Fund flow path determination method and device, equipment, storage medium and product

Abstract

The embodiment of the invention provides a fund flow path determination method and device, equipment, a storage medium and a product, and relates to the field of big data and financial science and technology. The method comprises the following steps: acquiring transaction data; constructing an initial transaction graph according to the transaction data; based on the initial transaction graph, generating a frequent sub-graph set corresponding to the initial transaction graph through a graph structure mining algorithm, and constructing a transitive closure matrix based on the frequent sub-graph set; and based on the transitive closure matrix, generating a partial order set and constructing a target transaction graph, the target transaction graph being used for representing a fund flow path corresponding to the transaction data. According to the method, the problem that the fund flow global topological evolution rule cannot be captured in the prior art is solved. The relevance of the transaction amount, the account attribute and the topological structure is ensured, and identification omission caused by static rules or single-dimension analysis is avoided. The method significantly improves the recognition capability of a complex fund circulation mode, and reduces the false alarm rate caused by local feature extraction.

Description

Technical Field

[0001] This application relates to the fields of big data and financial technology, and in particular to a method, apparatus, equipment, storage medium and product for determining the flow path of funds. Background Technology

[0002] As financial transactions become more complex and concealed, traditional methods for identifying fund flows are no longer sufficient to address new forms of fund transfers. For example, multi-layered nested accounts and cross-border fund transfers have led to highly dynamic and complex fund flows.

[0003] Current technologies for analyzing fund flows often rely on static rules or single-dimensional graph neural networks, making it difficult to capture the global evolution of fund flows. This is especially true when dealing with multi-hop paths, closed-loop structures, or cross-currency transactions, where key signals are easily missed. Furthermore, financial institutions' transaction data has high-dimensional attributes and dynamic temporal characteristics, and existing methods lack the ability to integrate and analyze such heterogeneous data, resulting in high false alarm rates and slow response times. Summary of the Invention

[0004] This application provides a method, apparatus, equipment, storage medium, and product for determining the path of fund transfer, in order to solve the technical problem that it is difficult to capture the global evolution pattern of fund transfer, resulting in a high false alarm rate and slow response speed.

[0005] Firstly, this application provides a method for determining the path of fund transfer, including:

[0006] Obtain transaction data;

[0007] Based on the transaction data, construct an initial transaction graph, which contains attributed nodes and weighted directed edges;

[0008] Based on the initial transaction graph, a set of frequent subgraphs corresponding to the initial transaction graph is generated through a graph structure mining algorithm, and a transitive closure matrix is ​​constructed based on the set of frequent subgraphs.

[0009] Based on the transitive closure matrix, a partially ordered set is generated and a target transaction graph is constructed. The target transaction graph is used to represent the capital flow path corresponding to the transaction data.

[0010] Secondly, this application provides a device for determining the path of fund transfer, comprising:

[0011] The data acquisition module is used to acquire transaction data;

[0012] The transaction graph construction module is used to construct an initial transaction graph based on transaction data. The initial transaction graph contains attribute nodes and weighted directed edges.

[0013] The transitive closure matrix construction module is used to generate a set of frequent subgraphs corresponding to the initial transaction graph based on the initial transaction graph using a graph structure mining algorithm, and to construct a transitive closure matrix based on the set of frequent subgraphs.

[0014] The target transaction graph construction module is used to generate a partially ordered set and construct a target transaction graph based on the transitive closure matrix. The target transaction graph is used to represent the capital flow path corresponding to the transaction data.

[0015] Thirdly, this application provides an electronic device, including: a processor and a memory communicatively connected to the processor;

[0016] The memory stores the instructions that the computer executes;

[0017] The processor executes computer-executable instructions stored in memory to implement any of the methods of the first aspect.

[0018] Fourthly, this application provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the method of any one of the first aspects.

[0019] Fifthly, this application provides a computer program product, including a computer program that, when executed by a processor, implements the method of any one of the first aspects.

