Cross-border transaction risk node identification method and device based on knowledge graph
By constructing a cross-border transaction knowledge graph and utilizing a group risk prediction model and centrality algorithm, the problem of incomplete identification of risk nodes in cross-border transactions was solved, enabling quantitative analysis and accurate identification of key nodes within risk groups.
Patent Information
- Application Number
- CN202411044076.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-31
- Publication Date
- 2024-11-22
AI Technical Summary
Existing technologies struggle to accurately identify risk nodes and key nodes within risk groups in cross-border transactions, fail to effectively identify all risk groups, and lack quantitative risk measurement for cross-border risk factors.
A cross-border transaction knowledge graph is constructed. By inputting group attribute features into the group risk prediction model, the risk level of each node in the risk group is determined by combining centrality algorithms and graph features.
It enables accurate identification of risk nodes in cross-border transactions and quantitative analysis of key nodes within risk groups, improving the ability to detect risks comprehensively and reducing the burden of business investigation.
Smart Images

Figure CN119005988A_ABST
Abstract
Description
Technical Field
[0001] This specification relates to the field of financial technology, and in particular to a method and device for identifying risk nodes in cross-border transactions based on a knowledge graph. Background Art
[0002] With the continuous innovation of financial technology and the increasingly diversified competitive environment of payment products, commercial banks are faced with many challenges such as business development and compliance operations, user experience and transaction security, data mining and privacy protection. Risk prevention and control work is facing tremendous pressure. Good and stable risk control capabilities have become an important support for financial security.
[0003] When building a customer graph, the existing technology uses a weakly connected graph algorithm to segment the large cross-border transaction graph to obtain several sub-graphs, and then builds a knowledge graph based on the sub-graphs. The accuracy of this segmentation method is poor in actual business scenarios. When dividing communities, it is difficult to divide nodes that are not directly related to seed nodes into communities, that is, they are basically 1-degree transactions. Therefore, it has limitations when looking for domestic funding source nodes, that is, the search capability is weak.
[0004] In addition, the existing technology lacks quantitative risk measurement for cross-border risk factors (such as risk groups), and it is impossible to know the degree of suspicion of risk activities in a certain group, explain which are the key nodes within the risk group, and the role of different nodes within the risk group, that is, which ones play the role of "bridge" in the risk process, and which nodes are the real fund collection nodes. A risk group mined through the knowledge graph involves hundreds or thousands of nodes, which is difficult to check in this case. Summary of the invention
[0005] In order to solve the problem that the risk node identification in the prior art is incomplete and the risks of each node within the risk group cannot be checked, the embodiments of this specification provide a method and device for identifying cross-border transaction risk nodes based on a knowledge graph.
[0006] The embodiments of the present specification provide a method for identifying cross-border transaction risk nodes based on a knowledge graph, the method comprising: constructing a cross-border transaction knowledge graph based on cross-border transaction behaviors between multiple nodes, wherein the cross-border transaction knowledge graph is constructed with transactions between nodes as relationships and accounts of each node as entities; obtaining group attribute features from the cross-border transaction knowledge graph, wherein the group attribute features correspond to a group composed of multiple nodes; inputting the group attribute features into a group risk prediction model to obtain a group risk prediction result, wherein the group risk prediction model is obtained by training a classification model based on group attribute sample features and sample risk labels; if the group risk prediction result indicates that the group is at risk, determining the risk level of each node in the risk group using a centrality algorithm and graph features in the cross-border transaction knowledge graph.
[0007] According to one aspect of the embodiments of this specification, the group risk prediction model is trained in the following manner: based on the cross-border transaction sample knowledge graph of the sample node, the group attribute sample characteristics and the sample risk label are obtained; the group attribute sample characteristics are input into the classification model to obtain the initial prediction result output by the classification model; based on the loss value between the initial prediction result and the sample risk label, the model parameters of the classification model are iteratively updated to obtain the group risk prediction model.
[0008] According to one aspect of an embodiment of the present specification, using a centrality algorithm and graph features in the cross-border transaction knowledge graph to determine the risk level of each node in a risk group includes: calculating the graph features of each node in the cross-border transaction knowledge graph based on the number of nodes and the number of edges in the cross-border transaction knowledge graph, the graph features including degree, out-degree and in-degree; determining the graph index of each node based on the degree, out-degree, in-degree and centrality algorithm; determining the risk level of each node in the group based on the graph index of each node.
[0009] According to one aspect of an embodiment of the present specification, determining the graph index of each node according to the graph feature and the centrality algorithm includes: determining the degree centrality of each node according to the out-degree of each node, the in-degree of each node and the number of nodes in the risk group: in, represents the in-degree of the ith node, represents the out-degree of the ith node, v i represents the degree of the ith node in the subgraph, and n represents the number of nodes in the population.
[0010] According to one aspect of an embodiment of this specification, determining the graph index of each node according to the graph feature and the centrality algorithm includes: determining the closeness centrality according to the following formula: Among them, C c (v i ) represents the i-th node v i The closeness centrality, I ij represents the i-th node v i and the jth node v j The shortest distance between Represents node v i The average shortest path between all other nodes, n is the number of nodes in the group.
