Method and system for detecting data tables

By constructing a graph structure for data tables for anomaly detection, the problems of low accuracy in local detection and limitations of large language models in existing technologies are solved, and more efficient anomaly detection is achieved.

CN122364210APending Publication Date: 2026-07-10ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
-1 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
ALIPAY (HANGZHOU) INFORMATION TECH CO LTD
Filing Date
2026-04-30
Publication Date
2026-07-10

AI Technical Summary

Technical Problem

Existing methods for detecting anomalies in data tables have low accuracy in local detection, making it difficult to capture subtle anomalies in the neighborhood. Furthermore, they are limited by the input length of large language models, resulting in limited detection effectiveness.

Method used

A graph structure for the data table is constructed, and sub-item data and their relationships are represented by nodes and edges. Overall analysis is performed to detect abnormal sub-item data, avoiding dependence on large language models.

Benefits of technology

It improves the accuracy and reliability of anomaly detection in data tables, enhances the generalization ability and effectiveness of detection, and can capture the correlation and differences between different sub-items of data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122364210A_ABST
    Figure CN122364210A_ABST
Patent Text Reader

Abstract

The present specification provides a data table detection method and system, comprising: obtaining a data table to be detected, the data table comprising a plurality of sub-item data, one sub-item data comprising an input variable and an output result generated based on the input variable, constructing an initial relationship graph corresponding to the data table, the initial relationship graph comprising nodes and edges, one node corresponding to one sub-item data, one node comprising a representation value representing the output result of the sub-item data corresponding thereto, and an edge between two nodes representing an association relationship between the input variables of the two sub-item data corresponding to the two nodes; and analyzing the initial relationship graph to obtain abnormal sub-item data in the data table. The accuracy and reliability of the detection of abnormal sub-item data can be improved.
Need to check novelty before this filing date? Find Prior Art