Incremental updating and querying method and device of multi-level timing diagram, equipment and medium
By converting multi-source heterogeneous data into standard event tuples and mapping them to multi-level time series graphs, the problems of data fragmentation and low query efficiency in blockchain risk prevention and control scenarios are solved, enabling real-time updates and fast queries.
Patent Information
- Application Number
- CN202611182825.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-05
- Publication Date
- 2026-11-17
AI Technical Summary
In blockchain risk prevention and control scenarios, existing technologies suffer from data fragmentation, high update costs, and low query efficiency, which cannot meet the needs of real-time risk identification.
Multi-source heterogeneous data is uniformly converted into standard event tuples and mapped into multi-level time series graphs. Local influence domains are determined through time proximity, feature similarity, and topological correlation for incremental updates. Fast queries are performed using a pre-maintained set of intermediate states and risk path indexes.
It achieves unified data representation and real-time updates, reduces update costs, improves query efficiency, and supports rapid risk path query and interpretation.
Smart Images

Figure CN123115236A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical fields of blockchain and data processing, and in particular to a method, apparatus, device and medium for incremental updating and querying of multi-level time sequence graphs. Background Technology
[0002] In blockchain risk control scenarios, multi-level data such as on-chain transactions, contract calls, node network status, consensus anomalies, cross-chain bridge events, and off-chain verification feedback continuously arrive. Existing technologies typically adopt the approach of "batch collection of on-chain data + offline graph calculation / rule scanning + risk control system query": first, blockchain data is obtained in batches through node synchronization or block parsing, then imported into a relational database or graph database, offline tasks calculate results such as address risk scores and fund paths, and finally the risk control system queries these offline results according to user requests.
[0003] However, existing solutions have significant shortcomings: First, the update cost is high. Adding new blocks or transactions can change local fund paths and address risks, but existing solutions often require periodically rerunning large-scale graph calculations, which cannot meet the needs of real-time risk identification. Second, there is multi-level semantic fragmentation. Transaction logs, contract calls, consensus anomalies, cross-chain bridge events, and off-chain evidence are often stored in different systems or tables, making it difficult to express the temporal evolution relationship of the same risk event at different levels. Third, queries lack reusable states. Queries such as fund tracking and identity association require re-expanding paths or re-matching patterns each time, resulting in high query latency and insufficient interpretation of evidence for complex paths. Summary of the Invention
[0004] This application provides a method, apparatus, device, and medium for incremental updating and querying of multi-level time series graphs, which can solve the problems of data fragmentation, high update cost, and low query efficiency, and realize real-time updates and fast queries.
[0005] In a first aspect, this application provides an incremental update and query method for a multi-level time series graph, comprising: acquiring multi-source heterogeneous data from different levels, uniformly converting the multi-source heterogeneous data into standard event tuples, and mapping the standard event tuples into a multi-level time series graph composed of multiple types of vertices and multiple types of temporal edges; when a batch of new standard event tuples arrives, determining the local influence domain of the batch of new standard event tuples in the multi-level time series graph based on the temporal proximity, feature similarity, and topological correlation between each new standard event tuple in the batch and historical standard event tuples, and performing incremental updates only within the local influence domain on a pre-maintained multi-level intermediate state set and a risk path index; responding to a risk query request, matching the risk path index according to the risk query request to obtain candidate paths and candidate subgraphs, calling the multi-level intermediate state set to execute query operators based on the candidate paths in the candidate subgraphs, and generating and outputting query results containing risk paths and explanatory information.
[0006] In one possible implementation, the step of uniformly converting the multi-source heterogeneous data into standard event tuples and mapping the standard event tuples into a multi-level temporal graph composed of multiple types of vertices and multiple types of temporal edges specifically includes: normalizing and encapsulating the multi-source heterogeneous data into unified standard event tuples, wherein the standard event tuples specifically include: chain identifier, time information, source object, target object, event type, event attribute, and risk information; generating or associating corresponding multiple types of vertices in the multi-level temporal graph based on the source object, the target object, and the event attribute in the standard event tuples; constructing temporal edges of the corresponding type between the vertices corresponding to the source object and the vertices corresponding to the target object based on the event type in the standard event tuples, and attaching at least some fields of the time information, the event attribute, and the risk information as attributes of the temporal edges.
[0007] In one possible implementation, determining the local influence domain of the batch of newly added standard event tuples in the multi-level time series graph based on the temporal proximity, feature similarity, and topological association between each newly added standard event tuple and historical standard event tuples in the batch specifically includes: calculating the temporal proximity between the newly added standard event tuples and historical standard event tuples based on the timestamp difference between the newly added standard event tuples and historical standard event tuples in the batch; and extracting the temporal proximity between the newly added standard event tuples and historical standard event tuples in the batch. Based on the matching degree of the event attributes, the feature similarity between the newly added standard event tuple and the historical standard event tuple is calculated; based on the graph connectivity in the multi-level time sequence graph, the topological correlation between the newly added standard event tuple and the historical standard event tuple in the batch of newly added standard event tuples is calculated; a comprehensive correlation score is calculated based on the temporal proximity, the feature similarity, and the topological correlation; the vertices and temporal edges involved in the newly added standard event tuple, as well as the vertices and temporal edges involved in the historical standard event tuple whose comprehensive correlation score is higher than a preset threshold, are included in the local influence domain.
[0008] In one possible implementation, the step of matching the risk path index with the risk query request to obtain candidate paths and candidate subgraphs specifically includes: extracting query conditions from the risk query request, wherein the query conditions include at least one of chain identifier, time window, start and end objects, risk type, path length range, and confidence range; matching and retrieving the query conditions with the risk path index, obtaining initial risk paths that satisfy the query conditions based on the matching results, and using them as candidate paths; extracting the vertex set involved in the candidate path, and obtaining the associated edges between the vertex sets, and constructing the candidate subgraph together with the vertex sets and the associated edges.
