Time-Weighted Diffusion Model for Enhanced Sequential Recommendations

The objective of sequential recommendation is to suggest items of interest to users based on their historical interaction behavior. However, user interaction data is often limited, and certain products may have sparse interactions, necessitating effective data augmentation techniques. To overcome these challenges, we propose T-DiffRec-Enhance, which integrates data augmentation with time-weighted diffusion recommendations to enhance data robustness. The diffusion models enhance data robustness through noise addition and denoising processes, while the incorporation of time weights ensures that recommendations are more closely aligned with users' recent interests. To validate the effectiveness of T-DiffRec-Ehance, we conduct experiments on two real-world datasets. The experimental results demonstrate that T-DiffRec-Ehance outperforms the baselines in recommendation performance.