[0011] According to one aspect of an embodiment of this specification, determining the graph index of each node according to the graph feature and the centrality algorithm includes: determining the betweenness centrality according to the following formula: Among them, C b (v i ) represents the i-th node vi The betweenness centrality, N jk Represents node v i and node v j The number of shortest paths between jk (i) represents node v i and node v j The shortest path between i The number of .
[0012] According to one aspect of the embodiments of this specification, the cross-border transaction knowledge graph is constructed in the following manner: determining the account information and the first transaction time of the domestic first-level node that conducts transactions with overseas accounts; determining the account information and the second transaction time of the domestic second-level node based on the transaction behavior of the domestic first-level node with other domestic nodes within the preset time range of the first transaction time; associating and expanding the transactions of the domestic second-level nodes within the preset time range of the second transaction time to determine multi-level transaction nodes and transaction times; and constructing a cross-border transaction knowledge graph by taking the transaction behavior between each node as an edge and the account of each node as an entity.
[0013] The embodiments of the present specification also disclose a cross-border transaction risk node identification device based on a knowledge graph, the device comprising: a graph construction unit, for constructing a cross-border transaction knowledge graph according to cross-border transaction behaviors between a plurality of nodes, wherein the cross-border transaction knowledge graph is constructed with transactions between nodes as relationships and accounts of each node as entities; an attribute acquisition unit, for acquiring group attribute features from the cross-border transaction knowledge graph, wherein the group attribute features correspond to a group composed of a plurality of nodes; a prediction result determination unit, for inputting the group attribute features into a group risk prediction model to obtain a group risk prediction result, wherein the group risk prediction model is obtained by training a classification model based on group attribute sample features and sample risk labels; a risk level determination unit, for determining the risk level of each node in the risk group using a centrality algorithm and graph features in the cross-border transaction knowledge graph if the group risk prediction result indicates that the group is at risk.
[0014] An embodiment of the present specification also provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the method for identifying cross-border transaction risk nodes based on the knowledge graph is implemented.
[0015] The embodiments of this specification also provide a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the cross-border transaction risk node identification method based on the knowledge graph.
[0016] The embodiments of this specification also provide a computer program product, which includes a computer program, and when the computer program is executed by a processor, it implements the cross-border transaction risk node identification method based on the knowledge graph.
[0017] This manual combines graph computing technology to quantify and analyze the different risks of members within a risk group, and can clearly identify the location of certain important nodes in the group. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0019] Figure 1 The figure shows a flow chart of a method for identifying risk nodes of cross-border transactions based on a knowledge graph according to an embodiment of the present specification;
[0020] Figure 2 The figure is a flow chart of a method for training a group risk prediction model according to an embodiment of the present specification;
[0021] Figure 3 The figure is a flow chart of a method for determining the risk level of each node in a risk group according to an embodiment of the present specification;
[0022] Figure 4 The figure is a flow chart of a method for constructing a cross-border transaction knowledge graph according to an embodiment of this specification;
[0023] Figure 5 The figure shows a schematic diagram of the structure of a cross-border transaction risk node identification device based on a knowledge graph according to an embodiment of the present specification;
[0024] Figure 6 The figure shows a schematic diagram of a transaction knowledge graph in a cross-border split remittance scenario in an embodiment of this specification;
[0025] Figure 7 The figure shows a schematic diagram of the structure of a computer device according to an embodiment of the present specification.
[0026] Description of the accompanying symbols:
[0027] 501, graph construction unit;
[0028] 502. attribute acquisition unit;
[0029] 503. prediction result determination unit;
[0030] 504. Risk level determination unit;
[0031] 702. Computer equipment;
[0032] 704, processor;
[0033] 706. Memory;
[0034] 708, driving mechanism;
[0035] 710, input / output module;
[0036] 712. Input devices;
[0037] 714. Output devices;
[0038] 716. Presentation equipment;
[0039] 718. Graphical user interface;
[0040] 720, network interface;
[0041] 722. Communication link;
[0042] 724. Communication bus. DETAILED DESCRIPTION
[0043] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this specification.
[0044] It should be noted that the terms "first", "second", etc. in the description and claims of this specification and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of this specification described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, device, product or equipment that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or equipment.
[0045] This specification provides method operation steps as described in the embodiments or flow charts, but more or fewer operation steps may be included based on routine or non-creative work. The order of steps listed in the embodiments is only one way of executing the order of many steps and does not represent the only execution order. When the system or device product is executed in practice, it can be executed in the order of the method shown in the embodiments or the drawings or in parallel.
[0046] It should be noted that the cross-border transaction risk node identification method based on knowledge graph described in this specification can be used in the field of financial technology, and this specification does not limit the application field of the cross-border transaction risk node identification method and device based on knowledge graph.
[0047] Figure 1 The flowchart of a method for identifying risk nodes of cross-border transactions based on a knowledge graph in an embodiment of this specification is shown, which specifically includes the following steps:
[0048] Step 101, construct a cross-border transaction knowledge graph based on cross-border transaction behaviors between multiple nodes, wherein the cross-border transaction knowledge graph is constructed with transactions between nodes as relationships and accounts of each node as entities. In this step, the cross-border transaction behaviors between multiple nodes are the cross-border transaction behaviors to be studied. The cross-border transaction behaviors between multiple nodes can be obtained through the banking system.