[0020] The method, apparatus, equipment, storage medium, and product for determining fund flow paths provided in this application solve the problem that existing technologies cannot capture the global topological evolution of fund flow through three steps: constructing a transaction graph, generating a set of frequent subgraphs and a transitive closure matrix, and generating a target transaction graph. By constructing the transaction graph, the correlation between transaction amount, account attributes, and topological structure is ensured, providing a complete data foundation for subsequent analysis. High-frequency transaction patterns are extracted using graph structure mining algorithms, avoiding omissions caused by static rules or single-dimensional analysis. Global topological modeling and dynamic path parsing significantly improve the ability to identify complex fund flow patterns while reducing the false positive rate caused by local feature extraction. Attached Figure Description

[0021] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0022] Figure 1 A schematic diagram illustrating a scenario for monitoring fund flows;

[0023] Figure 2 A flowchart illustrating a method for determining the flow path of funds provided in an embodiment of this application;

[0024] Figure 3 This is a schematic diagram of the structure of a fund transfer path determination device provided in an embodiment of this application;

[0025] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application.

[0026] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0027] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0028] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of the relevant data all comply with the relevant laws, regulations, and standards of the relevant countries and regions, have taken necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation access points for users to choose to authorize or refuse.

[0029] Furthermore, the technical solution involved in this application, which involves big data analysis of user information (including but not limited to personal biometrics, identity data, consumption data, asset data, electronic terminal operation data, etc.) and the use of artificial intelligence technology for automated decision-making, and makes decisions that have a significant impact on personal rights based on the results of automated decision-making, provides users with corresponding operation entry points for users to choose to agree to or reject the results of automated decision-making; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0030] It should be noted that the methods, apparatus, equipment, storage media and products for determining fund transfer paths provided in this application can be used in the fields of big data and fintech, as well as in any other fields. The application fields of the methods, apparatus, equipment, storage media and products for determining fund transfer paths in this application are not limited.

[0031] The specific application scenario of this application is the monitoring and risk warning of fund transfers in financial institutions such as banks. Figure 1 A schematic diagram illustrating a scenario for monitoring fund flows, such as... Figure 1 As shown, financial institutions typically monitor fund flows in two ways: 1. Static rule engine-based monitoring: This method uses preset rules to make hard judgments on transaction behavior data. For example, if the cumulative daily transfer amount exceeds a threshold (e.g., $50,000) or the counterparty is in a preset list, an alarm is triggered. This method relies on manual experience to set rules, but rule updates are lagging, making it difficult to adapt to the rapid iteration of new fund flow patterns, and it has weak recognition capabilities for complex topological structures (e.g., multi-layer nested closed loops). 2. Graph neural network anomaly detection: Graph neural networks extract node embedding features from the transaction graph corresponding to transaction behavior data and combine them with fully connected networks for risk classification. Its core process includes transaction graph construction, feature extraction, graph embedding vector generation, and risk classification. However, this method is usually only applicable to static graph structures, cannot effectively model the dynamic transmission closure of fund flows (e.g., multi-hop path reachability), and has insufficient recognition rate for nested closed loop structures with more than 3 layers. In addition, existing graph neural network (GNN) models often focus on specific tracking scenarios and lack adaptability to traditional banking transaction scenarios. This results in a disconnect between attribute dimensions (such as account type, geographic labels, etc.) and topological structure correlation analysis, leading to a high false alarm rate.

[0032] Therefore, existing methods for monitoring fund flows have the following shortcomings: Existing methods only focus on local transaction behaviors or static graph structures, failing to capture the global topological evolution of fund flows (such as multi-layered closed loops and cross-border jump paths), resulting in insufficient ability to identify complex fund flow patterns. Existing technologies lack the ability to model the dynamic evolution of risk patterns and cannot analyze the transitive closure relationships of fund flows (such as multi-hop path reachability), leading to a zero-sample recognition rate of less than 40% for novel fund flow patterns. Existing methods do not fully integrate the correlation between transaction amount, account attributes (such as regional tags and historical risk scores), and topological structure, resulting in a high false positive rate and difficulty in adapting to the analysis needs of dynamic time-series transaction data. The training and updating of existing GNN models are time-consuming, making it difficult to support minute-level risk pattern iteration needs, and computational efficiency is low when processing large-scale heterogeneous transaction data.

