CN-120851601-B - Intelligent risk auditing and early warning method and system integrating knowledge graphs

CN120851601BCN 120851601 BCN120851601 BCN 120851601BCN-120851601-B

Abstract

The invention discloses an intelligent risk auditing and early warning method and system integrating knowledge graphs, and relates to the technical field of intelligent early warning. The method can extract path nodes and dependency relations from historical archive risk events, construct a structural map with hierarchical semantics, realize quantitative analysis of path behavior rhythms and structures through periodic behavior tracks and track offset vectors, accurately identify path structure abnormal events and generate early warning candidate sets based on comparison of periodic offset coding sequences and structure consistency in the auditing process, realize structural labeling of risk states by combining jump nodes and superior path relations, trigger a chain response mechanism and form a complete linkage risk node set.

Inventors

Min Xurong
ZHANG TINGFU
YU XIAOGANG

Assignees

南京财信网络科技有限公司

Dates

Publication Date: 20260508
Application Date: 20250708

Claims (8)

1. The intelligent risk auditing and early warning method integrating the knowledge graph is characterized by comprising the following steps: In a knowledge graph of service data mapping, extracting related entity path nodes and path dependency relations based on historical archived risk event data, and constructing a hierarchical structure graph; aiming at path nodes marked as risks, sampling behavior occurrence tracks in different period sections in a hierarchical structure map, extracting time interval characteristics of path behaviors and track offset vectors, and constructing a period offset coding sequence; based on the hierarchical structure map and the periodic offset coding sequence, carrying out structural consistency comparison on the currently-occurring auditing path node sequence and the historical risk path, confirming a path structure abnormal event, and writing in an early warning candidate event set; Extracting upper and lower path relations for each risk node in the early warning candidate event set, marking the current node state as a structure deviation risk in the hierarchical structure map, triggering the upper path nodes communicated with the current node state to send out a chain early warning response, and encoding the current path nodes into a linkage risk node set; The construction of the hierarchical structure map comprises the steps of extracting entity objects, action and triggering sequence information from the archived risk event logs in historical wind control data to construct a triplet sequence set, limiting reasonable triggering relations among continuous action by using rule constraint based on the triplet sequence set, applying shielding treatment to path branches with unsatisfied rule constraint when generating path map nodes, and only reserving structure paths meeting the rule constraint dependence to form a candidate structure path set; The construction of the periodic offset coding sequence comprises the steps of extracting a time interval sequence of behaviors among nodes from each section of periodic track, calculating an average time interval characteristic value in the track, constructing a track difference vector by taking a reference periodic segment as a reference, and quantifying the time behavior offset degree; The method comprises the steps of comparing structural consistency, confirming a path structure abnormal event, namely constructing a structural vector according to a position relation in a hierarchical structure map according to a path node sequence formed in a current auditing behavior, wherein each dimension represents a structural displacement distance between a node and an upper node, mapping the structural vector to a history path set stored in the hierarchical structure map, comparing the structural vector with the matched history path structural vector in a bit-by-bit difference mode, and judging that the current path structure is in an abnormal state if the structural difference exceeds an offset threshold value and a behavior rhythm change value in a periodic offset coding sequence of the corresponding path node exceeds a drift amplitude.
2. The intelligent risk auditing and early warning method based on the fusion knowledge graph of claim 1 is characterized in that the rule constraint is as follows: Extracting all behavior actions to form a set based on the marked directional triplet set in the graph behavior sequence And constructing a trigger sequence matrix based on the co-occurrence sequence frequency in the history path Wherein the elements are Representing behavior In-behavior The probability density of the previous occurrence is determined, Composing the total number of different types of behavioral actions contained in the collection for the action; Based on a trigger sequence matrix If and only if And behaviors Post-behavior in hierarchical structure graphs Defining path dependence rule function as 1, otherwise path dependence rule function as 0, representing behavior Failure to satisfy the behavior of Triggering path constraint, thereby branching path ) A blocking mark is applied and the blocking mark is applied, Is the minimum trigger probability threshold.
3. The knowledge graph-fused intelligent risk auditing and early warning method as claimed in claim 1, characterized in that the path node sequence meeting the structural abnormality judgment condition Marked as abnormal path identification, extracting a jump node set And corresponding period offset coding In triplets of Is written into the early warning candidate event set.
4. The intelligent risk auditing and early warning method for fusing the knowledge graph as set forth in claim 1, wherein the labeling the current node state in the hierarchical structure graph as the structure offset risk comprises: for each group of triplets in the early warning candidate event set Traversing a set of hopping nodes Searching a direct superior path node set in the hierarchical structure map to form a communication path relation chain; And adding a structural deviation state label to the hierarchical structure map path nodes, and simultaneously calculating a risk level factor by combining the maximum rhythm drift value in the periodic deviation coding sequence.
5. The intelligent risk auditing and early warning method based on the integrated knowledge graph of claim 4, wherein the triggering and communicating the superior path node to send out the chain type early warning response comprises the following steps: and according to the communication path relation chain and the attached risk level factors, sending out structural deviation early warning to the upper path node set step by step, and recording the nodes marked by linkage as a linkage risk node set according to the early warning response sequence.
6. The intelligent risk auditing and early warning system integrating the knowledge graphs is characterized by further comprising the following steps of: the map construction module is used for extracting related entity path nodes and path dependency relations based on the risk event data archived in the history in the knowledge map mapped by the service data to construct a hierarchical structure map; the behavior sampling module is used for sampling behavior occurrence tracks in different period sections in the hierarchical structure map aiming at path nodes marked as risks, extracting time interval characteristics of the path behaviors and track offset vectors, and constructing a period offset coding sequence; The structure comparison module is used for comparing the structural consistency of the currently-occurring auditing path node sequence and the historical risk path based on the hierarchical structure map and the periodic offset coding sequence, confirming the path structure abnormal event and writing in the early warning candidate event set; The chain type linkage module is used for extracting upper and lower path relations for each risk node in the early warning candidate event set, marking the current node state in the hierarchical structure map as a structure deviation risk, triggering the upper path nodes communicated with the current node state to send out chain type early warning response, and encoding the current path nodes into a linkage risk node set.
7. The computer equipment comprises a memory and a processor, wherein the memory stores a computer program, and the computer equipment is characterized in that the processor realizes the steps of the intelligent risk auditing and early warning method integrating the knowledge graphs according to any one of claims 1-5 when executing the computer program.
8. The intelligent risk auditing and early warning method based on the integrated knowledge graph is characterized in that the intelligent risk auditing and early warning method based on the integrated knowledge graph is realized according to any one of claims 1-5 when the computer program is executed by a processor.