[0009] In one possible implementation, the multi-level intermediate state set includes node intermediate states, edge intermediate states, subgraph intermediate states, entity intermediate states, and risk path intermediate states. The step of executing a query operator based on the candidate path in the candidate subgraph, calling the multi-level intermediate state set, to generate and output query results containing risk paths and explanatory information, specifically includes: parsing the query type of the risk query request; scheduling the corresponding target query operator in the candidate subgraph according to the query type; reading corresponding state data from the multi-level intermediate state set based on the candidate path using the target query operator; during the execution of the target query operator, if it is determined that the state data cannot cover the query conditions of the risk query request or the version is too old, triggering supplementary calculation of the underlying data in the candidate subgraph; and generating and outputting query results containing risk paths, key evidence edges, and explanatory labels based on the state data or the result of the supplementary calculation.
[0010] In one possible implementation, the target query operator includes at least one of a path expansion operator, a pattern matching operator, an entity merging operator, a time filtering operator, a key evidence edge verification operator, or a version status verification operator; the step of scheduling the corresponding target query operator in the candidate subgraph according to the query type specifically includes: scheduling the path expansion operator when the query type is a funds tracking query; scheduling the entity merging operator when the query type is an anonymous identity association query; scheduling the pattern matching operator when the query type is a contract anomaly query; and scheduling the time filtering operator, as well as the key evidence edge verification operator or the version status verification operator, when the query type is an evidence chain verification query.
[0011] In one possible implementation, the incremental update and query method for a multi-level time sequence graph provided in this application further includes: receiving feedback results generated by external verification or handling; converting the feedback results into feedback event tuples; mapping the feedback event tuples to off-chain evidence edges, entity association edges, or risk event vertices; using the feedback event tuples as newly added standard event tuples; redefining the local influence domain; and correcting the multi-level intermediate state set and related risk thresholds within the local influence domain. The updated multi-level intermediate state set and risk path index have new version numbers, so that the state of the multi-level time sequence graph evolves with the feedback results.
[0012] Secondly, this application provides an incremental update and query device for a multi-level time series graph, comprising: a standard event tuple conversion module, a local influence domain determination module, and a query result generation module; wherein, the standard event tuple conversion module is used to acquire multi-source heterogeneous data from different levels, uniformly convert the multi-source heterogeneous data into standard event tuples, and map the standard event tuples into a multi-level time series graph composed of multiple types of vertices and multiple types of time series edges; the local influence domain determination module is used to, when a batch of new standard event tuples arrives, determine the local influence domain based on the relationship between each new standard event tuple in the batch and historical standard event tuples. The temporal proximity, feature similarity, and topological correlation among quasi-event tuples are used to determine the local influence domain of the newly added standard event tuple batch in the multi-level time series graph, and incremental updates are performed only within the local influence domain on the pre-maintained multi-level intermediate state set and risk path index; the query result generation module is used to respond to risk query requests, match the risk path index according to the risk query request to obtain candidate paths and candidate subgraphs, and call the multi-level intermediate state set to execute query operators based on the candidate paths in the candidate subgraphs to generate and output query results containing risk paths and explanatory information.
[0013] Thirdly, embodiments of this application also provide a computer device, which includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described method.
[0014] Fourthly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the above-described method.
[0015] This application provides a method, apparatus, device, and medium for incremental updating and querying of multi-level time series graphs, which has the following advantages compared with the prior art: The technical solution of this application solves the problem of data fragmentation by uniformly converting multi-source heterogeneous data into standard event tuples and mapping them to a multi-level time series graph, enabling data scattered across different systems to be uniformly represented in the same graph model. Secondly, when a batch of new standard event tuples arrives, a local influence domain is defined based on the temporal proximity, feature similarity, and topological correlation between each new standard event tuple and historical standard event tuples in the batch, and incremental updates are performed only within the local influence domain. This allows the batch of new standard event tuples to complete state maintenance without triggering a full recalculation, solving the problem of high update costs. Thirdly, through a pre-maintained set of intermediate states and path indexes, queries are performed by matching the index and calling the intermediate states, allowing queries to quickly return results with explanatory information without retracing the entire graph, solving the problem of low query efficiency. The technical solution of this application simultaneously solves the problems of data fragmentation, high update costs, and low query efficiency, achieving real-time updates and fast queries. Attached Figure Description
[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.
[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0018] One or more embodiments are illustrated by way of example with reference numerals in the accompanying drawings. These illustrations do not constitute a limitation on the embodiments. Elements with the same reference numerals in the drawings are denoted as similar elements. Unless otherwise stated, the figures in the drawings are not to be limited by scale.
[0019] Figure 1 This is a flowchart illustrating an embodiment of the incremental update and query method for a multi-level time sequence diagram provided in this application; Figure 2 This is a schematic diagram of an embodiment of the incremental update and query device for a multi-level time sequence diagram provided in this application; Figure 3 This is a schematic diagram of the structure of a computer device provided in this application. Detailed Implementation
[0020] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0021] The following disclosure provides numerous different embodiments or examples for implementing various structures of this application. To simplify the disclosure, specific examples of components and arrangements are described below. These are merely examples and are not intended to limit the scope of this application. Furthermore, reference numerals and / or letters may be repeated in different examples. Such repetition is for simplification and clarity and does not in itself indicate a relationship between the various embodiments and / or arrangements discussed.
[0022] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.
[0023] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.
[0024] It should also be further understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0025] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."
[0026] Example 1, see Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the incremental update and query method for multi-level time sequence graphs provided in this application, as shown below. Figure 1 As shown, the method includes steps 101-103, as detailed below: Step 101: Obtain multi-source heterogeneous data from different levels, convert the multi-source heterogeneous data into standard event tuples, and map the standard event tuples into a multi-level time sequence graph composed of multiple types of vertices and multiple types of temporal edges.
[0027] In one embodiment, the system first obtains multi-source heterogeneous data from multiple layers of the blockchain risk prevention and control scenario; wherein, the multiple layers include the storage and execution layer, the network and consensus layer, the smart contract layer, and the ecosystem application layer.
