithm based on Deep Deterministic Policy Gradient

Nowadays, personalized learning has become very important in educational sector for recommending knowledge points to individual learning needs adaptive systems that enhance learning efficiency and engagement drives the exploration. Many researchers have suggested various models for English learning using static models but failed to resolve existing challenges such as beginning, with limited empirical indication representing its effectiveness in English learning contexts. This study contributes by incorporating innovative Reinforcement Learning (RL) to continuously improve its recommendations based on learner feedback, making it highly responsive and personalized. The RL based knowledge recommendation framework for English learning specifically Deep Deterministic Policy Gradient (DDPG) with Knowledge Point Recommendation (KPR) algorithm. Initially, Coursera is employed for dataset and then data text cleaning, tokenization, padding is incorporated in pre-processing to convert the unstructured data to labeled data. Further, the features are selected through Recursive Feature Elimination (RFE) to eliminate unrelated features, decrease complexity and increases the model's capability to simplify to hidden data. Finally, the developed DDPG-KPR model was measured using accuracy of 97.89%. Hence, the proposed RL recommendation model provides higher accuracy value when compared to Long Short-Term Memory (LSTM).