[0033] The method, apparatus, equipment, storage medium, and product for determining the fund flow path provided in this application map transaction data into a transaction graph, extract a set of frequent subgraphs from the transaction graph using a graph structure mining algorithm, construct a transitive closure matrix based on the set of frequent subgraphs, transform the transitive closure matrix into a partially ordered set, and generate a Hasse graph corresponding to the target transaction graph, reflecting the fund flow path corresponding to the transaction. This aims to solve the above-mentioned technical problems of the prior art.

[0034] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0035] Figure 2 This is a flowchart illustrating a method for determining the flow path of funds provided in an embodiment of this application, as shown below. Figure 2 As shown, the method includes:

[0036] S201, Obtain transaction data.

[0037] In one example, transaction data may include data recorded by financial institutions regarding transaction activities, such as account information, transaction amount, timestamps, etc., like transfer transaction data in bank statements.

[0038] S202. Construct an initial transaction graph based on the transaction data.

[0039] In this embodiment of the application, the initial transaction graph includes attributed nodes and weighted directed edges.

[0040] In one example, the initial transaction graph comprises a data structure consisting of attributed nodes and weighted directed edges. Attributed nodes represent entities involved in fund flows, while weighted directed edges represent transaction relationships. Both attributed nodes and weighted directed edges contain attribute information. The transaction data is mapped to the initial transaction graph, where the attributes of attributed nodes include account type, geographic label, and historical risk score, and the attributes of weighted directed edges include transaction amount, frequency, and flow direction.

[0041] S203. Based on the initial transaction graph, generate a set of frequent subgraphs corresponding to the initial transaction graph through a graph structure mining algorithm, and construct a transitive closure matrix based on the set of frequent subgraphs.

[0042] In one example, a set of frequent subgraphs is extracted from the initial transaction graph using a graph structure mining algorithm. This set of frequent subgraphs contains subgraph patterns that satisfy a minimum support threshold, such as multi-layered closed-loop transaction paths or jump paths. A transitive closure matrix is ​​constructed based on the set of frequent subgraphs. An improved multi-source shortest path planning algorithm is used to calculate the reachability of multi-hop paths between nodes or subgraphs, and a pre-defined consistency constraint is introduced to avoid cross-currency path interference.

[0043] Among them, the graph structure mining algorithm is used to characterize the algorithm for extracting recurring subgraph patterns from the transaction graph. It generates candidate subgraphs through edge expansion and filters frequent subgraphs. The transitive closure matrix is ​​applied to a Boolean matrix representing the reachability relationships between nodes or subgraphs in the transaction graph, such as calculating the reachability of multi-hop paths.

[0044] S204. Based on the transitive closure matrix, generate a partially ordered set and construct the target transaction graph.

[0045] In this embodiment of the application, the target transaction graph is used to represent the fund flow path corresponding to the transaction data.

[0046] In one example, the target transaction graph can be a Hasse graph. By transforming the transitive closure matrix into a partially ordered set and generating the Hasse graph through topological sorting, the target transaction graph is generated. This graph reflects the evolution path of the capital flow pattern through the node hierarchy and edge connections, such as the change from a single-layer closed loop to a multi-layer nested topological structure.

[0047] The partially ordered set is an ordered set generated based on the transitive closure matrix, used to represent partial order relationships between frequent subgraphs, such as generating a Hasse graph through topological sorting. The Hasse graph is a visual graph structure built based on the partially ordered set, used to represent the evolutionary paths between frequent subgraphs, such as reflecting dynamic changes in risk patterns through node hierarchy and edge connections.

[0048] In one implementation scenario, constructing an initial transaction graph may include frequent edge filtering. The frequency of weighted directed edges is counted based on a sliding time window. Weighted directed edges that satisfy the following conditions can be identified as frequent edges, i.e., the first transaction edges: In the formula, α is the dynamic adjustment coefficient. The total number of transactions. Generating a set of frequent subgraphs corresponding to the initial transaction graph using a graph structure mining algorithm can include: expanding candidate subgraphs from the frequent edges obtained above using a frequent subgraph mining algorithm, retaining subgraphs that satisfy the minimum support threshold β, and obtaining the set of frequent subgraphs. Constructing a transitive closure matrix based on the frequent subgraph set can include: applying an improved multi-source shortest path planning algorithm to the frequent subgraph set, adding a constraint on the reachability of fund flows. Generating a partially ordered set and constructing the target transaction graph based on the transitive closure matrix can include: converting the transitive closure matrix into a partially ordered set, constructing a risk pattern evolution path graph through topological sorting, which is the target transaction graph.