[0028] In one embodiment, the multi-source heterogeneous data is normalized and encapsulated into a unified standard event tuple. The standard event tuple specifically includes: chain identifier, time information, source object, target object, event type, event attribute, and risk information. Based on the source object, target object, and event attribute in the standard event tuple, corresponding multi-type vertices are generated or associated in the multi-level time sequence graph. Based on the event type in the standard event tuple, a time sequence edge of the corresponding type is constructed between the vertex corresponding to the source object and the vertex corresponding to the target object, and at least some fields of the time information, event attribute, and risk information are attached as attributes of the time sequence edge.
[0029] Specifically, the standard event tuple adopts the following format: e =<cid,t,src,dst,type,attr,risk> Wherein, cid is the chain identifier or network identifier, used to distinguish between public chains, consortium chains, sidechains, and cross-chain target chains, and can be extracted from block headers, node configurations, cross-chain bridge events, or consortium chain channel information; t is the event occurrence time or the time of system confirmation, which can include block time, transaction confirmation time, log time, and verification feedback time, and can be obtained from block timestamps, transaction receipts, node logs, and business processing records; src is the source object, representing the transaction initiating address, calling contract, propagation source node, or off-chain evidence source, and can be obtained from transaction inputs, contract callers, node IDs, or business system records; dst is the target object, representing the receiving address, called contract, propagation target node, associated entity, or evidence target, and can be obtained from transaction outputs, contract callers, node IDs ... The event type is obtained from the contract address, node ID, or KYC matching object; `type` is the event type, used to distinguish between transfers, contract calls, co-occurrence within the same block, entity associations, cross-chain mappings, network propagation, consensus anomalies, and off-chain evidence, which can be obtained from block parsing results, contract logs, network monitoring, or off-chain verification; `attr` is the event attribute set, recording information such as amount, function selector, contract address, node role, transaction hash, blacklist label, and evidence number, which can be obtained from transaction fields, event logs, risk control label library, and case or handling system; `risk` is the risk information, recording risk type, risk score, rule hit items, model output, or manual confirmation results, which can be obtained from risk rules, model scores, blacklists and graylists, expert review, and verification by public security or financial institutions. Through the above standardized encapsulation, multi-source heterogeneous data from different sources and in different formats are converted into a unified structured event tuple, which serves as a standard input for subsequent time series graph construction, incremental updates, and risk queries.
[0030] Specifically, the source object `src`, target object `dst`, and key objects appearing in the event attribute set `attr` of the standard event tuple are mapped to multiple types of vertices. These multiple types of vertices include address vertices, transaction vertices, contract vertices, asset vertices, block vertices, node vertices, entity vertices, and risk event vertices. For example, a normal transfer event can map the transaction initiating address and receiving address to address vertices, the transaction hash to transaction vertices, and the asset corresponding to the transfer amount to asset vertices. A contract call event can map the caller address to address vertices, the called contract address to contract vertices, and the transaction hash to transaction vertices. An off-chain verification event can map the entities involved in the verification to entity vertices and the risk event corresponding to the verification result to risk event vertices.
[0031] Specifically, for vertices that already exist in the graph, the system directly associates them with the existing vertices; for vertices that have not yet appeared in the graph, the system adds corresponding vertices to the graph.
[0032] Specifically, the event type determines the edge type; for example, a transfer event is mapped to a transfer edge, a contract log event is mapped to a call edge, highly correlated transactions within the same block or the same time window are mapped to co-occurrence edges, address clustering or KYC-assisted matching are mapped to entity association edges, cross-chain bridge locking and releasing events are mapped to cross-chain mapping edges, inter-node propagation or abnormal communication is mapped to network propagation edges, consensus anomalies are mapped to consensus anomaly edges, and expert review, public security verification, or financial institution handling records are mapped to off-chain evidence edges.
[0033] Specifically, while constructing time-series edges, the system attaches the time information in the event tuple as the time attribute of the time-series edge, and attaches at least some fields from the event type, event attributes, and risk information as the state attributes of the time-series edge. For example, each time-series edge stores time attributes such as occurrence time and effective start time, as well as state attributes such as effective status, invalid status, invalidation time, confidence level, risk score, data source, version number, and evidence citation. Among them, the version number is used to distinguish different states of the same relation after multiple incremental updates or manual feedback, and the confidence level is used to indicate the reliability of model judgment, rule hit, or manual confirmation. Through the above mapping process, multi-source heterogeneous data scattered in different systems are integrated into a unified time-series graph structure, forming the graph data foundation for subsequent incremental updates and risk queries.
[0034] Step 102: When a batch of new standard event tuples arrives, based on the temporal proximity, feature similarity, and topological correlation between each new standard event tuple in the batch and the historical standard event tuples, determine the local influence domain of the batch of new standard event tuples in the multi-level time sequence diagram, and perform incremental updates only on the pre-maintained multi-level intermediate state set and risk path index within the local influence domain.
[0035] In one embodiment, the local influence domain of the newly added standard event tuple batch in the multi-level time series graph is determined based on the temporal proximity, feature similarity, and topological association between each newly added standard event tuple and the historical standard event tuple in the newly added standard event tuple batch. Specifically, this includes: calculating the temporal proximity between the newly added standard event tuple and the historical standard event tuple based on the timestamp difference between the newly added standard event tuple and the historical standard event tuple in the newly added standard event tuple batch; and extracting the event attributes of the newly added standard event tuple and the historical standard event tuple in the newly added standard event tuple batch. Based on the matching degree of the event attributes, the feature similarity between the newly added standard event tuple and the historical standard event tuple is calculated; based on the graph connectivity in the multi-level time sequence graph, the topological association between the newly added standard event tuple and the historical standard event tuple in the batch of newly added standard event tuples is calculated; a comprehensive correlation score is calculated based on the temporal proximity, the feature similarity, and the topological association; vertices and temporal edges involved in the newly added standard event tuple, as well as vertices and temporal edges involved in the historical standard event tuple whose comprehensive correlation score is higher than a preset threshold, are included in the local influence domain.
