Author(s)

Lyu Jiaqi

Corresponding Author

Lyu Jiaqi

Abstract

In recent years, the rapid growth of cloud computing, big data, mobile Internet and various video services has led to the explosive growth of micro-video data. How to quickly and effectively find videos that users are interested in and provide personalized recommendation services has become an inevitable trend of micro-video business development. In this article, a three-dimensional DNN is proposed to extract the depth features of micro-videos, and a user behavior preference model with joint depth feature labels is constructed to better describe the user's behavior preferences. The test results show that, compared with FCA, this method has obvious advantages in the later stage of operation, and the error is reduced by 33.84%. The accuracy of this algorithm for micro-video image feature recognition is higher, which is 21.82% higher than that of the comparison algorithm. The model abstract, by introducing the relevant weight update feedback, updates the user's behavior preference in real time, and significantly improves the accuracy of personalized recommendation results of micro-videos.

Keywords

Micro-Video; Personalized Recommendation; Neural Network