[0049] For example, the similarity between the image to be detected and the frequent subgraphs can also be calculated using a multimodal similarity calculation model. The similarity between the image to be detected and the frequent subgraphs can be obtained by calculating the composite metric formula as follows:

[0050]

[0051] In the formula, The number of nodes / edges in the common substructure; For transaction volume volatility, γ is the volatility penalty factor (default value). ); All are weighted coefficients (optimized values) ).

[0052] For example, the risk level corresponding to the transaction graph to be detected is obtained through the following risk level decision tree:

[0053] graph TD

[0054] Does A[maximum similarity frequent subgraph] --> B{have multiple loops?}

[0055] B --> | Yes | C [Directly mark L3 high risk]

[0056] B -->|No| D{Number of associated subgraphs > 1?}

[0057] D -->|Yes| E [Tracing upstream nodes in the Hasse diagram]

[0058] D -->|No| F [Classified by similarity threshold]

[0059] E --> G [Take the highest risk level in the path]

[0060] F --> H[Sim>0.85: L3]

[0061] F --> I[0.6≤Sim≤0.85:L2]

[0062] F --> J[Sim<0.6:L1]

[0063] The method for determining fund flow paths provided in this embodiment solves the problem that existing technologies cannot capture the global topological evolution of fund flows through three steps: constructing a transaction graph, generating a set of frequent subgraphs and a transitive closure matrix, and generating a target transaction graph. By constructing the transaction graph, the correlation between transaction amounts, account attributes, and the topological structure is ensured, providing a complete data foundation for subsequent analysis. High-frequency trading patterns are extracted using graph structure mining algorithms, avoiding omissions caused by static rules or single-dimensional analysis. Global topological modeling and dynamic path parsing significantly improve the ability to identify complex fund flow patterns while reducing the false positive rate caused by local feature extraction.

[0064] Optionally, based on the initial transaction graph, a set of frequent subgraphs corresponding to the initial transaction graph is generated using a graph structure mining algorithm, including: extracting a set of candidate subgraphs from the initial transaction graph through edge expansion; filtering the set of candidate subgraphs to obtain subgraphs that meet a preset support threshold as frequent subgraphs, thereby generating a set of frequent subgraphs corresponding to the initial transaction graph.

[0065] In one example, the edge expansion algorithm is a graph traversal algorithm that generates candidate subgraphs by progressively expanding edges. The support threshold can include a minimum support threshold, a preset condition for filtering frequent subgraphs; for example, subgraphs with support ≥ 5% are retained.

[0066] For example, the edge expansion algorithm starts with an edge in the transaction graph and gradually expands to generate candidate subgraphs. For instance, starting with a single edge "Account A → Account B", it generates a multi-edge subgraph (such as "Account A → Account B → Account C") by adding adjacent edges. During the screening phase, candidate subgraphs are retained based on a minimum support threshold (e.g., support ≥ 5%) to ensure that the generated set of frequent subgraphs contains only significant transaction patterns.

[0067] By employing an edge expansion algorithm and support filtering, accurate extraction of high-frequency topological patterns in transaction graphs is achieved. Dynamic edge expansion generates candidate subgraphs, ensuring coverage of multi-layered nested structures. Simultaneously, a support threshold filters redundant patterns, avoiding wasted computational resources and thus improving the ability to identify complex fund flow patterns.

[0068] Optionally, constructing a transitive closure matrix based on a set of frequent subgraphs includes: traversing the set of frequent subgraphs, obtaining the path reachability between each frequent subgraph, and obtaining an initial transitive closure matrix; updating the initial transitive closure matrix based on a preset consistency constraint, and constructing the transitive closure matrix.

[0069] In one example, path reachability calculation is used to compute reachability between nodes or subgraphs in a transaction graph. Preset consistency constraints may include restricting the reachability calculation of cross-currency paths, such as allowing only transaction paths of the same currency to participate in transitive closure construction.