[0036] Specifically, when a batch of new standard event tuples arrives, the system does not directly put all events into the whole graph for recalculation. Instead, it first locates the vertices and edges directly involved by each new standard event tuple in the batch, including source object src, target object dst, transaction object, contract object, asset object, node object, entity object, and risk event object, forming an initial set Ω0, and calculates the correlation between each new standard event tuple in the batch and the historical standard event tuples.
[0037] Specifically, the system calculates the temporal proximity T between the newly added standard event tuple and the historical standard event tuple. Temporal proximity measures the closeness of two events in time; a typical implementation uses a sliding time window, for example, when the difference between the timestamps of two events is less than a preset window W, the temporal proximity is high, and the larger the time difference, the lower the temporal proximity; alternatively, the exponential decay function T = exp(-|t1-t2| / τ) can be used for calculation, where τ is the time decay parameter. The system calculates the temporal proximity between the newly added standard event tuple and the historical standard event tuple based on the difference between the timestamp t1 of the newly added standard event tuple and the timestamp t2 of the historical standard event tuple.
[0038] Specifically, the system calculates the feature similarity S between the newly added standard event tuple and the historical standard event tuple; this feature similarity is used to measure whether event attributes are similar. The system extracts feature vectors from the event attribute set `attr` corresponding to each of the newly added and historical standard event tuples. These features include amount range, function selector, contract type, node role, risk label, transaction frequency, and number of inputs and outputs. The system uses cosine similarity, Jaccard similarity, or rule matching scores to calculate the feature similarity between the newly added and historical standard event tuples.
[0039] Specifically, the system calculates the topological correlation R between the newly added standard event tuple and the historical standard event tuple. This topological correlation measures the degree of structural association between events in the current time sequence graph. The system calculates this based on the graph connectivity in the multi-level time sequence graph, specifically including: determining whether two events share addresses, contracts, transactions, entities, or cross-chain bridges; determining whether the relevant vertices of the two events are connected within a k-hop range; and counting the number of common nodes, common entities, or common risk tags on the association paths of the two events.
[0040] Specifically, after calculating the temporal proximity T, feature similarity S, and topological correlation R respectively, the system calculates a comprehensive correlation score based on these three factors. The comprehensive correlation score is calculated using a weighted sum: ,in, These are weighting coefficients that can be configured according to the business scenario. For example, real-time money laundering fund flow identification can increase the weight of temporal proximity and topological correlation, while contract bulk attack identification can increase the weight of feature similarity.
[0041] Specifically, when the overall relevance score is higher than a preset threshold θ, the system classifies the vertices, temporal edges, and associated structures involved in the corresponding historical standard event tuple into the local influence domain; when a newly added standard event tuple is related to multiple subgraphs simultaneously, the system assigns it to the main subgraph based on the highest overall relevance score, and retains the reference relationship in other related subgraphs. The preset threshold θ can be dynamically adjusted according to throughput, false alarm rate, subgraph size limit, and query latency requirements.
[0042] Specifically, the system uses the newly added event time t as the center and searches for historical events and pending events related to vertices or edges in the initial set Ω0 within the [t-Wpre, t+Wpost] window, adding them to the local influence domain. Here, Wpre is the size of the forward window, and Wpost is the size of the backward window. If an address in the initial set Ω0 already belongs to an entity, or if the newly added event triggers entity associations such as KYC, blacklists, or off-chain evidence, then the associated addresses, related contracts, and historical risk events under the same entity are added to the local influence domain. Starting from a vertex in the initial set Ω0, the system expands k hops along edge types such as transfer edges, call edges, cross-chain mapping edges, and entity association edges in the current time series graph, filtering by edge type, time range, and confidence level to obtain the final local influence domain Ω. The system only performs subsequent incremental update operations within the final local influence domain Ω; vertices, edges, intermediate states, and indexes outside the local influence domain Ω remain unchanged from their original versions.
[0043] In one embodiment, to avoid repeatedly calculating from the original transactions and logs during queries, the system maintains five types of intermediate states during incremental updates. These intermediate states are used both to receive new events and for direct reuse during real-time queries.
[0044] In one embodiment, the multi-level intermediate state set includes five types of intermediate states: node intermediate state Mv, edge intermediate state Me, subgraph intermediate state Mg, entity intermediate state Mi, and risk path intermediate state Mp.
[0045] Specifically, the node intermediate state Mv includes vertex ID, vertex type, recent active time, inflow and outflow amount statistics, transaction frequency, contract call frequency, black and gray list tags, time decay risk score, and version number. It is used to quickly determine the current risk profile of an address, contract, node, or entity, avoiding the need to re-aggregate the node's historical behavior during queries.
[0046] Specifically, the edge intermediate state Me includes edge ID, edge type, source vertex, target vertex, cumulative amount, recent event time, valid state, confidence level, risk weight, evidence source, and version number. It is used to quickly obtain the current state of fund flow, call flow, propagation flow, or evidence relationship, and supports path expansion and evidence chain generation.
[0047] Specifically, the subgraph intermediate state Mg contains the subgraph ID, time window, member vertex summary, member edge summary, splitting and aggregation mode, circular transfer mode, batch call mode, cross-chain jump mode, and local risk score, which are used to store the results of local risk patterns. When querying, suspicious patterns can be directly hit without having to do full pattern matching again.
[0048] Specifically, the entity intermediate state Mi includes entity ID, associated address set, KYC auxiliary label, off-chain identity clues, entity confidence level, historical risk events, feedback status, and version number, which are used to support anonymous identity association and address clustering queries, and can continuously integrate on-chain addresses and off-chain verification results.
[0049] Specifically, the intermediate state Mp of the risk path includes path ID, start point, end point, path edge sequence, key evidence edge, cross-chain node, cumulative risk score, minimum confidence level, time span, explanation label and version number, which are used to directly support fund tracking, cross-chain path reconstruction and evidence chain verification, significantly shortening the query latency of complex paths.