Description

Intelligent risk auditing and early warning method and system integrating knowledge graphs Technical Field The invention relates to the technical field of intelligent early warning, in particular to an intelligent risk auditing and early warning method and system integrating knowledge graphs. Background Along with the wide application of artificial intelligence, big data and knowledge graph technology in the field of enterprise wind control, the intelligent and structural trend of enterprise risk auditing and early warning is increasingly obvious. Traditional rule-driven wind control methods are difficult to meet the requirements of dynamic and changeable business environments, and a atlas wind control model based on historical behavior tracks, structural evolution modes and multi-level path dependency relations becomes an emerging research hotspot. The knowledge graph can bear a large amount of heterogeneous data and behavior logic through entity relation modeling, and becomes an important infrastructure for structural modeling and context reasoning in business auditing step by step. However, most of current knowledge graph application still stays in a static clustering or rule screening stage, a dynamic risk behavior modeling mechanism for multi-period and structural evolution is not established, and particularly, a recognition and response mechanism for advanced semantic risk features such as path node structural abnormality and track offset is lacking. The comparison document CN116645189A discloses an enterprise risk early warning method, an enterprise risk early warning device, an electronic device and a readable storage medium, wherein a plurality of enterprises in an enterprise knowledge graph are aggregated into an entity set based on geographic features and/or business features, and correlation analysis is carried out on an object to be early warned through an entity attribution relationship so as to output early warning results such as risk items, suspected risk items and the like. According to the scheme, the real-time performance and the efficiency of risk early warning are improved, and early warning can be realized for a new customer under the condition of lacking historical data. However, the method still uses static characteristics (such as geographic or commercial information) to drive analysis, does not model multi-period path behaviors of enterprises, lacks dynamic tracking of hierarchical path relationships and structure change characteristics among entities, and is difficult to identify a risk recurrence mode based on structure evolution. Therefore, limitations still exist in terms of path-dependent structural analysis, periodic behavior change modeling, risk linkage early warning and the like. The comparison document CN119990719A provides a receipt information input method and system based on big data processing, and combines a knowledge graph to realize risk category identification and automatic early warning. The method has the advantages that the method can support multidimensional compliance check and dynamic rule update, and the auditing accuracy is improved through the auditing rule driven by big data. However, the early warning logic of the method still depends on the direct matching of the attribute of the bill object and the sensitive element, lacks a multiplexing and dynamic alignment mechanism of a historical risk path structure, does not relate to abnormal track identification based on a path structure map, does not provide linkage relation modeling capability between risk nodes, and is difficult to cope with chained risk response requirements in the structural risk evolution process. Disclosure of Invention The invention is provided in view of the fact that most of existing risk auditing methods are based on static attribute characteristics for analysis, and are difficult to support deep risk identification based on path structure evolution. Therefore, the problem to be solved by the invention is how to realize the active identification of the abnormal path nodes and the linkage response of the risk chains by a periodic track sampling and structure consistency comparison technology. In order to solve the technical problems, the invention provides the following technical scheme: The invention provides an intelligent risk auditing and early warning method integrating knowledge graphs, which comprises the steps of extracting related entity path nodes and path dependency relations based on historical archived risk event data in a knowledge graph of service data mapping, constructing a hierarchical structure graph, sampling behavior occurrence tracks in different period sections in the hierarchical structure graph aiming at path nodes marked as risks, extracting time interval characteristics and track offset vectors of the path behaviors, constructing a periodic offset coding sequence, comparing structural consistency of the current auditing path node sequence and a historical