Methods, devices, electronic equipment, storage media, and program products for identifying abnormal transaction objects
By constructing a curvature relationship graph of trading objects and using local neighborhood density to identify abnormal trading objects, the problem of low risk identification accuracy and insufficient contextual association in existing technologies is solved, and efficient risk screening and location are achieved.
Patent Information
- Application Number
- CN202611481913.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-22
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies rely on keyword matching and single threshold comparison when identifying the risks of trading objects, resulting in low accuracy in risk identification, inability to establish contextual relationships between multiple trading objects, and difficulty in accurately locating abnormal objects.
By acquiring the feature vectors of multiple trading objects, a curvature relationship graph is constructed, the target curvature of the changing vector sequence is calculated, and abnormal trading objects are identified by utilizing local neighborhood density, thus establishing contextual relationships between trading objects.
It improves the accuracy of identifying abnormal transaction targets and the efficiency of risk screening, and can accurately pinpoint the time and cause of abnormal behavior.
Smart Images

Figure CN123118259A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of big data technology, specifically to a method, apparatus, electronic device, storage medium, and program product for identifying abnormal transaction objects. Background Technology
[0002] Currently, after a transaction is completed by a company or individual, financial institutions such as banks usually conduct a period of observation on the transaction counterparty to identify any potential risks and take corresponding risk control measures.
[0003] Currently, relevant technologies primarily rely on two methods to determine risk. One is text matching using preset keywords; when risk-related keywords appear in the text related to a trading entity, a risk is identified. The other is setting a fixed threshold for a single feature indicator; when the feature value exceeds or falls below the threshold, a risk warning is triggered. However, keyword matching cannot identify risk expressions not predefined in the rule base, and single threshold comparison cannot comprehensively assess information from multiple dimensions, resulting in low accuracy in risk identification. Moreover, when a risk warning is triggered, the system cannot trace the specific time point and data source that led to the risk assessment result, making risk screening inefficient. Furthermore, when a transaction involves multiple trading entities, the relevant technologies can only analyze each entity independently, failing to establish contextual relationships between multiple entities and making it difficult to accurately locate the specific entity where the problem occurred. Summary of the Invention
[0004] In view of the above problems, this application provides a method, apparatus, electronic device, storage medium and program product for identifying abnormal transaction objects.
[0005] According to the first aspect of this application, a method for determining abnormal transaction objects is provided, comprising: obtaining multiple feature vectors corresponding to multiple time points within a preset time period after multiple transaction objects occur; sorting the multiple transaction objects according to time, taking the sum of the feature vectors of adjacent transaction objects at each time point as the change vector corresponding to the time point, and concatenating the multiple change vectors corresponding to multiple time points sequentially according to time to obtain a change vector sequence, and calculating the target curvature of the change vector sequence; constructing a curvature relationship graph based on multiple transaction objects and the target curvature, wherein the curvature relationship graph uses transaction objects as nodes, and determines the target distance between adjacent nodes in the curvature relationship graph based on the target curvature corresponding to adjacent transaction objects, and the target curvature is positively correlated with the target distance; determining a search area with nodes in the curvature relationship graph as centers according to a preset search radius, and taking the number of other nodes in the search area as the local neighborhood density corresponding to the node; determining a target node from the multiple nodes contained in the curvature relationship graph based on the local neighborhood density, and determining the transaction object corresponding to the target node as an abnormal transaction object.
[0006] According to an embodiment of this application, obtaining multiple feature vectors corresponding to multiple time points within a preset time period after multiple trading objects have engaged in the same trading activity includes: obtaining the original text data of the multiple trading objects within a preset time period after engaging in the trading activity; dividing the original text data into blocks according to multiple time points within the preset time period to obtain text blocks corresponding to each time point; inputting the text blocks into a preset feature vector library for retrieval to obtain multiple associated feature vectors that are semantically related to the text blocks; and fusing the multiple associated feature vectors to obtain the feature vector corresponding to the time point.
[0007] According to an embodiment of this application, calculating the target curvature of a change vector sequence includes: obtaining multiple continuous change vector groups in the change vector sequence, each continuous change vector group including at least three continuous change vectors; determining multiple initial curvature values based on the multiple continuous change vector groups, and using the average of the multiple initial curvature values as the target curvature of the change vector sequence.