[0050] Specifically, based on each newly added standard event tuple in the batch of newly added standard event tuples, the system recalculates the affected node intermediate states Mv, edge intermediate states Me, subgraph intermediate states Mg, entity intermediate states Mi, and risk path intermediate states Mp within the subgraph defined by the local influence domain Ω. For example, when a new transfer event falls within the local influence domain Ω, the system updates the inflow and outflow amount statistics and transaction frequency in the node intermediate state Mv of the relevant address vertex, updates the cumulative amount and recent event time in the edge intermediate state Me of the relevant transfer edge, updates the member vertex summary and member edge summary in the subgraph intermediate state Mg of the subgraph to which it belongs, and if the address is associated with an entity, it synchronously updates the associated address set and historical risk events in the entity intermediate state Mi, and updates the path edge sequence and cumulative risk score in the involved risk path intermediate state Mp. At the same time, the system performs incremental updates to the risk path index within the local influence domain Ω, recalculating and updating the corresponding index values in the risk path index based on the updated intermediate states, including candidate paths, key evidence edges, cumulative risk scores, and version numbers.
[0051] In one embodiment, the system performs the recalculation operation only within the local influence domain Ω. Vertices, edges, intermediate states, and indexes outside the local influence domain Ω retain their original versions. By restricting the update calculation to the local influence domain, for high-frequency on-chain transaction flows, most new events only affect a limited number of addresses, contracts, entities, and paths. This reduces redundant calculations, improves real-time update capabilities, and prevents existing stable results from being disturbed by irrelevant events. The updated intermediate states and risk path indexes carry new version numbers; subsequent queries will read the latest version and can retrospectively reflect the state changes before and after.
[0052] Step 103: In response to the risk query request, the risk path index is matched according to the risk query request to obtain candidate paths and candidate subgraphs. Based on the candidate paths in the candidate subgraphs, the multi-level intermediate state set is called to execute the query operator, and query results containing risk paths and explanatory information are generated and output.
[0053] In one embodiment, obtaining candidate paths and candidate subgraphs by matching the risk path index with the risk query request specifically includes: extracting query conditions from the risk query request, wherein the query conditions include at least one of chain identifier, time window, start and end objects, risk type, path length range, and confidence range; matching and retrieving the query conditions with the risk path index, obtaining initial risk paths that satisfy the query conditions based on the matching results, and using them as candidate paths; extracting the vertex set involved in the candidate path, and obtaining the associated edges between the vertex sets, and constructing the candidate subgraph together with the vertex sets and the associated edges.
[0054] Specifically, the system pre-builds and maintains a risk path index. The risk path index consists of an index key and an index value; it is used to transform complex risk queries from full graph search to candidate subgraph selection + local operator execution.
[0055] Specifically, the index key includes the chain identifier (cid), time window, start and end objects, risk type, path length range, and confidence level range. The chain identifier distinguishes between single-chain, cross-chain, and consortium chain channels; the time window limits the time frame within which the path occurs or becomes effective; start and end objects include the starting address, ending address, entity, contract, or risk event; risk types include money laundering, fraud, abnormal contract calls, cross-chain transfers, and consensus anomalies; the path length range limits whether the path is 1-hop, k-hop, or a multi-segment cross-chain path; and the confidence level range limits the credibility of the model, rules, or human verification results.
[0056] Specifically, the index values include candidate paths, key evidence edges, cumulative risk scores, and version numbers. Among them, candidate paths store path summaries composed of vertices and edges; key evidence edges store transfer edges, call edges, cross-chain mapping edges, or off-chain evidence edges that play a key role in risk assessment; cumulative risk scores store the comprehensive results of the risks of nodes, edges, subgraphs, and entities on the path; and the version number records the version of the path index that was generated or updated, facilitating the tracking of changes after feedback.
[0057] Specifically, the system matches the extracted query conditions with the index key of the risk path index, that is, it matches the risk path index based on conditions such as chain identifier, time window, risk type, start and end objects, and obtains the initial risk path that meets the query conditions from the index value as a candidate path.
[0058] Specifically, the candidate path itself already contains a path summary composed of vertices and edges. The system extracts a set of all vertices and a set of all edges connecting these vertices from this path summary. The candidate subgraph is constructed by combining the vertex set and the associated edges. In this way, the system transforms complex risk queries from full graph search to candidate subgraph selection plus local operator execution, significantly reducing the computational scope required for the query.
[0059] In one embodiment, the step of calling the multi-level intermediate state set to execute a query operator based on the candidate path in the candidate subgraph to generate and output query results containing risk paths and explanatory information specifically includes: parsing the query type of the risk query request; scheduling the corresponding target query operator in the candidate subgraph according to the query type; reading the corresponding state data from the multi-level intermediate state set based on the candidate path using the target query operator; during the execution of the target query operator, when it is determined that the state data cannot cover the query conditions of the risk query request or the version is too old, triggering supplementary calculation of the underlying data in the candidate subgraph; and generating and outputting query results containing risk paths, key evidence edges, and explanatory labels based on the state data or the result of the supplementary calculation.
[0060] In one embodiment, the target query operator includes at least one of a path expansion operator, a pattern matching operator, an entity merging operator, a time filtering operator, a key evidence side verification operator, or a version status verification operator.
[0061] In one embodiment, scheduling the corresponding target query operator in the candidate subgraph according to the query type specifically includes: scheduling the path expansion operator when the query type is a fund tracking query; scheduling the entity merging operator when the query type is an anonymous identity association query; scheduling the pattern matching operator when the query type is a contract anomaly query; and scheduling the time filtering operator, as well as the key evidence edge verification operator or version status verification operator when the query type is an evidence chain verification query.
[0062] Specifically, after scheduling the corresponding target query operator, the system reads the corresponding state data from the multi-level intermediate state set based on the candidate path and using the target query operator. During the query, the system prioritizes reading the node intermediate state Mv, edge intermediate state Me, subgraph intermediate state Mg, entity intermediate state Mi, and risk path intermediate state Mp that match the query conditions. For example, a fund tracking query first reads the risk path intermediate state Mp and edge intermediate state Me; an identity association query first reads the entity intermediate state Mi; and a contract anomaly query first reads the node intermediate state Mv and subgraph intermediate state Mg of the contract vertex.
