This project presents a personalized content recommendation engine for an e-learning platform using deep learning techniques. It implements and compares two prominent models — Neural Collaborative Filtering (NCF) and Light Graph Convolutional Network (LightGCN) — to recommend educational content to users based on their engagement patterns.
The system is designed to address challenges of sparsity and personalization in online learning environments. Key features include:
- Comparative implementation of NCF and LightGCN
- Performance evaluation using accuracy, precision, recall, NDCG, MAP, and ROC-AUC
- Integration-ready for use with a web-based e-learning platform
- Visualizations of model results using bar plots and heatmaps
The dataset used is the Online Course Engagement Dataset from Kaggle, which includes features such as course completion, views, user feedback, and engagement scores.
- Implemented in TensorFlow/Keras
- Embedding-based user-item interaction
- Dense MLP layers for non-linear learning
- Implemented in PyTorch
- Simplified graph convolution architecture
- Optimized for sparse recommendation tasks
| Metric | Description |
|---|---|
| Accuracy | Classification accuracy for implicit ratings |
| Precision@K | Relevance of top-K recommendations |
| Recall@K | Coverage of relevant items in top-K |
| NDCG@K | Ranking quality of recommendations |
| MAP@K | Mean Average Precision |
| ROC-AUC | Binary classification performance (NCF) |
| Model | Precision | Recall | NDCG | MAP | ROC-AUC | Accuracy |
|---|---|---|---|---|---|---|
| NCF | 0.5723 | 0.5403 | — | — | 0.6226 | 0.5653 |
| LightGCN | 0.6800 | 0.6800 | 3.628 | 0.6521 | — | — |
LightGCN outperformed NCF in sparse and top-K ranking scenarios, making it a better fit for educational recommendation use cases.
- Python
- TensorFlow / Keras
- PyTorch
- NumPy, Pandas, Matplotlib, Seaborn
- Scikit-learn (for evaluation metrics)