[0008] According to an embodiment of this application, the target curvature of the changing vector sequence is calculated, and multiple initial curvature values are determined based on multiple continuous changing vector groups, including: constructing a target arc based on three continuous changing vectors; and determining the reciprocal of the radius of the target arc as the initial curvature value corresponding to the continuous changing vector group.
[0009] According to an embodiment of this application, determining the target distance between adjacent nodes in the curvature relationship graph based on the target curvature corresponding to adjacent trading objects includes: determining the repulsion strength between adjacent nodes corresponding to adjacent trading objects based on the target curvature corresponding to adjacent trading objects, wherein the target curvature and the repulsion strength are positively correlated; and using the product of the repulsion strength between adjacent nodes and the preset distance as the target distance between adjacent nodes.
[0010] According to an embodiment of this application, after determining the target node from multiple nodes contained in the curvature relationship graph, the method further includes: in the curvature relationship graph, determining at least one node adjacent to the target node; determining the target curvature corresponding to the target node and the at least one adjacent node as the abnormal curvature; determining three consecutive change vectors corresponding to the abnormal curvature, and determining the change vector located in the middle position among the three consecutive change vectors as the abnormal vector; determining the time point corresponding to the abnormal vector as the time point when the abnormal transaction object exhibits abnormal behavior within a preset time period.
[0011] The second aspect of this application provides an apparatus for determining abnormal transaction objects, comprising: an acquisition module for acquiring multiple feature vectors corresponding to multiple time points within a preset time period after the same transaction occurs for multiple transaction objects; a processing module for sorting the multiple transaction objects according to time, taking the sum of the feature vectors of adjacent transaction objects at each time point as the change vector corresponding to the time point, and concatenating the multiple change vectors corresponding to multiple time points sequentially according to time to obtain a change vector sequence, and calculating the target curvature of the change vector sequence; a construction module for constructing a curvature relationship graph based on multiple transaction objects and the target curvature, wherein the curvature relationship graph uses transaction objects as nodes, and determines the target distance between adjacent nodes in the curvature relationship graph based on the target curvature corresponding to adjacent transaction objects, and the target curvature is positively correlated with the target distance; a first determination module for determining a search area with nodes in the curvature relationship graph as centers and according to a preset search radius, and taking the number of other nodes in the search area as the local neighborhood density corresponding to the node; and a second determination module for determining a target node from the multiple nodes contained in the curvature relationship graph based on the local neighborhood density, and determining the transaction object corresponding to the target node as an abnormal transaction object.
[0012] A third aspect of this application provides an electronic device comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method described above.
[0013] A fourth aspect of this application also provides a computer-readable storage medium having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the above-described method.
[0014] The fifth aspect of this application also provides a computer program product, including a computer program or instructions that, when executed by a processor, implement the steps of the above-described method. Attached Figure Description
[0015] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:
[0016] Figure 1 The illustrations depict application scenarios of the method, apparatus, electronic device, storage medium, and program product for determining abnormal transaction objects according to embodiments of this application.
[0017] Figure 2 A flowchart illustrating a method for determining abnormal transaction objects according to an embodiment of this application is shown schematically;
[0018] Figure 3 A schematic diagram of curvature relationships according to an embodiment of this application is shown.
[0019] Figure 4 This schematic diagram illustrates the structural block diagram of a device for determining abnormal transaction objects according to an embodiment of this application;
[0020] Figure 5 A block diagram schematically illustrates an electronic device suitable for implementing a method for determining abnormal transaction objects according to an embodiment of this application. Detailed Implementation
[0021] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.
[0022] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.
[0023] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.
[0024] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).
[0025] It should be noted that in the embodiments of this application, certain software, components, models and other existing solutions in the industry may be mentioned. These should be regarded as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. However, it does not mean that the applicant has used or necessarily used the solution.
[0026] In the technical solution of this application, the user information (including but not limited to user personal information, user image information, user device information, such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved 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 relevant laws, regulations, and standards, and necessary measures have been taken to ensure that they do not violate public order and good morals. Corresponding operation entry points are provided for users to choose to authorize or refuse.
[0027] In scenarios where personal information is used for automated decision-making, the methods, devices, and systems provided in this disclosure all provide users with corresponding operation entry points for users to choose to agree to or reject the automated decision results; if the user chooses to reject, the process proceeds to the expert decision-making process.
[0028] Figure 1 The illustration shows an application scenario diagram of the method for determining abnormal transaction objects according to an embodiment of this application.
