How Does the Short Video Recommendation System Understand My Interest

Taking TikTok as an Example

Abstract

This paper takes TikTok as an example to study how the short video recommendation system understands user interests. Under the background of rapid development of Internet technology and widespread popularity of smartphones, short videos have become an important part of people's daily life, and TikTok uses collaborative filtering, content filtering and deep learning technology to realize accurate modeling of user interests and personalized content recommendation through data collection, user profile construction, content feature extraction and recommendation algorithm application. The study adopts literature review, case study and algorithm simulation to reveal how TikTok recommendation system improves recommendation accuracy and user engagement through multimodal data fusion and dynamic weight allocation mechanism. The experimental results show that the system demonstrates significant advantages in core indicators such as accuracy and recall rate, effectively promoting user experience and platform activity.