[0049] In this step, by defining the knowledge graph point-edge schema table structure, taking the accounts of multiple nodes as entities and the transactions between multiple nodes as relationships, the corresponding attributes of the entity relationship (for example, customer name, transaction amount, etc.) are designed, and knowledge processing and extraction are performed according to the designed graph schema to construct a cross-border transaction knowledge graph.
[0050] The following table is a statistical table of entity relationships of a cross-border transaction knowledge graph in an embodiment of this specification:
[0051] Table 1 Instance entity relationship statistics of cross-border transaction knowledge graph
[0052]
[0053] Table 1 records the 440 million basic data in the two transaction scenarios of cross-border remittance and cross-border remittance, and analyzes the corresponding number of initial entities, initial relationships, associated entities, associated relationships, sub-graph entities and sub-graph relationships. Specifically, suspicious (8,100, 5,000) transactions are accurately identified from the massive cross-border transaction relationships (440 million transaction flows).
[0054] From the perspective of cross-border outward and inward remittance scenarios, more than three times the entities (i.e., from 24,000 to 131,000, from 2,000 to 12,000) and more than three times the relationships (i.e., from 62,000 to 234,000, from 6,800 to 26,000) can be expanded from the original cross-border flow to identify potential risks as much as possible. And through the established filtering conditions, the suspicious groups with risks are accurately screened out from the associated and expanded entities and relationships, retaining about 3% of the entities (from 131,000 to 3,800, from 12,000 to 3,000) and relationships (from 234,000 to 8,100, from 26,000 to 5,000), which can be accurately identified to the greatest extent and reduce the burden of business investigation.
[0055] Step 102, obtaining group attribute features from the cross-border transaction knowledge graph, wherein the group attribute features correspond to a group composed of multiple nodes. In an embodiment of the present specification, multiple nodes in the cross-border transaction knowledge graph may constitute a group, and the cross-border transaction knowledge graph may reflect group attribute features. Among them, group attribute features include, but are not limited to: one or any combination of customer identity features, transaction scale features, address indicator features, judicial information features, account attribute features, etc. Among them, customer identity features include: the number of customers in the group, the average age of customers, etc.; transaction scale features include the total number of transactions in the group, the total transaction amount in the group, etc.; address indicator features include: the number of all MAC addresses used in the group, the number of all IP addresses used in the group, etc.; judicial information features include: the total number of judicial inquiries involving customers in the group, the total number of judicial freezes involving customers in the group, etc.; account attribute features include: the number of days from the earliest account opening time in the group, the proportion of a type of account in the group, etc.
[0056] Step 103: input the group attribute characteristics into a group risk prediction model to obtain a group risk prediction result. The group risk prediction model is obtained by training a classification model based on group attribute sample information and sample risk labels.
[0057] This step uses the trained group risk prediction model to predict the group attribute characteristics input into the model, and predicts the group risk related to the cross-border transaction behavior between multiple nodes. For example, the group risk prediction model predicts that the cross-border transaction behavior of a group A composed of multiple nodes on February 18, 2024 is risky; for another example, the group risk prediction model predicts that the cross-border transaction behavior of a group B composed of multiple nodes between March 14 and March 15, 2024 is normal and risk-free.
[0058] In the embodiments of this specification, the group risk prediction model is obtained by training the classification model based on the group attribute sample characteristics and the pre-labeled sample risk labels. For details on the process of training the classification model, see Figure 2 describe.
[0059] In the embodiments of this specification, different scoring levels can be set according to the prediction results of the group risk prediction model. For example, the first scoring level is set to 0.8 and the second scoring level is set to 0.6. When the score corresponding to the group risk prediction probability output by the group risk prediction model is higher than 0.8, the group risk is considered to be high risk; if the corresponding score is between 0.6 and 0.8, the group risk is considered to be medium risk; if the corresponding score is less than 0.6, the group risk is considered to be low risk. This specification can adjust the number of scoring levels and the specific scores of each scoring level according to business needs, and this application does not limit this.
[0060] In the embodiment of this specification, according to the prediction results of the group risk prediction model, the group risk prediction probability is obtained, and according to the preset scoring levels: high risk (>0.8 points), medium risk (0.6-0.8), and low risk (0-0.6), the data distribution of the group risk quantity shown in Table 2 is obtained for the two transaction scenarios of cross-border remittance and cross-border remittance as shown in Table 2:
[0061] Table 2 Data distribution of group risk quantity
[0062] Scene Classification Number of groups High risk Medium risk Low risk Cross-border remittance 24,000 17 78 24480 Cross-border remittance 20,000 2 14 2004
[0063] Step 104: If the group risk prediction result indicates that the group is at risk, the risk level of each node in the risk group is determined using a centrality algorithm and graph features in the cross-border transaction knowledge graph.