[0070] For example, a path reachability algorithm is used to traverse the set of frequent subgraphs and calculate the multi-hop path reachability between any two subgraphs. For instance, if subgraph A is reachable from subgraph B to subgraph C, and A, B, and C are all transaction type 1, then A→C is marked as reachable in the transitive closure matrix. Pre-defined consistency constraints ensure that cross-currency paths do not participate in the transitive closure calculation, avoiding interference with topology analysis due to currency conversion.

[0071] By traversing the set of frequent subgraphs using a path reachability algorithm, the reachability of multi-hop paths between any two subgraphs is calculated, achieving accurate modeling of multi-hop paths in fund transfers. Invalid paths are filtered out by pre-set consistency constraints, ensuring that cross-currency paths do not participate in the transitive closure calculation, avoiding interference with topology analysis due to currency conversion, and improving the accuracy of identifying cross-border transfer paths.

[0072] Optionally, based on the transitive closure matrix, a partially ordered set is generated and a target transaction graph is constructed, including: generating a partially ordered set based on the reachability relationships in the transitive closure matrix; and sorting the partially ordered set using a topological sorting algorithm to construct the target transaction graph.

[0073] In one example, the topological sorting algorithm is used to linearly sort the elements in a partially ordered set. Boolean values ​​in the transitive closure matrix (e.g., 1 for A→B) indicate that subgraph A is reachable from subgraph B. Based on the reachability relationships in the transitive closure matrix, a partially ordered set is generated, where subgraph A≤B if and only if A is reachable from B. The subgraphs in the partially ordered set are sorted using the topological sorting algorithm to construct a Hasse graph, i.e., the target transaction graph. For example, a multi-layered closed-loop transaction pattern can be used as the parent node, and a single-layered closed-loop pattern as the child node, reflecting the evolution path of the capital flow pattern through hierarchical relationships.

[0074] By generating a partially ordered set based on the transitive closure matrix, dynamic modeling of the evolution path of risk patterns is achieved. The subgraphs in the partially ordered set are then sorted using topological sorting to obtain the target transaction graph, clarifying the partially ordered relationships between the subgraphs. This supports zero-sample identification of novel capital flow patterns, thereby enhancing the dynamic analysis capability of capital flow patterns.

[0075] Optionally, after constructing the target transaction graph, the similarity between the target transaction graph and the preset frequent subgraphs is obtained through multimodal similarity calculation; based on the similarity, the level corresponding to the target transaction graph is determined.

[0076] In one example, multimodal similarity calculation can be performed by combining a multimodal similarity calculation model with node / edge number matching and attribute feature similarity calculation methods, such as weighted similarity formulas.

[0077] For example, similarity is calculated between the target transaction graph and a preset frequent subgraph, combining topological structure (such as the number of common nodes / edges) and attribute features. For instance, if the common node ratio between the target graph and a certain frequent subgraph is 80%, and the volatility difference is less than a preset threshold, then the similarity is considered high. The model can generate the final similarity value using a weighted formula, which is then used to determine the risk level.

[0078] By calculating multimodal similarity, multi-dimensional matching between transaction graphs and risk patterns is achieved. Through weighted fusion of multimodal features, false alarms caused by similar topological structures but different attribute features are reduced, as are misjudgments caused by single-dimensional analysis. This significantly improves the accuracy of risk identification and supports the analysis needs of dynamic time-series transaction data.

[0079] Optionally, before extracting a set of candidate subgraphs from the initial transaction graph through edge expansion, the method further includes: obtaining the frequency of transaction edges in the initial transaction graph based on a preset sliding time window; and selecting a first transaction edge based on the frequency of the transaction edges according to a dynamic adjustment coefficient, wherein the first transaction edge is used to characterize the transaction edges whose frequency meets the preset conditions.

[0080] In one example, the sliding time window is a window of transaction data that is dynamically updated at fixed time intervals. The dynamic adjustment coefficient can be a threshold coefficient that is dynamically adjusted based on the distribution of historical transaction data. The sliding time window dynamically updates the transaction data and counts the frequency of each type of transaction edge (e.g., "Account A → Account B"). The dynamic adjustment coefficient α can be calculated in real time based on historical transaction patterns, for example, α_min = μ - σ, α_max = μ + σ. Only transaction edges with a frequency percentage ≥ α·N_total are retained as high-frequency edges to ensure that candidate subgraphs are generated based on the latest transaction trends.