[0029] like Figure 1 As shown, application scenario 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 serves as a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables.
[0030] Users can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 via the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social media platform software, etc. (for example only).
[0031] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with displays and support web browsing, including but not limited to smartphones, tablets, laptops, and desktop computers.
[0032] Server 105 can be a server that provides various services, such as a backend management server that supports websites browsed by users using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (this is just an example). The backend management server can analyze and process data such as received user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal devices.
[0033] It should be noted that the method for determining abnormal transaction objects provided in this application embodiment can generally be executed by server 105. Correspondingly, the apparatus for determining abnormal transaction objects provided in this application embodiment can generally be located in server 105. The method for determining abnormal transaction objects provided in this application embodiment can also be executed by a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105. Correspondingly, the apparatus for determining abnormal transaction objects provided in this application embodiment can also be located in a server or server cluster that is different from server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or server 105.
[0034] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.
[0035] The following will be based on Figure 1 The described scene, through Figures 2-3 The method for determining abnormal transaction objects according to the disclosed embodiments is described in detail.
[0036] Figure 2 A flowchart illustrating a method for determining abnormal transaction objects according to an embodiment of this application is shown schematically.
[0037] like Figure 2 As shown, the method 200 for determining abnormal transaction objects in this embodiment includes steps S210 to S250.
[0038] In step S210, multiple feature vectors corresponding to multiple time points within a preset time period after the same transaction occurs for multiple transaction objects are obtained.
[0039] For example, a transaction counterparty refers to the various entities participating in the same transaction. In a financial product transaction scenario, the transaction counterparty may include the buyer company, its legal representative, or its actual controller. A feature vector is the result of digitally representing the state of a transaction counterparty at a specific point in time. This feature vector may contain numerical features obtained by converting the textual description information of the transaction counterparty. A preset time period refers to a time window for continuous observation of the transaction counterparty after the transaction occurs, such as 12 months or 6 months after the transaction is completed. Multiple time points refer to multiple observation points set within the preset time period, such as obtaining the feature vector for each month to capture the trend of the transaction counterparty's state changing over time.
[0040] In step S220, multiple transaction objects are sorted according to time, the sum of the feature vectors of adjacent transaction objects at each time point is taken as the change vector corresponding to the time point, and the multiple change vectors corresponding to multiple time points are concatenated in time to obtain a change vector sequence, and the target curvature of the change vector sequence is calculated.
[0041] For example, sorting multiple trading objects by time means sorting them according to the chronological order of operations performed on each trading object in the trading process.
[0042] Taking two adjacent transaction objects after sorting as an example, assume that transaction object M and transaction object N each have a feature vector V at the same point in time. M and V N Then, adding the two feature vectors together yields the change vector V corresponding to that time point. MN =V M +V N The change vectors reflect the overall state orientation of the two trading objects at that point in time. By concatenating the change vectors corresponding to multiple points in time within a preset time period in chronological order, a change vector sequence is obtained. The target curvature of this change vector sequence is then calculated. Curvature describes the degree of bending of the sequence in space; a greater curvature indicates a more drastic shift in the state change trajectory of the trading objects, meaning a more significant change in the cooperative relationship between the two trading objects during that time period.
[0043] In step S230, a curvature relationship graph is constructed based on multiple transaction objects and target curvature. The curvature relationship graph uses transaction objects as nodes, and the target distance between adjacent nodes in the curvature relationship graph is determined based on the target curvature corresponding to adjacent transaction objects. The target curvature and the target distance are positively correlated.
[0044] For example, such as Figure 3 As shown, based on the sorting results, each pair of adjacent trading objects corresponds to a target curvature, which quantifies the degree of curvature of the trajectory of state change of the pair of trading objects within a preset time period. A curvature relationship graph is constructed using these trading objects as nodes. The target curvature between each pair of adjacent trading objects determines the target distance of the line connecting them, and the target curvature and target distance are positively correlated. That is, the greater the target curvature, the farther apart the two nodes are in the graph. In this way, the abstract relationship between trading objects is mapped to a computable geometric space.
[0045] In step S240, the search area is determined with the nodes in the curvature graph as the center and according to the preset search radius, and the number of other nodes in the search area is used as the local neighborhood density of the node.
[0046] For example, local neighborhood density quantifies the number of other nodes gathered around each node. The lower the local neighborhood density, the farther the node is geometrically from other nodes, and the more its state behavior deviates from the group.
