Personalized recommendation of english learning resources based on collaborative filtering algorithm in english teaching scenarios

Traditional user-based Collaborative Filtering (CF) often struggles with sparse rating data in English teaching scenarios, making it difficult to identify relevant resources and limiting recommendation accuracy. To overcome this limitation, this study proposes an improved Alternating Least Squares (ALS)-based collaborative filtering algorithm designed to alleviate data sparsity and enhance recommendation performance. The dataset used for evaluation was collected from multiple English online learning platforms and included over 400,000 learner interaction records and more than 130,000 explicit ratings on courses and articles. After constructing the learner–resource rating matrix, the ALS-based model was trained to minimize prediction error and generate personalized recommendations. Experimental results demonstrated high performance, with a Matthews Correlation Coefficient (MCC) of 0.97 for advanced grammar resources and a Mean Absolute Error (MAE) of 0.01 when tested with large-scale data (up to 1500GB). User feedback from 1,140 valid questionnaires revealed that more than 40% of learners rated the system with the highest satisfaction score, validating both the accuracy and usability of the approach. The findings imply that the improved ALS method can effectively address data sparsity issues, deliver stable and precise recommendations in diverse English teaching contexts, and provide educators and learners with a scalable tool to personalize learning pathways, ultimately improving engagement and learning outcomes.