[0081] By using a sliding time window and dynamically adjusting coefficients, real-time adaptation to changes in the time series of transaction data is achieved. Dynamically filtering high-frequency edges avoids omissions caused by fixed thresholds, thereby enhancing the comprehensive capture capability of risk signals. This significantly improves the sensitivity and adaptability of frequent edge filtering, ensuring comprehensive capture of risk signals.

[0082] Optionally, traversing the set of frequent subgraphs to obtain the path reachability between each frequent subgraph and obtaining the initial transitive closure matrix, also includes: using a distributed computing framework to traverse the set of frequent subgraphs and obtain the path reachability between each frequent subgraph in parallel; and updating the initial transitive closure matrix based on a preset incremental update mechanism.

[0083] In one example, the distributed computing framework serves as a computing platform for parallel processing tasks. The incremental update mechanism involves triggering local transitive closure updates only for newly added or modified subgraphs, avoiding global recomputation. The distributed computing framework divides the set of frequent subgraphs into multiple subtasks, utilizing a resilient distributed dataset to process the reachability calculations of each subtask in parallel. By pre-setting an incremental update mechanism, local updates are triggered only for newly added or modified subgraphs (such as newly emerging transaction graphs), avoiding the recomputation of the global transitive closure matrix and significantly reducing computational resource consumption.

[0084] By dividing the set of frequent subgraphs into multiple subtasks and processing the transitive closure calculations of each subtask in parallel, the computational bottleneck of a single node is reduced. By triggering local transitive closure updates only for newly added or modified frequent subgraphs, global recalculation is avoided, significantly reducing computational resource consumption. The combination of an incremental update mechanism and a sliding time window ensures that the update of the transitive closure matrix is ​​synchronized with the dynamic changes in transaction data, meeting the requirement for minute-level response.

[0085] Optionally, the level corresponding to the target transaction map is determined based on similarity, including: using a level decision tree to classify the target transaction map based on similarity, and determining the level corresponding to the target transaction map. The level decision tree includes several preset judgment rules.

[0086] In one example, the hierarchical decision tree can include a hierarchical decision model based on preset decision rules, such as marking multi-layered closed loops as the first level. The hierarchical decision tree is used to classify the target transaction graph based on multimodal similarity results. For example, if the similarity between the target transaction graph and a frequent subgraph is ≥0.85 and it contains a multi-layered closed loop structure, it is marked as L3 high risk; if the similarity is between 0.6 and 0.85 and the number of associated subgraphs is >1, the upstream nodes of the target transaction graph are traced for risk classification.

[0087] The hierarchical decision tree enables refined classification of different risk levels. Pre-defined judgment rules quickly identify high-risk transactions, while Hasse chart traceability supports dynamic analysis of low-risk patterns, thereby improving the real-time nature and accuracy of risk response.

[0088] Figure 3 This is a schematic diagram of the structure of a fund transfer determination device provided in an embodiment of this application, as shown below. Figure 3 As shown, the fund transfer determination device 30 provided in this embodiment includes:

[0089] Data acquisition module 301 is used to acquire transaction data;

[0090] The transaction graph construction module 302 is used to construct an initial transaction graph based on transaction data. The initial transaction graph contains attribute nodes and weighted directed edges.

[0091] The transitive closure matrix construction module 303 is used to generate a set of frequent subgraphs corresponding to the initial transaction graph based on the initial transaction graph through a graph structure mining algorithm, and to construct a transitive closure matrix based on the set of frequent subgraphs.

[0092] The target transaction graph construction module 304 is used to generate a partially ordered set and construct a target transaction graph based on the transitive closure matrix. The target transaction graph is used to represent the capital flow path corresponding to the transaction data.

[0093] In one possible implementation, the transitive closure matrix construction module 303 is specifically used to: extract a set of candidate subgraphs from the initial transaction graph through edge expansion; filter the set of candidate subgraphs to obtain subgraphs that meet a preset support threshold as frequent subgraphs, and generate a set of frequent subgraphs corresponding to the initial transaction graph.

[0094] In one possible implementation, the transitive closure matrix construction module 303 is further specifically used to: traverse the set of frequent subgraphs, obtain the path reachability between each frequent subgraph, and obtain the initial transitive closure matrix; update the initial transitive closure matrix based on a preset consistency constraint, and construct the transitive closure matrix.