[0064] If it is determined based on the above steps that the group currently being studied has transaction risks, the centrality algorithm and the graph features in the knowledge graph are used to determine the risk level or importance of each node in the group with transaction risks. In the embodiments of this specification, the risk level can be further divided into different risk levels according to the size of the risk level value. For example, primary risk, secondary risk, tertiary risk, etc.
[0065] In the embodiments of this specification, the centrality algorithm includes calculating degree centrality, betweenness centrality, closeness centrality, etc. Correspondingly, the graph features in the cross-border transaction knowledge graph include but are not limited to: node degree, node out-degree, node in-degree, etc. Among them, the degree of a node represents the number of edges connected to the node; the out-degree of a node indicates the sum of the number of times a node in a directed graph is the end point of an edge in the graph; the in-degree of a node indicates the sum of the number of times a node in a directed graph is the starting point of an edge in the graph.
[0066] In this manual, the group risk prediction model is used to quickly identify whether there are risk factors in the account group displayed in the cross-border transaction knowledge graph, which helps to quickly screen and identify whether the group account is suspected of gang crime. It is important for further judging the risk nodes in the group account and the importance or risk level of the key nodes that act as "bridges".
[0067] Figure 2 The flowchart of a method for training a group risk prediction model according to an embodiment of the present specification is shown, which specifically includes the following steps:
[0068] Step 201, according to the cross-border transaction sample knowledge graph of the sample node, obtain the group attribute sample characteristics and sample risk labels. In this step, by studying the transactions of multiple sample nodes in history, the cross-border transaction sample knowledge graph corresponding to multiple sample nodes is constructed using the same knowledge graph construction method as in step 101. Further, the group attribute sample characteristics are obtained from the cross-border transaction sample knowledge graph. In addition, the risk labels are marked for the samples through the transaction behaviors of the sample nodes. For example, if the total number of times one or several sample nodes have been involved in judicial inquiries exceeds 5 times, and the total number of times one or several sample nodes have been involved in judicial freezing exceeds 2 times, then the risk label is marked for this one or several sample nodes; for another example, if a node has a record of illegal reporting in the bank, the risk label is marked for the node, which is recorded as "1", otherwise it is recorded as "0"; for another example, if the customer identity characteristics, transaction scale characteristics, address indicator characteristics, judicial information characteristics, and account attribute characteristics of one or several sample nodes are all within the normal range, then the risk-free label is marked for this one or several sample nodes.
[0069] Step 202, input the group attribute sample features into the classification model to obtain the initial prediction results output by the classification model. In this step, the XGBoost algorithm is used to fit the input group attribute sample features, and the performance of the algorithm model is measured using indicators such as precision, recall, F1-Score, and AUC value.
[0070] Step 203, iteratively update the model parameters of the classification model according to the loss value between the initial prediction result and the sample risk label to obtain the group risk prediction model. In this step, the prediction result generated after each iteration of the classification model is used to calculate the difference or loss value with the sample risk label to determine the distance between the prediction result and the sample risk label. When the loss value or difference is less than a preset range, or the number of iterations of the classification model reaches a preset threshold, the classification model training is determined to be completed, and the group risk prediction model is obtained.
[0071] In this manual, for the constructed cross-border transaction knowledge graph, a group risk prediction model is used to mine the importance of different nodes (accounts) within the knowledge graph, that is, to identify which of the multiple nodes are key nodes (also called risk group subgraphs) and further investigate the relevant circumstances of their risk activities.
[0072] Figure 3 The flowchart of a method for determining the risk level of each node in a risk group according to an embodiment of the present specification specifically includes the following steps:
[0073] Step 301, calculate the graph features of each node in the cross-border transaction knowledge graph according to the number of nodes and the number of edges in the cross-border transaction knowledge graph, wherein the graph features include degree, out-degree and in-degree. This step is performed after using the group risk prediction model to predict the risk of the cross-border transaction knowledge graph and determining that the group in the cross-border transaction knowledge graph has transaction risks. In the embodiment of this specification, the group in the cross-border transaction knowledge graph with transaction risks can be regarded as a risk team or "gang". For each node within the risk team, the corresponding graph index can be calculated, and the values of the calculated graph index can be sorted in order from large to small, so as to determine the importance ranking of different nodes under the current graph index.
[0074] In the embodiments of this specification, the graph indicators of the nodes include, but are not limited to, degree centrality, closeness centrality, betweenness centrality, etc. In addition, the graph indicators also include eigenvector centrality, pagerank centrality, etc.
[0075] In the embodiments of this specification, degree centrality, closeness centrality and betweenness centrality are all used to indicate the importance of a node. Centrality is used to measure the centrality of a node in the entire network graph. To distinguish, degree centrality indicates the number of relationships on each node; closeness centrality indicates the closeness between a node and other nodes in the network; betweenness centrality indicates the number of times a vertex appears on the shortest path between any two other vertex pairs.
[0076] Step 302, determine the graph index of each node according to the graph features and centrality algorithm. In this step, the graph features include but are not limited to: node degree, out-degree, in-degree, paths between nodes, shortest paths between nodes, etc. Among them, degree represents the number of edges connected to the node; out-degree represents the sum of the number of times a node in a directed graph is the end point of the edge in the graph; in-degree represents the sum of the number of times a node in a directed graph is the starting point of the edge in the graph.
