A material recommendation method, device, equipment and computer readable storage medium

CN116049763BActive Publication Date: 2026-07-03MICRO DREAM TECHTRONIC NETWORK TECH CHINACO
View PDF -1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
MICRO DREAM TECHTRONIC NETWORK TECH CHINACO
Filing Date
2023-01-09
Publication Date
2026-07-03

AI Technical Summary

Technical Problem

Existing multi-objective fusion methods are complex and time-consuming in recommendation systems, and their online performance does not match offline metrics. They also consume a lot of bandwidth and are difficult to achieve efficient fusion of multi-task learning models.

Method used

By obtaining a combination of predicted data from multiple deep models, the fusion score is calculated using a preset mapping relationship. The relationship between the fusion score and the predicted data combination is automatically fitted, reducing manual parameter adjustment. A high-dimensional piecewise function is used to fit the fusion function, thereby achieving automated multi-objective fusion.

Benefits of technology

It shortens the cycle of multi-objective fusion, saves online traffic, improves the matching between online results and offline metrics, and simplifies the model adjustment process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN116049763B_ABST
    Figure CN116049763B_ABST
Patent Text Reader

Abstract

This application provides a material recommendation method applied to a recommendation system, comprising: obtaining a first estimated data combination of candidate materials from M deep learning models, wherein the first estimated data combination includes at least two of the following: estimated click-through rate, estimated interaction rate, and estimated consumption duration, where M is an integer greater than or equal to 2; determining a first fusion score corresponding to the first estimated data combination based on the mapping relationship between the estimated data combination and the fusion score, wherein the first fusion score is used to represent the target user's degree of interest in the candidate materials; and recommending materials to the target user based on the first fusion score. This application can shorten the cycle and save online traffic.
Need to check novelty before this filing date? Find Prior Art