Versioned Graph Data Retrieval for Time-Sequence Feature Computing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to efficiently manage and search large-scale graph data within service systems, making it difficult to compute graph features that are necessary for complex services like data security, risk control, and advertising.
Innovation Solution
A system and method for generating and storing graph data in multiple versions based on service data from consecutive time intervals, using a data storage system and graph computing engine to efficiently search and compute graph features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If graph data from multiple time intervals are stored and managed, then the ability to perform time sequence prediction and compute graph features across versions is improved, but the system complexity and data management burden increase
Solution Approach 1:
The patent segments graph data into multiple versions corresponding to different time intervals. Each version represents graph data at a specific time point, allowing the system to perform time sequence prediction by comparing and analyzing features across these segmented versions without managing the entire dataset as a single complex entity
Solution Approach 2:
The patent performs preliminary actions by pre-computing and storing graph features for each version of graph data. This includes calculating node features, edge features, and subgraph features in advance, so that when time sequence prediction is needed, the system can directly retrieve and compare pre-computed features rather than computing them from raw data each time
2Measurement precision
If graph features are computed for multiple versions of graph data, then the accuracy of prediction services is improved, but the computing resources and time required increase
Solution Approach 1:
The patent applies preliminary action by pre-computing graph features for multiple versions of graph data and storing them in advance. This includes calculating various types of features (node-level, edge-level, subgraph-level) before they are needed for prediction, significantly reducing the time required when actual prediction services are executed
Solution Approach 2:
The patent computes different types of graph features with different levels of detail based on local requirements. Not all versions require the same level of feature computation - the system adjusts the depth and type of feature extraction based on the specific prediction task and available resources, optimizing the balance between accuracy and computing cost
Data Source
AI summary
Implementations of this specification provide graph data processing systems and methods. An example graph data processing system includes a data storage system and a graph computing engine. The data storage system is configured to store a plurality of pieces of graph data corresponding to a plurality of versions, where the plurality of pieces of graph data are generated based on a plurality of groups of service data generated by a target service system in a plurality of consecutive time intervals. The graph computing engine is configured to receive a graph feature computing request. The graph computing engine is further configured to search the plurality of pieces of graph data stored in the data storage system for a plurality of pieces of target graph data corresponding to the plurality of target versions, and compute graph features of the plurality of pieces of target graph data.