Research on personalized distance education recommendation system based on deep learning

Online learning environments must be personalized with intelligent systems that propose courses according to learners’ individual preferences, activity patterns, and proficiency levels. Current recommendation models, however, fail to capture dynamic learner behavior and do not scale in big data. The research proposes a hybrid Putterfish Optimization Algorithm with Apriori (POA–Apriori)–Map-Reduce based Convolutional Neural Network (MR-CNN) methodology to address these issues, which integrates the Pufferfish Optimization Algorithm (POA) for global parameter tuning, the Apriori algorithm for association rule mining, and MapReduce-based Convolutional Neural Network (MR-CNN) for parallel deep learning. Data preprocessing was performed through normalization, feature encoding, and temporal smoothing and then model training with Udemy and Coursera datasets. The proposed model performed well with strong predictive performance of Precision@10 = 0.8769, Recall@10 = 0.8896, Normalized Discounted Cumulative Gain (NDCG)@10 = 0.8765, and lower error metrics with Mean Absolute Error (MAE)@10 = 0.2550 and Root Mean Squared Error (RMSE)@10 = 0.3568 compared to traditional models such as Support Vector Machines (SVM), K-Nearest Neighbors (KNN), and Deep Neural Networks (DNN) using both data sets. Statistical tests using paired t-tests (p < 0.05) confirmed the significance of these results. Experiments demonstrate that the hybrid POA–Apriori–MR-CNN approach significantly enhances recommendation accuracy, scalability, and robustness for massive-scale personalized education systems. The integration of optimization, association mining, and distributed deep learning marks a significant milestone towards adaptive and data-oriented e-learning environments.