[0095] In one possible implementation, the target transaction graph construction module 304 is specifically used to: generate a partially ordered set based on the reachability relations in the transitive closure matrix; and sort the partially ordered set using a topological sorting algorithm to construct the target transaction graph.

[0096] In one possible implementation, the fund flow determination device is further specifically used to: after constructing the target transaction graph, obtain the similarity between the target transaction graph and the preset frequent subgraph through multimodal similarity calculation; and determine the level corresponding to the target transaction graph based on the similarity.

[0097] In one possible implementation, the fund flow determination device is further specifically used to: obtain the frequency of transaction edges in the initial transaction graph based on a preset sliding time window; and, based on the frequency of the transaction edges and a dynamic adjustment coefficient, filter to obtain a first transaction edge, which is used to characterize transaction edges whose frequency meets preset conditions.

[0098] In one possible implementation, the transitive closure matrix construction module 303 is also specifically used to: traverse the set of frequent subgraphs using a distributed computing framework, and obtain the path reachability between each frequent subgraph in parallel; and update the initial transitive closure matrix based on a preset incremental update mechanism.

[0099] In one possible implementation, the fund flow determination device is further specifically used to: classify the target transaction map based on similarity using a graded decision tree, and determine the grade corresponding to the target transaction map. The graded decision tree includes several preset judgment rules.

[0100] The fund transfer determination device provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0101] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 40 may include a memory 401 and a processor 402. Optionally, the electronic device may also include a transceiver 403, wherein the memory 401 and the processor 402 communicate with each other; for example, the memory 401, the processor 402 and the transceiver 403 may communicate via a communication bus 404, the memory 401 is used to store a computer program, and the processor 402 executes the computer program to implement the method of the above embodiments.

[0102] Optionally, the aforementioned processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps in the method embodiments disclosed in this application can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0103] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the methods in any of the above method embodiments.

[0104] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the methods in any of the above method embodiments.

[0105] All or part of the steps in the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a readable memory. When the program is executed, it performs the steps of the above method embodiments; and the aforementioned memory (storage medium) includes: read-only memory (ROM), RAM, flash memory, hard disk, solid-state drive, magnetic tape, floppy disk, optical disk, and any combination thereof.

[0106] This application describes embodiments with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processing unit of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processing unit of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0107] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0108] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0109] Obviously, those skilled in the art can make various modifications and variations to the embodiments of this application without departing from the spirit and scope of this application. Therefore, if these modifications and variations to the embodiments of this application fall within the scope of the claims of this application and their equivalents, this application also intends to include these modifications and variations.

[0110] In this application, the term "comprising" and its variations can refer to non-limiting inclusion; the term "or" and its variations can refer to "and / or". The terms "first", "second", etc., in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. In this application, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0111] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to this application.

[0112] It should be further noted that although the steps in the flowchart are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowchart may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.

[0113] It should be understood that the above-described device embodiments are merely illustrative, and the device of this application can also be implemented in other ways. For example, the division of units / modules in the above embodiments is only a logical functional division, and there may be other division methods in actual implementation. For example, multiple units, modules, or components may be combined, or integrated into another system, or some features may be ignored or not executed.

[0114] Furthermore, unless otherwise specified, the functional units / modules in the various embodiments of this application can be integrated into one unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated together. The integrated units / modules described above can be implemented in hardware or as software program modules.

[0115] When integrated units / modules are implemented in hardware, the hardware can be digital circuits, analog circuits, etc. The physical implementation of the hardware structure includes, but is not limited to, transistors, memristors, etc. Unless otherwise specified, the processor can be any suitable hardware processor, such as a CPU, GPU, FPGA, DSP, and ASIC, etc. Unless otherwise specified, the storage unit can be any suitable magnetic or magneto-optical storage medium, such as Resistive Random Access Memory (RRAM), Dynamic Random Access Memory (DRAM), Static Random Access Memory (SRAM), Enhanced Dynamic Random Access Memory (EDRAM), High-Bandwidth Memory (HBM), Hybrid Memory Cube (HMC), etc.

[0116] If the integrated unit / module is implemented as a software program module and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). 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 memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0117] In the above embodiments, the descriptions of each embodiment have their own emphasis. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments. The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0118] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this application are indicated by the following claims.