[0063] Specifically, during the execution of the target query operator, when the system determines that the status data cannot cover the query conditions of the risk query request or the version is too old, the system triggers supplementary calculation of the underlying data in the candidate subgraph; that is, the system only falls back to the candidate subgraph to perform supplementary calculation when the intermediate state cannot cover the query conditions or the version is too old.
[0064] Specifically, when the system generates and outputs query results containing risk paths, key evidence edges, and explanation labels based on the state data or the results of the supplementary calculations, since the intermediate state Mp of the risk path contains path ID, start point, end point, path edge sequence, key evidence edge, cross-chain node, cumulative risk score, minimum confidence level, time span, explanation label, and version number, the system directly reads the above information and outputs it, so that the query results not only contain the risk path itself, but also simultaneously provide key evidence edges and explanation labels, which facilitates subsequent auditing, analysis, and evidence solidification.
[0065] In one embodiment, feedback results generated by external verification or handling are received, the feedback results are converted into feedback event tuples, and the feedback event tuples are mapped to off-chain evidence edges, entity association edges, or risk event vertices; the feedback event tuples are used as new standard event tuples to redetermine the local influence domain, and the multi-level intermediate state set and related risk thresholds are corrected within the local influence domain. The updated multi-level intermediate state set and risk path index have new version numbers, so that the state of the multi-level sequence diagram evolves with the feedback results.
[0066] Specifically, the system treats external feedback such as expert review, financial institution processing, public security verification, and user confirmation results as new graph events, rather than simply marking them as "verified" on the interface.
[0067] Specifically, the system receives feedback results from external verification or processing, and converts the feedback results into feedback event tuples; the feedback event tuples record information such as feedback source, feedback time, associated address / contract / entity / path, confirmation result, evidence number, and confidence level.
[0068] Specifically, after standardizing the feedback event tuples into standard event tuples, they are mapped to off-chain evidence edges, entity association edges, or risk event vertices. For example, when experts confirm that an address belongs to an entity involved in a case, the system maps the feedback to an entity association edge, associating the address with the corresponding entity vertex; when the public security bureau confirms that a transaction is risky, the system maps the feedback to an off-chain evidence edge, attaching it to the timeline path of the transaction; when a financial institution's handling results form a new risk event, the system maps it to a risk event vertex.
[0069] Specifically, the system uses the feedback event tuple as a new standard event tuple to redetermine the local influence domain. This local influence domain typically includes the verified address, the entity to which it belongs, related risk paths, similar risk subgraphs, and associated evidence edges. After determining the local influence domain, the system corrects the multi-level intermediate state set and related risk thresholds within the local influence domain.
[0070] Specifically, the system modifies the entity intermediate state Mi, risk path intermediate state Mp, node intermediate state Mv, and related risk thresholds within the redefined local influence domain. For example, after confirming that an address belongs to the entity involved in the case, the system updates the set of associated addresses and historical risk events in the entity intermediate state Mi corresponding to that entity, and simultaneously adjusts the risk scores and related path rankings of other addresses under that entity.
[0071] Specifically, the updated multi-level intermediate state set and risk path index have new version numbers. Subsequent queries will read the latest version and can trace back the changes in risk paths before and after feedback. Through the above closed-loop feedback mechanism, the system can continuously inject real handling results into the underlying graph state, so that the state of the multi-level time sequence graph evolves with the feedback results. This allows the risk identification and tracking capabilities to continuously evolve with case analysis and business handling, rather than remaining in a static rules or static tag library.
[0072] Example 2, see Figure 2 , Figure 2 This is a schematic diagram of an embodiment of an incremental update and query device for multi-level time series graphs provided in this application. Corresponding to the above-described incremental update and query method for multi-level time series graphs, this application also provides an incremental update and query device for multi-level time series graphs. This device includes modules for executing the above-described incremental update and query method for multi-level time series graphs. The device can be deployed in servers, cloud platforms, computing clusters, edge computing devices, or electronic devices with computing and storage capabilities. Specifically, the device includes a standard event tuple conversion module 201, a local influence domain determination module 202, and a query result generation module 203.
[0073] The standard event tuple conversion module 201 is used to acquire multi-source heterogeneous data from different levels, convert the multi-source heterogeneous data into standard event tuples, and map the standard event tuples into a multi-level time sequence graph composed of multi-type vertices and multi-type temporal edges.
[0074] The local influence domain determination module 202 is used to determine the local influence domain of the new standard event tuple batch in the multi-level time sequence diagram based on the temporal proximity, feature similarity and topological correlation between each new standard event tuple in the new standard event tuple batch and the historical standard event tuple when the new standard event tuple batch arrives, and to perform incremental updates on the pre-maintained multi-level intermediate state set and risk path index only within the local influence domain.
[0075] The query result generation module 203 is used to respond to a risk query request, match the risk path index according to the risk query request to obtain candidate paths and candidate subgraphs, call the multi-level intermediate state set to execute query operators based on the candidate paths in the candidate subgraphs, and generate and output query results containing risk paths and explanatory information.
[0076] In one embodiment, the standard event tuple conversion module 201 is used to uniformly convert the multi-source heterogeneous data into standard event tuples and map the standard event tuples into a multi-level time series graph composed of multiple types of vertices and multiple types of temporal edges. Specifically, it includes: normalizing and encapsulating the multi-source heterogeneous data into a unified standard event tuple, wherein the standard event tuple specifically includes: chain identifier, time information, source object, target object, event type, event attribute, and risk information; generating or associating corresponding multiple types of vertices in the multi-level time series graph based on the source object, the target object, and the event attribute in the standard event tuple; constructing a temporal edge of the corresponding type between the vertex corresponding to the source object and the vertex corresponding to the target object based on the event type in the standard event tuple, and attaching at least some fields of the time information, the event attribute, and the risk information as attributes of the temporal edge.
