Knowledge Graph Embedding Fusion for Recommendation Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current knowledge graph-based recommendation systems only analyze low-order information, leading to low accuracy in recommended commodity results and inefficient use of computing resources due to repeated prediction processes.
Innovation Solution
The method involves obtaining account and item entity relations from a knowledge graph, converting these into embedding vectors, and fusing them to create more expressive account and item representations, which improves the accuracy of recommendations by capturing high-order structured information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If only low-order information in the knowledge graph is analyzed, then the processing complexity is reduced, but the accuracy of recommended commodity results deteriorates
Solution Approach 1:
The patent transitions from analyzing only low-order information to incorporating high-order structured information by converting knowledge triplets into embedding vectors and performing multi-dimensional fusion. This dimensional expansion enables the system to capture complex relationships between accounts and items while maintaining manageable processing complexity through efficient vector operations.
Solution Approach 2:
The patent creates composite representations by fusing multiple types of information (account embedding vectors, item embedding vectors, and relation embedding vectors) into unified account and item representations. This composite approach integrates diverse data sources to improve recommendation accuracy without proportionally increasing processing complexity.
2Use of energy by stationary object
If low-order information is used for analysis, then the computing resources required are reduced, but repeated prediction processes are triggered resulting in waste of computing resources
Solution Approach 1:
The patent performs preliminary conversion of knowledge triplets into embedding vectors and pre-computation of relation embeddings before the actual recommendation process. This preliminary action prepares the data in an optimized format that enables single-pass high-accuracy predictions, eliminating the need for repeated prediction processes and wasting computing resources.
Solution Approach 2:
The patent changes the representation parameters from raw low-order information to transformed high-order embedding vectors. This parameter transformation enables the system to achieve high accuracy in a single prediction pass, improving productivity while maintaining efficient computing resource usage through optimized vector operations.
3Measurement precision
If high-order structured information is captured through embedding vector fusion, then the accuracy of recommendations is improved, but the processing complexity increases
Solution Approach 1:
The patent replaces complex mechanical information processing with mathematical embedding vector operations. By substituting traditional complex analysis mechanisms with efficient vector transformations and fusion operations, the system achieves high-order structured information capture without proportionally increasing processing complexity.
Solution Approach 2:
The embedding vector representation serves multiple functions simultaneously: it captures account information, item information, and relation information in a unified structure. This multi-functionality enables high-accuracy recommendations while avoiding the need for separate complex processing steps for each information type.
Data Source
AI summary
An account entity relation between a target account entity and a neighbor account entity is obtained. An item entity relation between a target item entity and a neighbor item entity is obtained. The account entity relation is converted into an account relation embedding vector. The item entity relation is converted into an item relation embedding vector. Under supervision of a target item embedding vector, a target account embedding vector and a neighbor account embedding vector are fused through the account relation embedding vector into a target account representation. Under supervision of a target account embedding vector, a target item embedding vector and a neighbor item embedding vector are fused through the item relation embedding vector into a target item representation. A target item for a target account of the target account entity is determined from the target item entity based on the target account representation and the target item representation.