[0119] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.

Claims

1. A method for determining a fund flow path, characterized in that, The method comprises: acquiring transaction data; constructing an initial transaction graph according to the transaction data, the initial transaction graph comprising attribute nodes and weighted directed edges; generating a set of frequent subgraphs corresponding to the initial transaction graph through a graph structure mining algorithm based on the initial transaction graph, and constructing a transitive closure matrix based on the set of frequent subgraphs; generating a partial order set and constructing a target transaction graph based on the transitive closure matrix, the target transaction graph being used to represent a fund flow path corresponding to the transaction data.

2. The method of claim 1, wherein, The generating of the set of frequent subgraphs corresponding to the initial transaction graph through the graph structure mining algorithm based on the initial transaction graph comprises: extracting a set of candidate subgraphs from the initial transaction graph through edge expansion; screening the set of candidate subgraphs to obtain subgraphs satisfying a preset support threshold as frequent subgraphs, thereby generating the set of frequent subgraphs corresponding to the initial transaction graph.

3. The method of claim 2, wherein, The construction of the transitive closure matrix based on the set of frequent subgraphs comprises: traversing the set of frequent subgraphs to obtain path reachability between the frequent subgraphs, thereby obtaining an initial transitive closure matrix; updating the initial transitive closure matrix based on a preset consistency constraint, thereby constructing the transitive closure matrix.

4. The method of claim 3, wherein, The generating of the partial order set and the constructing of the target transaction graph based on the transitive closure matrix comprises: generating the partial order set based on the reachability relationship in the transitive closure matrix; sorting the partial order set through a topological sorting algorithm to construct the target transaction graph.

5. The method of claim 4, wherein, After the constructing of the target transaction graph, a multi-modal similarity calculation is performed to obtain a similarity between the target transaction graph and a preset frequent subgraph; based on the similarity, determining a level corresponding to the target transaction graph.

6. The method of claim 2, wherein, Before the extracting of the set of candidate subgraphs from the initial transaction graph through edge expansion, the method further comprises: based on a preset sliding time window, acquiring a frequency of transaction edges in the initial transaction graph; based on the frequency of the transaction edges, screening a first transaction edge according to a dynamic adjustment coefficient, the first transaction edge being used to represent a transaction edge satisfying a preset condition in terms of frequency.

7. The method of claim 3, wherein, The traversing of the set of frequent subgraphs to obtain path reachability between the frequent subgraphs, thereby obtaining the initial transitive closure matrix, further comprises: using a distributed computing framework to traverse the set of frequent subgraphs and obtain path reachability between the frequent subgraphs in parallel; updating the initial transitive closure matrix based on a preset incremental update mechanism.

8. The method of claim 5, wherein, The determining of the level corresponding to the target transaction graph based on the similarity comprises: based on the similarity, using a level decision tree to perform hierarchical judgment on the target transaction graph, thereby determining the level corresponding to the target transaction graph, the level decision tree comprising a plurality of preset judgment rules.

9. A fund flow determination apparatus characterized by comprising: The device comprises: a data acquisition module configured to acquire transaction data; a transaction graph construction module configured to construct an initial transaction graph according to the transaction data, the initial transaction graph comprising attribute nodes and weighted directed edges; The transmission closure matrix construction module is configured to generate a frequent subgraph set corresponding to the initial transaction graph by a graph structure mining algorithm based on the initial transaction graph, and construct a transmission closure matrix based on the frequent subgraph set. The target transaction graph construction module is configured to generate a partial order set and construct a target transaction graph based on the transmission closure matrix, and the target transaction graph is used to represent a fund flow path corresponding to the transaction data.

10. An electronic device, comprising: The method comprises: a processor, and a memory connected to the processor in communication; the memory stores computer execution instructions; the processor executes the computer execution instructions stored in the memory to implement the method of any one of claims 1 to 8.

11. A computer readable storage medium, characterized in that, The computer readable storage medium stores computer execution instructions, and the computer execution instructions are executed by the processor to implement the method of any one of claims 1 to 8.

12. A computer program product, characterised in that, The computer program is executed by the processor to implement the method of any one of claims 1 to 8.

Images