[0077] In one embodiment, the local influence domain determination module 202 is used to determine the local influence domain of the new standard event tuple batch in the multi-level time series graph based on the temporal proximity, feature similarity, and topological correlation between each new standard event tuple and historical standard event tuple in the new standard event tuple batch. Specifically, this includes: calculating the temporal proximity between the new standard event tuple and historical standard event tuple based on the timestamp difference between the new standard event tuple and historical standard event tuple in the new standard event tuple batch; and extracting the temporal proximity between the new standard event tuple and historical standard event tuple in the new standard event tuple batch. Based on the event attributes of the event tuples, the feature similarity between the newly added standard event tuples and historical standard event tuples is calculated; based on the graph connectivity in the multi-level time sequence graph, the topological association between the newly added standard event tuples and historical standard event tuples in the batch of newly added standard event tuples is calculated; a comprehensive correlation score is calculated based on the temporal proximity, the feature similarity, and the topological association; vertices and temporal edges involved in the newly added standard event tuples, as well as vertices and temporal edges involved in historical standard event tuples whose comprehensive correlation score is higher than a preset threshold, are included in the local influence domain.
[0078] In one embodiment, the query result generation module 203 is used to obtain candidate paths and candidate subgraphs by matching the risk path index with the risk query request. Specifically, it includes: extracting query conditions from the risk query request, wherein the query conditions include at least one of chain identifier, time window, start and end objects, risk type, path length range, and confidence range; matching and retrieving the query conditions with the risk path index, obtaining an initial risk path that satisfies the query conditions based on the matching result, and using it as the candidate path; extracting the vertex set involved in the candidate path, and obtaining the associated edges between the vertex sets, and constructing the candidate subgraph together with the vertex sets and the associated edges.
[0079] In one embodiment, the multi-level intermediate state set includes node intermediate states, edge intermediate states, subgraph intermediate states, entity intermediate states, and risk path intermediate states.
[0080] In one embodiment, the query result generation module 203 is used to call the multi-level intermediate state set to execute a query operator based on the candidate path in the candidate subgraph, and generate and output query results containing risk paths and explanation information. Specifically, this includes: parsing the query type of the risk query request; scheduling the corresponding target query operator in the candidate subgraph according to the query type; reading the corresponding state data from the multi-level intermediate state set based on the candidate path using the target query operator; during the execution of the target query operator, when it is determined that the state data cannot cover the query conditions of the risk query request or the version is too old, triggering supplementary calculation of the underlying data in the candidate subgraph; and generating and outputting query results containing risk paths, key evidence edges, and explanation labels based on the state data or the result of the supplementary calculation.
[0081] In one embodiment, the target query operator includes at least one of a path expansion operator, a pattern matching operator, an entity merging operator, a time filtering operator, a key evidence side verification operator, or a version status verification operator.
[0082] In one embodiment, the query result generation module 203 is used to schedule the corresponding target query operator in the candidate subgraph according to the query type, specifically including: scheduling the path expansion operator when the query type is a fund tracking query; scheduling the entity merging operator when the query type is an anonymous identity association query; scheduling the pattern matching operator when the query type is a contract anomaly query; and scheduling the time filtering operator, as well as the key evidence edge verification operator or version status verification operator when the query type is an evidence chain verification query.
[0083] In one embodiment, the incremental update and query device for a multi-level time sequence graph provided in this application further includes: a multi-level time sequence graph state update module; wherein, the multi-level time sequence graph state update module is used to receive feedback results generated by external verification or handling, convert the feedback results into feedback event tuples, and map the feedback event tuples as off-chain evidence edges, entity association edges, or risk event vertices; using the feedback event tuples as newly added standard event tuples, redetermine the local influence domain, and correct the multi-level intermediate state set and related risk thresholds within the local influence domain; the updated multi-level intermediate state set and risk path index have new version numbers, so that the state of the multi-level time sequence graph evolves with the feedback results.
[0084] The above-described incremental update and query apparatus for multi-level time series diagrams can implement the incremental update and query method for multi-level time series diagrams described in the above method embodiments. The options described in the above method embodiments are also applicable to this embodiment, and will not be detailed here.
[0085] like Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a computer device provided in this application; it includes a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112 and the memory 113 communicate with each other through the communication bus 114, and the memory 113 is used to store computer programs.
[0086] In one embodiment of this application, when the processor 111 executes the program stored in the memory 113, it implements the incremental update and query method of the multi-level timing diagram provided in any of the foregoing method embodiments.
[0087] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program may be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.
[0088] Therefore, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, wherein when the computer program is executed by a processor, it implements the steps of the incremental update and query method for multi-level timing diagrams as provided in any of the foregoing method embodiments.
[0089] The storage medium is a physical, non-transient storage medium, such as a USB flash drive, external hard drive, read-only memory (ROM), magnetic disk, or optical disk, or any other physical storage medium capable of storing program code. The computer-readable storage medium can be non-volatile or volatile.
[0090] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this application.
[0091] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is merely a logical functional division, and there may be other division methods in actual implementation. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.
[0092] The steps in the methods of this application embodiment can be adjusted, merged, or deleted according to actual needs. The units in the apparatus of this application embodiment can be merged, divided, or deleted according to actual needs. Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.
[0093] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.
[0094] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.
[0095] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Since these modifications and variations fall within the scope of the claims and their equivalents, this application also intends to include these modifications and variations.
[0096] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
Claims
1. A method for incremental update and query of multi-level timing diagrams, characterized in that, include: Acquire multi-source heterogeneous data from different levels, convert the multi-source heterogeneous data into standard event tuples, and map the standard event tuples into a multi-level time sequence graph composed of multi-type vertices and multi-type temporal edges; When a batch of new standard event tuples arrives, the local influence domain of the batch of new standard event tuples in the multi-level time sequence diagram is determined based on the temporal proximity, feature similarity, and topological correlation between each new standard event tuple in the batch and the historical standard event tuples. Incremental updates are then performed only within the local influence domain on the pre-maintained multi-level intermediate state set and risk path index. In response to a risk query request, the risk path index is matched according to the risk query request to obtain candidate paths and candidate subgraphs. Based on the candidate paths in the candidate subgraphs, the multi-level intermediate state set is called to execute query operators, and query results containing risk paths and explanatory information are generated and output.
