Research Article
TDFM and TAFM: Time-Aware and Feature Fusion-Based Deep Recommendation Models for Short Videos
@INPROCEEDINGS{10.1007/978-3-031-80713-8_20, author={Bing Li and Yuqi Hou and Biao Yang}, title={TDFM and TAFM: Time-Aware and Feature Fusion-Based Deep Recommendation Models for Short Videos}, proceedings={Data Information in Online Environments. 4th EAI International Conference, DIONE 2023, Nanchang, China, November 25--27, 2023, Proceedings}, proceedings_a={DIONE}, year={2025}, month={2}, keywords={Time-Aware Feature fusion Short Video Recommendation}, doi={10.1007/978-3-031-80713-8_20} }- Bing Li
Yuqi Hou
Biao Yang
Year: 2025
TDFM and TAFM: Time-Aware and Feature Fusion-Based Deep Recommendation Models for Short Videos
DIONE
Springer
DOI: 10.1007/978-3-031-80713-8_20
Abstract
With the increasing growth of technological innovation and the emergence of technological achievements, 5G technology is driving the transformation of social and information dissemination from text to video, and video information, live streaming and short video will usher in new development opportunities. With the continuous development of the two head platforms TikTok and Kwai, the short video market pattern is gradually stabilizing, the user coverage rate is increasing and the growth rate is starting to slow down. How to recommend short videos that are more in line with users’ preferences has become a topic of more concern for platforms and users nowadays. However, the existing short video recommendation model is only based on user and item features, and lacks the perception of time features, resulting in the existing short video recommendation model is not ideal and needs to be improved. A time-aware and feature fusion-based deep recommendation model for short videos, including the TDFM model and TAFM model, is proposed in the paper. The two proposed methods are compared iteratively with traditional recommendation algorithms. This paper proposes a better method with better results. We make our experimental code public so that our experiments can be verified and reproduced. (https://github.com/lbxd123/Time-aware-recommendation.git)
