Video recommendation method and device

By building an attention model of user behavior sequence and optimized attention, the problems of low accuracy and insufficient personalization of video recommendations in the existing technology are solved, and more accurate and personalized video recommendations are achieved, taking into account the dynamic changes and differences of user interests. change.

CN119316671APending Publication Date: 2025-01-14CHINA CONSTRUCTION BANK +1
0 Cites -1 Cited by

Patent Information

Application Number
CN202411380764.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-30
Publication Date
2025-01-14

AI Technical Summary

Technical Problem

The existing video recommendation method adopts static recommendation, which lacks considerations on the impact of time and video type on user interests, the dynamic dependence between user historical behavior and user interests, and user differentiation, resulting in low accuracy and insufficient personalization of video recommendations.

Method used

By obtaining the user's historical viewing records, dividing them into time periods, building a user behavior sequence, and using an optimized attention model, determining the recommendation probability of pre-recommended videos based on the user's short-term interest feature vector, and sorting and displaying them in time periods.

Benefits of technology

It improves the accuracy and personalization of video recommendations, takes into account the dynamic relationship between user's interest changes and historical behaviors over different time periods, and enhances support for user differentiation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119316671A_ABST
    Figure CN119316671A_ABST
Patent Text Reader

Abstract

The invention discloses a video recommendation method and device, and relates to the technical field of video recommendation, and the method comprises the steps: obtaining a historical record of watching a video by a user; dividing the historical record according to a preset time period, and constructing a user behavior sequence according to the divided historical record; determining a pre-recommended video according to the user behavior sequence; inputting the user behavior sequence into a pre-optimized video recommendation model, and outputting a recommendation probability of a pre-recommended video in each time period; according to the recommendation probability of the pre-recommended videos in each time period, sorting the pre-recommended videos according to time periods, and establishing a time-period video recommendation list; and displaying the pre-recommended videos to the user in sequence according to the time-phased video recommendation list. The technical problems of low video recommendation accuracy and insufficient individuation caused by lack of consideration of time, influence of video types on user interests, a dynamic dependency relationship between user historical behaviors and user interests and user differentiation in static recommendation in the prior art can be solved.
Need to check novelty before this filing date? Find Prior Art