[0077] According to one or more graph features, at least one graph index for each node can be obtained. Graph indexes include but are not limited to: degree centrality, closeness centrality, betweenness centrality, etc. Specifically, the degree centrality of a node can be determined based on the out-degree and in-degree of the node; the closeness centrality of a node can be determined based on the shortest distance and the shortest average path between two nodes; the betweenness centrality of the node can be determined based on the number of shortest paths between two nodes and the number of nodes through which the shortest path between the nodes passes.
[0078] Step 303: Determine the risk level of each node in the group according to the graph index of each node.
[0079] In this step, the risk level of the node in the group can be determined based on a single graph indicator; or the risk level of the node in the group can be determined comprehensively by combining multiple graph indicators.
[0080] In some embodiments of the present specification, the values of the same graph index of each node can be uniformly calculated, and the values of each node under the graph index can be sorted in descending order. In this step, a preset proportion of nodes with graph index values ranked high can be selected to determine that these nodes are at risk. The preset proportion can be 20%, 25%, 30% or other proportions.
[0081] In some other embodiments of the present specification, the size of each graph index of each node can be compared with the risk threshold. If at least one of the graph indexes is greater than the risk threshold, it is determined that the node is at risk. The graph indexes of all nodes in the graph are sorted from high to low. If the value of at least one graph index of a node is high and exceeds the preset threshold, it means that the node is at risk. If the values of all graph indexes of a node are within the normal range, it means that the risk level of the node is low.
[0082] In the embodiments of the present specification, the risk level of the node is determined by comparing the size of each graph indicator of each node with the risk threshold. Among them, if there is at least one indicator in the graph indicators that is greater than the risk threshold, it is determined that the node is at risk. Furthermore, the risk threshold can be set to different levels, such as the first-level risk threshold, the second-level risk threshold, the third-level risk threshold, etc. The risk threshold can be set to correspond to the level of the risk level. For example, the first-level risk threshold corresponds to the first-level risk, the second-level risk threshold corresponds to the second-level risk, and the third-level risk threshold corresponds to the third-level risk. Compare each graph indicator with the risk threshold of different levels to determine which level of risk threshold the value of the graph indicator falls within, thereby determining the risk level of the node. For example, if the value of the betweenness centrality of a node is within the second-level risk threshold, the risk level of the node is determined to be a second-level risk.
[0083] Furthermore, this specification also needs to combine transaction data to analyze the importance of nodes in the group reflected by different graph indicators. For example, in a transaction relationship network of a risk group, a node has a high intermediary centrality. Combined with its transactions, the node may be a high-risk account belonging to the "owner", and a large amount of funds are collected or split through these accounts.
[0084] For example, a node with a high degree centrality may represent that the node is a core member or a key role in the risk group. Combined with transactions, these nodes may be organizers of risky activities, transit points for fund transfers, or people who have close ties with other risky nodes.
[0085] In the embodiments of this specification, the degree centrality of each node is determined based on the out-degree of each node, the in-degree of each node, and the number of nodes in the risk group. Degree centrality is used to measure the number of relationships on each node, thereby characterizing the importance of the node. Specifically, the degree centrality of a node can be determined by the following formula:
[0086] in, represents the in-degree of the ith node, represents the out-degree of the ith node, v i It represents the degree of the ith node in the subgraph, and n represents the number of nodes in the group. When the value of degree centrality exceeds the preset threshold, it means that the node is at risk and is a "key figure" in the group, which can guide further practical investigation.
[0087] In the embodiment of this specification, the shortest distance and the shortest average path between two nodes determine the closeness centrality of a node. Specifically, it can be determined according to the following formula: Among them, C c (v i ) represents the i-th node v i The closeness centrality, I ij represents the i-th node v i and the jth node v j The shortest distance between Represents node v i The average shortest path between a node and all other nodes, where n is the number of nodes in the population. In this specification, a node v i The smaller the average shortest path between all other nodes, the higher the proximity centrality of the node. When the value of proximity centrality exceeds the preset threshold, it means that the node is at risk and is a "key figure" in the group, which can guide further practical investigation.
[0088] In the embodiment of this specification, the betweenness centrality of a node is determined based on the number of shortest paths between two nodes and the number of nodes through which the shortest path between the nodes passes. The number of shortest paths passing through a node is used to characterize the importance index of a node.
[0089] If a member is located on multiple shortest paths to other members, then the member is a core member and has a large betweenness centrality. Specifically, the betweenness centrality of a node is determined according to the following formula:
[0090]
[0091] Among them, C b (v i ) represents the i-th node v i The betweenness centrality, N jk Represents node v i and node v j The total number of shortest paths between jk (i) represents node v i and node v j The shortest path between i The betweenness centrality refers to the ratio of the shortest path that passes through a point and connects two points in the network to the total number of shortest paths between the two points. When the betweenness centrality value of a node exceeds the preset threshold, it means that the node is risky and is a key node in the group. A large amount of funds may be split or circulated at the node, and further investigation is necessary.
