Object Relationship Diagram Meta-Path Attention for Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current deep learning methods for predicting target indicators, such as fund performance, often overlook inherent association relationships between objects, leading to suboptimal prediction accuracy by ignoring existing connections between data points.
Innovation Solution
The proposed method constructs an object relationship diagram to represent association relationships between objects and uses meta-paths to determine attention weights, fusing this information into representation vectors to improve prediction accuracy through a feature extraction and prediction network.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If deep learning methods ignore association relationships between objects, then the model complexity is reduced, but prediction accuracy deteriorates
Solution Approach 1:
The patent introduces an object relationship diagram as an intermediary structure that captures association relationships between objects. This diagram serves as a mediator between the input data and the prediction model, allowing the model to leverage relational information without directly increasing model complexity. The relationship diagram pre-processes and organizes association information in a structured format that can be efficiently integrated into the deep learning pipeline.
Solution Approach 2:
The patent performs preliminary construction of the object relationship diagram before the main prediction process. By pre-computing and storing association relationships in the relationship diagram, the model avoids the need to compute these relationships during prediction, thus maintaining low inference complexity while improving accuracy through pre-prepared relational information.
2Measurement precision
If the object relationship diagram is constructed to capture all association relationships, then prediction accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the association relationship extraction process by focusing on specific meta-paths rather than computing all possible relationships. The object relationship diagram is constructed by identifying and processing only the relevant meta-paths that connect the target object to other objects, dividing the complex relationship extraction task into manageable segments based on predefined path patterns.
Solution Approach 2:
The patent applies local quality by determining attention weights specifically for associations along identified meta-paths rather than uniformly processing all possible object associations. The attention mechanism selectively enhances the quality of relationship representation for relevant paths while ignoring less important connections, thus improving accuracy for critical relationships without proportionally increasing overall computational complexity.
Data Source
AI summary
A method is provided. The method includes: obtaining an object relationship diagram; for a target object of a plurality of first objects, obtaining at least one meta-path corresponding to the target object in the object relationship diagram; for each meta-path, performing the following operations: determining a plurality of first attention weights of the target object based on inherent attribute data of the target object and inherent attribute data of each of a plurality of second objects on the meta-path; obtaining a second representation vector of the target object based on a first representation vector of the target object and the plurality of first attention weights; and obtaining a target indicator prediction result of the target object based at least on at least one second representation vector of the target object corresponding to the at least one meta-path.