Local Resource Recommendation Model for Privacy-Preserving Personalization

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current resource recommendation systems in terminal devices become cumbersome due to the large number of historical usage records, requiring users to sift through numerous resources, and there is a lack of personalized and intelligent recommendations.

Innovation Solution

Implement a local recommendation model in terminal devices that updates using an online learning algorithm, leveraging current running information to provide personalized recommendations, and integrate a server-based basic model for initial training, with local updates and parameter fusion to maintain user privacy.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If resources are sorted by most recent usage time, then the resource list comprehensively reflects historical usage records, but the number of recommended resources increases causing users to need extensive searching or browsing

Engineering Contradiction:

Improvecomprehensiveness of resource recommendationVSAvoidtime for searching or browsing resources

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent changes the recommendation parameter from simple chronological sorting to probability-based sorting. The resource list is reordered according to calculated recommendation probabilities that consider multiple factors including usage frequency, recency, and resource characteristics, thereby optimizing the balance between comprehensiveness and search efficiency

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces the mechanical chronological sorting mechanism with an intelligent probability calculation system. Instead of mechanically ordering resources by timestamp, the system uses computational models to calculate recommendation probabilities and sorts resources accordingly, reducing user search time while maintaining comprehensiveness

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Object-affected harmful factors

If user data is processed locally in terminal device, then privacy leakage risk is reduced, but computational resources and model maintenance burden increase

Engineering Contradiction:

Improveprivacy leakage riskVSAvoidmodel maintenance burden

Core Design Contradiction:

Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The patent segments the recommendation system into two parts: a lightweight local model deployed in the terminal device for privacy-preserving real-time recommendations, and a cloud-based training system for model updates. This segmentation allows local processing to protect privacy while distributing the heavy computational burden of model training to the cloud

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an online learning algorithm as an intermediary mechanism that enables the local model to continuously adapt and improve using user interaction data without requiring data to leave the device. This intermediary allows the system to maintain model performance while preserving privacy and reducing maintenance complexity

Inventive Principle:

Principle #24Intermediary (Mediator)

Data Source

PatentUS12619671B2Resource list recommendation method, terminal device, and server

Publication Date: 2026.05.05 ZTE CORP

AI summary

Disclosed are method for recommending a resource list, a terminal device, and a server. The method may include: acquiring current running information in response to detecting that the interface is invoked; inputting the running information into the local recommendation model to obtain recommendation probabilities of resources in the resource list; and selecting a to-be-recommended resource according to the recommendation probabilities and displaying the to-be-recommended resource.