[0092] In this specification, the risk characteristics of typical cases are further analyzed based on the preset suspicious scores. Specifically, the analysis includes the following table 3:
[0093] Table 3 Analysis of case characteristics of cross-border split remittance scenarios
[0094] statistics Case 1 Case 2 Case 3 Verify account 4201078XXXX 977496XXXX 685684XXXX Cross-border transaction amount 1399411 1739240.7 2652684.5 Number of cross-border trading counterparties 5 5 7 Number of cross-border transactions 5 13 8 The number of domestic accounts with cross-border transactions greater than 300,000 4 4 6 Number of subgraph entities 10 26 29 Number of subgraph relationships 22 177 91 Same number of physical branches 8 7 21 The same number of physical devices 4 3 3 Number of entities with the same place of ownership 0 3 7 … … … … Suspicious Rating 0.864 0.941 0.928
[0095] As can be seen from the table, in cases 1 and 2, the cross-border transaction amounts are large, there are many cross-border transaction counterparties, and there is a suspicious risk of shared equipment among the transaction counterparts, which shows the characteristics of multiple accounts using the same equipment to split remittances; while in case 3, the cross-border transaction amount is large, there are many cross-border transaction counterparties, and the number of branches with the same account entity opening is greater than 20, which shows the characteristics of multiple accounts in the same opening branch using the same equipment to split remittances.
[0096] In the embodiment of this specification, the internal nodes in a cross-border transaction knowledge graph are mined using a graph algorithm, and the results shown in Table 4 below can be obtained:
[0097] Table 4 Graph indicators and graph features of cross-border transaction knowledge graph
[0098] Various indicators Node 1 Node 2 … Node 11 out degree 8 2 … 15 degree 3 10 … 12 Degree Centrality 0.11 0.12 … 0.27 Closeness Centrality 0.133 0.008 … 0.194 … … … … … Betweenness Centrality 0.257 0.002 … 0.107
[0099] From this, we can analyze that the various centralities of node 11 are high, so this node may be of high importance in the risk group; the betweenness centrality of node 1 is high, and combined with the out-degree and in-degree, this node may mainly play a bridging role in the risk group, that is, a "transition account"; the various centrality indicators of node 2 are low, but the in-degree is high. Combined with the subgraph, this node is likely to be a fund collection account.
[0100] Figure 4 The flowchart of a method for constructing a cross-border transaction knowledge graph in an embodiment of this specification is shown, which specifically includes the following steps:
[0101] Step 401, determine the account information and the first transaction time of the domestic first-level node that has transactions with the overseas account. In this step, extract the cross-border remittance transactions within the first preset time period, take the overseas collection node or the overseas remittance node as the leaf node, and take the cross-border transaction as the first-level edge, wherein the first preset time period can be determined according to actual business needs, for example, 365 days, 180 days, 90 days, etc.
[0102] According to the domestic node that has a transaction with the leaf node (i.e., the overseas payment collection node or the overseas remittance node) within a preset time period, the domestic node is determined as a domestic first-level node. The account information of the domestic first-level node and the time when the domestic first-level node and the overseas payment collection node or the overseas remittance node traded are obtained, and the time is determined as the first transaction time. In the embodiment of this specification, the number of domestic first-level nodes can be at least one.
[0103] Step 402, based on the transaction behavior of the domestic first-level node with other domestic nodes within the preset time range of the first transaction time, determine the account information and second transaction time of the domestic second-level node. In this step, taking the second transaction time of each transaction of the domestic first-level node in the country and the transaction account of the domestic first-level node as the starting point, select the second preset time period before and after the transaction time as the time window for all transaction records of the transfer of the domestic account, and group and count them according to the account numbers of the two parties to the transaction. Among them, the second preset time period can be determined according to actual business needs, for example, 7 days, 15 days, 30 days, etc.
[0104] Step 403, associate and expand the transactions of the domestic secondary nodes within the preset time range of the second transaction time, and determine the multi-level transaction nodes and transaction time.
[0105] In this step, the earliest transaction time, the latest transaction time and the target account (remittance account / remittance account) of all transaction information in each account of the domestic secondary node are taken as the starting point, and all transaction records of transfers to the account of the domestic secondary node within the third preset time period before the earliest transaction time and the third preset time period after the latest time are selected, and grouped and counted according to the accounts of both parties to the transaction to determine the domestic third-level node. Among them, the third preset time period can be determined according to actual business needs, for example, 7 days, 15 days, 30 days, etc. In accordance with the aforementioned steps, the domestic third-level nodes that have transactions with the domestic second-level nodes, the domestic fourth-level nodes that have transactions with the domestic third-level nodes, etc. are expanded, so as to respectively determine the transaction time, transaction amount and other information of the multi-level transaction nodes and each transaction node.
[0106] Step 404, the transaction behaviors between nodes are used as edges and the accounts of each node are used as entities to construct a cross-border transaction knowledge graph. In the embodiment of this specification, since cross-border transactions are an extension of the dynamic time window, there are duplicate transactions, so when summarizing the transaction results of the associated extension, it is necessary to deduplicate the transactions.
