Recommendation model training method and device, recommendation method and device and electronic equipment
CN121145995APending Publication Date: 2025-12-16NAT UNIV OF DEFENSE TECH
Patent Information
- Application Number
- CN202511245423.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-02
- Publication Date
- 2025-12-16
AI Technical Summary
Technical Problem
Existing graph contrastive learning methods do not consider the temporal constraints of behavioral sequences, causing the model to identify non-causal relationships as genuine interests, which weakens the scenario adaptability of the recommendation system and results in poor recommendation accuracy.
Method used
By generating first and second enhanced contrast views, and combining Bayesian personalized ranking loss function and time-aware graph contrast learning, a multi-task learning framework is constructed, incorporating time constraints to alleviate the problem of insufficient temporal correlation.
Benefits of technology
This improved the noise resistance and time constraint of the recommendation model, and enhanced the accuracy of the recommendation results.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN121145995A_ABST
Abstract
The invention discloses a recommendation method, and particularly relates to a training method and device of a recommendation model, a recommendation method and device and electronic equipment. The recommendation model training method comprises the steps of obtaining a to-be-trained recommendation model and an original training interaction information graph; generating two enhanced contrast views based on the original training interaction information graph; inputting the original training interaction information graph into a to-be-trained recommendation model for training, calculating to obtain sorting loss based on a Bayesian personalized sorting loss function, and inputting the two enhanced comparison views into the to-be-trained recommendation model for comparison learning training to obtain comparison loss; determining a total loss based on the sorting loss and the comparison loss; and under the condition that the total loss meets a preset condition, determining the currently trained recommendation model to be trained as a recommendation model. According to the method, the performance of the recommendation model can be improved, and a recommendation result with higher accuracy is obtained.
Need to check novelty before this filing date? Find Prior Art
Citation Information
Cited By
Interaction prediction model training method and device, equipment and storage medium
CN121328646A