2. The method of claim 1, wherein, The step of uniformly converting the multi-source heterogeneous data into standard event tuples and mapping the standard event tuples into a multi-level time series graph composed of multiple types of vertices and multiple types of temporal edges specifically includes: The multi-source heterogeneous data is normalized and encapsulated into a unified standard event tuple, which specifically includes: chain identifier, time information, source object, target object, event type, event attribute, and risk information; Based on the source object, the target object, and the event attributes in the standard event tuple, generate or associate corresponding multi-type vertices in the multi-level sequence graph; Based on the event type in the standard event tuple, a time-series edge of the corresponding type is constructed between the vertex corresponding to the source object and the vertex corresponding to the target object, and at least some fields of the time information, the event attributes, and the risk information are mounted as attributes of the time-series edge.
3. The method of claim 1, wherein, The step of determining the local influence domain of the newly added standard event tuple batch in the multi-level time series diagram based on the temporal proximity, feature similarity, and topological correlation between each newly added standard event tuple and historical standard event tuples in the batch specifically includes: Based on the timestamp difference between the newly added standard event tuples and the historical standard event tuples in the batch of newly added standard event tuples, calculate the temporal proximity between the newly added standard event tuples and the historical standard event tuples; Extract the event attributes of the newly added standard event tuples and the historical standard event tuples from the batch of newly added standard event tuples, and calculate the feature similarity between the newly added standard event tuples and the historical standard event tuples based on the degree of matching of the event attributes. Based on the graph connectivity in the multi-level time sequence graph, calculate the topological association between the newly added standard event tuples and the historical standard event tuples in the batch of newly added standard event tuples; A comprehensive correlation score is calculated based on the temporal proximity, the feature similarity, and the topological correlation. The vertices and temporal edges involved in the newly added standard event tuples, as well as the vertices and temporal edges involved in the historical standard event tuples whose comprehensive correlation scores are higher than a preset threshold, are included in the local influence domain.
4. The method of claim 1, wherein, The step of matching the risk path index with the risk query request to obtain candidate paths and candidate subgraphs specifically includes: Extract the query conditions from the risk query request. The query conditions include at least one of the following: chain identifier, time window, start and end objects, risk type, path length range, and confidence range. The query conditions are matched and retrieved with the risk path index. Based on the matching results, an initial risk path that meets the query conditions is obtained as the candidate path. Extract the set of vertices involved in the candidate path and obtain the associated edges between the vertex sets. The candidate subgraph is constructed by the vertex sets and the associated edges.
5. The method of claim 1, wherein, The multi-level intermediate state set includes node intermediate states, edge intermediate states, subgraph intermediate states, entity intermediate states, and risk path intermediate states; The step of calling the multi-level intermediate state set to execute a query operator based on the candidate path in the candidate subgraph to generate and output query results containing risk paths and explanatory information specifically includes: The query type of the risk query request is parsed, and the corresponding target query operator is scheduled in the candidate subgraph according to the query type; Based on the candidate path, the target query operator is used to read the corresponding state data from the multi-level intermediate state set; During the execution of the target query operator, if it is determined that the status data cannot cover the query conditions of the risk query request or the version is too old, supplementary calculation of the underlying data is triggered in the candidate subgraph. Based on the state data or the results of the supplementary calculations, generate and output query results containing risk paths, key evidence edges, and explanatory labels.
6. The method of claim 5, wherein, The target query operator includes at least one of the following: path expansion operator, pattern matching operator, entity merging operator, time filtering operator, key evidence side verification operator, or version status verification operator; The step of scheduling the corresponding target query operator in the candidate subgraph according to the query type specifically includes: When the query type is a fund tracking query, the path expansion operator is scheduled; When the query type is an anonymous identity association query, the entity merging operator is scheduled; When the query type is a contract exception query, the pattern matching operator is scheduled; When the query type is an evidence chain verification query, the time filtering operator, as well as the key evidence edge verification operator or version status verification operator, are scheduled.
7. The method of claim 1, wherein, Also includes: Receive feedback results from external verification or processing, convert the feedback results into feedback event tuples, and map the feedback event tuples as off-chain evidence edges, entity association edges, or risk event vertices; Using the feedback event tuple as a new standard event tuple, the local influence domain is redefined, and the multi-level intermediate state set and related risk thresholds are corrected within the local influence domain. The updated multi-level intermediate state set and risk path index have new version numbers, so that the state of the multi-level time sequence diagram evolves with the feedback results.
8. A device for incremental update and query of a multi-level timing graph, characterized by include: The module includes a standard event tuple conversion module, a local influence domain determination module, and a query result generation module. The standard event tuple conversion module is used to acquire multi-source heterogeneous data from different levels, convert the multi-source heterogeneous data into standard event tuples, and map the standard event tuples into a multi-level time sequence graph composed of multi-type vertices and multi-type temporal edges. The local influence domain determination module is used to determine the local influence domain of the new standard event tuple batch in the multi-level time sequence diagram based on the temporal proximity, feature similarity and topological correlation between each new standard event tuple in the new standard event tuple batch and the historical standard event tuple when the new standard event tuple batch arrives, and to perform incremental updates on the pre-maintained multi-level intermediate state set and risk path index only within the local influence domain. The query result generation module is used to respond to a risk query request, match the risk path index to obtain candidate paths and candidate subgraphs according to the risk query request, call the multi-level intermediate state set to execute query operators based on the candidate paths in the candidate subgraphs, and generate and output query results containing risk paths and explanatory information.
9. A computer device, characterized in that, The computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the method as described in any one of claims 1-7.
10. A computer-readable storage medium, characterized in that, The storage medium stores a computer program that, when executed by a processor, can implement the method as described in any one of claims 1-7.
