A big data-based data cleaning system
By standardizing multi-source heterogeneous data and constructing a weighted directed graph, and using dynamic graph neural networks and variational energy models for data cleaning, the problem of data quality control in dynamic spatiotemporal coupling environments is solved, and automated data repair and adaptive maintenance are achieved.
Patent Information
- Application Number
- CN202610696712.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2026-05-20
- Publication Date
- 2026-08-18
- Estimated Expiration
- 2046-05-20
AI Technical Summary
Existing technologies struggle to eliminate dimensional differences in multi-source heterogeneous data under dynamic spatiotemporal coupling environments, neglect topological connections and temporal evolution characteristics between nodes, and lack closed-loop control mechanisms for automatically correcting anomalous data, making it difficult for data quality control to meet the needs of large-scale dynamic systems.
The data preprocessing center standardizes multi-source heterogeneous data, constructs a weighted directed graph, generates state space vectors using dynamic graph neural networks, calculates global consistency energy using variational energy models, repairs data using energy gradient scalars, and sets up a feedback control unit to achieve automated closed-loop control.
It effectively eliminates dimensional differences, accurately captures features between nodes, realizes the verification and automatic repair of deep logical conflicts, improves the accuracy of data processing and the convergence speed of repair, and has adaptive learning capabilities to adapt to environmental changes.