[0107] Specifically, according to the description in step 401 to step 403, a cross-border transaction knowledge graph is constructed with four layers of edges, and for each point in the cross-border transaction knowledge graph, the cumulative amount of transfers in, the cumulative amount of transfers out, the cumulative number of transfers in, the cumulative number of transfers out, the cumulative number of transfers in, the cumulative number of transfers out, and other information are respectively counted. Further, points in the cross-border transaction knowledge graph with an in-degree or out-degree greater than or equal to N accounts, or points with a unilateral cumulative amount greater than RMB M, can be marked. Among them, according to actual business needs and data distribution, N can be determined as a value such as 3, 5, 10, etc.
[0108] like Figure 5 The figure shows a schematic diagram of the structure of a cross-border transaction risk node identification device based on a knowledge graph in an embodiment of the present specification. In this figure, the basic structure of the cross-border transaction risk node identification device based on a knowledge graph is described. The functional units and modules therein can be implemented in software, or a general chip or a specific chip can be used to implement cross-border transaction risk node identification based on a knowledge graph. The device specifically includes:
[0109] A graph construction unit 501 is used to construct a cross-border transaction knowledge graph based on cross-border transaction behaviors between multiple nodes, wherein the cross-border transaction knowledge graph is constructed with transactions between nodes as relationships and accounts of each node as entities;
[0110] An attribute acquisition unit 502, configured to acquire a group attribute feature from the cross-border transaction knowledge graph, wherein the group attribute feature corresponds to a group consisting of a plurality of nodes;
[0111] A prediction result determination unit 503 is used to input the group attribute characteristics into a group risk prediction model to obtain a group risk prediction result, wherein the group risk prediction model is obtained by training a classification model based on group attribute sample characteristics and sample risk labels;
[0112] The risk level determination unit 504 is used to determine the risk level of each node in the risk group by using the centrality algorithm and the graph features in the cross-border transaction knowledge graph if the group risk prediction result indicates that the group is at risk.
[0113] Figure 6 The figure shows a schematic diagram of a transaction knowledge graph in a cross-border split remittance scenario in an embodiment of this specification. The figure shows overseas nodes and three layers of nodes in China, a total of four layers of nodes. The three layers of nodes in China are marked as node 1, node 2, and node 3, respectively, and node 1, node 2, and node 3 all represent multiple nodes.
[0114] like Figure 7 The diagram shows a schematic diagram of a computer device provided by an embodiment of the present specification. The cross-border transaction risk node identification method based on the knowledge graph described in the present specification can be executed by the computer device. The computer device 702 may include one or more processors 704, such as one or more central processing units (CPUs), each of which may implement one or more hardware threads. The computer device 702 may also include any memory 706 for storing any kind of information such as code, settings, data, etc. Non-limitingly, for example, the memory 706 may include any one or more combinations of the following: any type of RAM, any type of ROM, flash memory device, hard disk, optical disk, etc. More generally, any memory may use any technology to store information. Further, any memory may provide volatile or non-volatile retention of information. Further, any memory may represent a fixed or removable component of the computer device 702. In one case, when the processor 704 executes an associated instruction stored in any memory or combination of memories, the computer device 702 may perform any operation of the associated instruction. The computer device 702 also includes one or more drive mechanisms 708 for interacting with any storage, such as a hard disk drive mechanism, an optical disk drive mechanism, etc.
[0115] The computer device 702 may also include an input / output module 710 (I / O) for receiving various inputs (via input devices 712) and for providing various outputs (via output devices 714). A specific output mechanism may include a presentation device 716 and an associated graphical user interface (GUI) 718. In other embodiments, the input / output module 710 (I / O), the input device 712, and the output device 714 may not be included, and the computer device 702 may be used as a computer device in a network. The computer device 702 may also include one or more network interfaces 720 for exchanging data with other devices via one or more communication links 722. One or more communication buses 724 couple the components described above together.
[0116] The communication link 722 may be implemented in any manner, for example, through a local area network, a wide area network (e.g., the Internet), a point-to-point connection, etc., or any combination thereof. The communication link 722 may include any combination of hardwired links, wireless links, routers, gateway functions, name servers, etc. governed by any protocol or combination of protocols.
[0117] Corresponds to Figures 1 to 4 The method in the embodiment of the present specification also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are executed.
[0118] The embodiment of the present specification also provides a computer-readable instruction, wherein when the processor executes the instruction, the program therein causes the processor to execute the following Figures 1 to 4 The method shown.
[0119] The embodiment of the present specification also provides a computer program product, which includes a computer program. When the computer program is executed by a processor, Figures 1 to 4 The method shown.
[0120] It should be understood that in the various embodiments of this specification, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this specification.
[0121] It should also be understood that in the embodiments of this specification, the term "and / or" is only a description of the association relationship of the associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this specification generally indicates that the associated objects before and after are in an "or" relationship.
[0122] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed in this specification can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this specification.
[0123] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.
[0124] In the several embodiments provided in this specification, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or it can be an electrical, mechanical or other form of connection.
[0125] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of this specification.
[0126] In addition, each functional unit in each embodiment of this specification may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units.
[0127] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this specification is essentially 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, and the computer software product is stored in a storage medium, including a number of instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of this specification. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.
