Social governance element risk assessment method based on dynamic gating mechanism

By processing heterogeneous graph data of social governance through dynamic gating and relational attention mechanisms, the problem of low accuracy in risk prediction caused by imbalance in node characteristics is solved, thereby improving the accuracy and robustness of node risk assessment.

CN121616094APending Publication Date: 2026-03-06TIANJIN UNIV
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202511777012.X
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-11-28
Publication Date
2026-03-06

AI Technical Summary

Technical Problem

In social governance scenarios, the imbalance in the distribution of node type characteristics leads to low accuracy in risk prediction, and the problems of noise propagation and representation bias are difficult to solve.

Method used

By employing a dynamic gating mechanism, node feature processing and unified projection are performed on heterogeneous graph datasets to generate dynamic sparse activation weights and attention aggregation results. Combined with a relational attention mechanism, a risk prediction model is trained to achieve node risk assessment.

Benefits of technology

It effectively alleviates the problems of noise propagation and representation bias caused by imbalanced node features, improves the classification accuracy and robustness of multi-type nodes, is suitable for heterogeneous graph scenarios with imbalanced features, and can adaptively adjust neighbor weights and enhance the representation ability of weak semantic nodes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121616094A_ABST
    Figure CN121616094A_ABST
Patent Text Reader

Abstract

The invention provides a social governance element risk assessment method based on a dynamic gating mechanism, and relates to the technical field of risk control, and the method comprises the steps: carrying out the data collection of a social governance system, a public security alarm record, public opinion data and an administrative report, carrying out the node feature processing and unified projection of a heterogeneous graph data set, and obtaining a node initial representation; according to the characteristics of the node initial representation, generating activation probabilities of an incoming edge and an outgoing edge to obtain a dynamic sparse activation weight, processing the node initial representation by using a relation attention aggregation mechanism to obtain an attention aggregation result, and training a risk prediction model to obtain a trained risk prediction model; and analyzing nodes of the heterogeneous graph data set by using the trained risk prediction model to obtain a social governance factor risk assessment result of each node, and completing social governance factor risk assessment. According to the invention, the problem of low risk prediction accuracy caused by difficulty in realizing balanced and robust node representation learning in the prior art is solved.
Need to check novelty before this filing date? Find Prior Art