Enhancing Sequence Movie Recommendation System Using Deep Learning and KMeans

Enhancing Sequence Movie Recommendation System Using Deep Learning and KMeans

15 March 2024 | Sophort Siet, Sony Peng, Sadriddinov Ilkhomjon, Misun Kang, Doo-Soon Park
The paper presents a deep learning-based movie recommendation system that combines deep learning techniques with KMeans clustering to enhance the accuracy and diversity of recommendations. The system addresses common challenges in traditional recommendation systems, such as data sparsity, cold start, and scalability, by leveraging user behavior sequences and demographic information. The key contributions include: 1. **Model Architecture**: The system uses a transformer architecture with multi-head attention and positional encoding to capture temporal dependencies in user movie-watching history. User demographic information and movie embeddings are integrated into the model to improve prediction accuracy. 2. **KMeans Clustering**: After training the transformer model, KMeans clustering is applied to movie genre embeddings to group similar movies. This ensures that recommendations are diverse and cover a wide range of genres. 3. **Evaluation**: The system is evaluated on two MovieLens datasets (100K and 1M) using metrics such as RMSE, MAE, precision, recall, and F1-score. The results show significant improvements over baseline models, achieving better performance in all metrics. 4. **Performance Analysis**: The paper discusses the impact of hyperparameters, such as latent dimensions, on the model's performance. Higher latent dimensions generally lead to better results, especially on larger datasets. 5. **Conclusion**: The proposed system effectively addresses the challenges of traditional recommendation systems and provides more accurate and diverse movie recommendations, making it a valuable tool for personalized movie suggestions in the era of big data and information overload.The paper presents a deep learning-based movie recommendation system that combines deep learning techniques with KMeans clustering to enhance the accuracy and diversity of recommendations. The system addresses common challenges in traditional recommendation systems, such as data sparsity, cold start, and scalability, by leveraging user behavior sequences and demographic information. The key contributions include: 1. **Model Architecture**: The system uses a transformer architecture with multi-head attention and positional encoding to capture temporal dependencies in user movie-watching history. User demographic information and movie embeddings are integrated into the model to improve prediction accuracy. 2. **KMeans Clustering**: After training the transformer model, KMeans clustering is applied to movie genre embeddings to group similar movies. This ensures that recommendations are diverse and cover a wide range of genres. 3. **Evaluation**: The system is evaluated on two MovieLens datasets (100K and 1M) using metrics such as RMSE, MAE, precision, recall, and F1-score. The results show significant improvements over baseline models, achieving better performance in all metrics. 4. **Performance Analysis**: The paper discusses the impact of hyperparameters, such as latent dimensions, on the model's performance. Higher latent dimensions generally lead to better results, especially on larger datasets. 5. **Conclusion**: The proposed system effectively addresses the challenges of traditional recommendation systems and provides more accurate and diverse movie recommendations, making it a valuable tool for personalized movie suggestions in the era of big data and information overload.
Reach us at info@study.space
[slides and audio] Enhancing Sequence Movie Recommendation System Using Deep Learning and KMeans