[0128] Specific embodiments are used in this specification to illustrate the principles and implementation methods of this specification. The description of the above embodiments is only used to help understand the methods and core ideas of this specification. At the same time, for those skilled in the art, according to the ideas of this specification, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on this specification.
Claims
1. A method for identifying risk nodes in cross-border transactions based on knowledge graph, characterized in that: The method comprises: According to the cross-border transaction behaviors between multiple nodes, a cross-border transaction knowledge graph is constructed, wherein the cross-border transaction knowledge graph is constructed with the transactions between the nodes as relationships and the accounts of the nodes as entities; Acquire a group attribute feature from the cross-border transaction knowledge graph, wherein the group attribute feature corresponds to a group consisting of a plurality of nodes; Inputting the group attribute characteristics into a group risk prediction model to obtain a group risk prediction result, wherein the group risk prediction model is obtained by training a classification model based on group attribute sample characteristics and sample risk labels; If the group risk prediction result indicates that the group is at risk, the risk level of each node in the risk group is determined using the centrality algorithm and the graph features in the cross-border transaction knowledge graph.
2. The cross-border transaction risk node identification method based on knowledge graph according to claim 1 is characterized in that: The group risk prediction model is trained in the following way: According to the cross-border transaction sample knowledge graph of the sample node, obtain the group attribute sample characteristics and sample risk labels; Inputting the sample characteristics of the group attributes into the classification model to obtain an initial prediction result output by the classification model; According to the loss value between the initial prediction result and the sample risk label, the model parameters of the classification model are iteratively updated to obtain the group risk prediction model.
3. The cross-border transaction risk node identification method based on knowledge graph according to claim 2 is characterized in that: Using the centrality algorithm and the graph features in the cross-border transaction knowledge graph, the risk level of each node in the risk group is determined, including: Calculate the graph features of each node in the cross-border transaction knowledge graph according to the number of nodes and the number of edges in the cross-border transaction knowledge graph; Determine the graph index of each node according to the graph characteristics and centrality algorithm; According to the graph index of each node, the risk level of each node in the group is determined.
4. The cross-border transaction risk node identification method based on knowledge graph according to claim 3 is characterized in that: The graph indicators of each node determined according to the graph characteristics and centrality algorithm include: According to the out-degree of each node, the in-degree of each node and the number of nodes in the risk group, the degree centrality of each node is determined: Among them, C d (v i ) represents the degree centrality of the ith node, represents the in-degree of the ith node, represents the out-degree of the ith node, v i represents the degree of the ith node in the subgraph, and n represents the number of nodes in the population.
5. The cross-border transaction risk node identification method based on knowledge graph according to claim 3 is characterized in that: The graph indicators of each node determined according to the graph characteristics and centrality algorithm include: The closeness centrality is determined according to the following formula: Among them, C c (v i ) represents the i-th node v i The closeness centrality, I ij represents the i-th node v i and the jth node v j The shortest distance between Represents node v i The average shortest path between all other nodes, n is the number of nodes in the group.
6. The cross-border transaction risk node identification method based on knowledge graph according to claim 5 is characterized in that: The graph indicators of each node determined according to the graph characteristics and centrality algorithm include: The betweenness centrality is determined according to the following formula: Among them, C b (v i ) represents the i-th node v i The betweenness centrality, N jk Represents node v i and node v j The number of shortest paths between jk (i) represents node v i and node v j The shortest path between i The number of .
7. The cross-border transaction risk node identification method based on knowledge graph according to claim 1 is characterized in that: The cross-border transaction knowledge graph is constructed in the following way: Determine the account information and first transaction time of the domestic first-level node that conducts transactions with overseas accounts; Determine the account information and second trading time of the domestic second-level node based on the trading behavior of the domestic first-level node with other domestic nodes within the preset time range of the first trading time; Associating and expanding the transactions of the domestic secondary nodes within the preset time range of the second transaction time, and determining the multi-level transaction nodes and transaction time; The transaction behaviors between nodes are taken as edges and the accounts of each node are taken as entities to construct a cross-border transaction knowledge graph.
8. A cross-border transaction risk node identification device based on knowledge graph, characterized in that: The device comprises: A graph construction unit, used to construct a cross-border transaction knowledge graph based on cross-border transaction behaviors between multiple nodes, wherein the cross-border transaction knowledge graph is constructed with transactions between nodes as relationships and accounts of each node as entities; An attribute acquisition unit, used to acquire group attribute features from the cross-border transaction knowledge graph, wherein the group attribute features correspond to a group composed of a plurality of nodes; A prediction result determination unit, used for inputting the group attribute characteristics into a group risk prediction model to obtain a group risk prediction result, wherein the group risk prediction model is obtained by training a classification model based on group attribute sample characteristics and sample risk labels; The risk level determination unit is used to determine the risk level of each node in the risk group by using the centrality algorithm and the graph features in the cross-border transaction knowledge graph if the group risk prediction result indicates that the group is at risk.
9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
11. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.
Citation Information
Cited By
Data cross-border abnormal behavior detection method and system based on knowledge graph
CN120567587A