[0047] For example, such as Figure 3 As shown, nodes A and B each define their respective search areas with the same preset search radius. For node A, its search area includes two other nodes besides itself. This means that node A has a large number of neighbors and a high local neighborhood density. Since node A is located near the starting point of the entire trajectory and is close to its surrounding nodes, it indicates that the state change pattern of the transaction object corresponding to node A is similar to that of many surrounding transaction objects, placing it in a normal position within the group. For node B, its search area includes only one other node besides itself. This means that node B has a small number of neighbors and a low local neighborhood density. This indicates that the state change pattern of the transaction object corresponding to node B differs significantly from that of most transaction objects in the group. Since the spatial position of nodes in the curvature relationship graph is determined by the target curvature between transaction objects, nodes with low local neighborhood density are located at the edge or in isolated positions within the group, indicating that the state change trajectory of this transaction object deviates from the mainstream trend of the group within a preset time period after the transaction occurs. This transaction object is therefore an abnormal transaction object.
[0048] In step S250, based on the local neighborhood density, a target node is determined from multiple nodes contained in the curvature relationship graph, and the transaction object corresponding to the target node is determined as an abnormal transaction object.
[0049] For example, based on the local neighborhood density of each node, nodes with a local neighborhood density less than or equal to a density threshold are identified as target nodes, and the trading objects corresponding to the target nodes are identified as abnormal trading objects in the trading behavior. Since the target node is located at the edge or in an isolated position in the curvature relationship graph, if the state change trajectory of its corresponding trading object deviates from the mainstream trend of the group within a preset time period after the trading behavior occurs, and meets certain preset conditions, then the trading object is considered an abnormal trading object.
[0050] For example, in a scenario where multiple trading entities engage in the same transaction, the state change trajectories of most entities remain relatively consistent, exhibiting low curvature and clustering together in the curvature graph, resulting in high local neighborhood density. However, if a particular trading entity's state change trajectory exhibits an anomaly—high curvature—it is pushed out of the group in the curvature graph, resulting in low local neighborhood density. If this density falls below a threshold, the trading entity is more likely to have engaged in risky trading behavior, thus identifying it as an abnormal trading entity. Financial institutions can then focus on and further investigate this entity.
[0051] The above steps, by using the sum of the feature vectors of adjacent trading objects as the change vector and calculating the curvature, integrate the state change information of multiple trading objects at multiple time points into a quantifiable curvature value. Then, the local neighborhood density of nodes in the curvature relationship graph automatically identifies trading objects whose state changes deviate from the group. This approach does not rely on any manually preset risk keywords, overcoming the problem of low risk identification accuracy caused by relying on keyword matching and single threshold comparison, and improving the accuracy of identifying abnormal trading objects. Furthermore, a curvature relationship graph is constructed by ranking and pairing trading objects, establishing contextual associations between multiple trading objects within the same graph. Local neighborhood density accurately identifies specific trading objects whose state behavior deviates from the group, overcoming the problem in existing technologies that cannot establish contextual associations between multiple objects, thus making it difficult to locate specific problem objects, further improving the accuracy and efficiency of risk screening.
[0052] According to the method for determining abnormal transaction objects in this application, multiple feature vectors corresponding to multiple time points within a preset time period after multiple transaction objects have performed the same transaction are obtained. The method includes: obtaining the original text data of multiple transaction objects within a preset time period after performing the transaction; dividing the original text data into blocks according to multiple time points within the preset time period to obtain text blocks corresponding to each time point; inputting the text blocks into a preset feature vector library for retrieval to obtain multiple associated feature vectors that are semantically related to the text blocks; and fusing the multiple associated feature vectors to obtain the feature vector corresponding to the time point.
[0053] For example, raw text data refers to unstructured natural language text obtained from multiple data sources, including both structured and unstructured text data, which undergoes data cleaning after collection. In financial product trading scenarios, raw text data may include industry research reports, official website updates, registration information, and related documents of the trading entity. A text block is a text segment obtained by dividing raw text data into multiple time points within a preset time period, with each time point corresponding to a text block. For example, all relevant raw text of a company in the first month may be divided into text blocks corresponding to that month. A preset feature vector library is a pre-built vector set that stores a large number of text blocks and their corresponding feature vectors, used for semantic-level retrieval and matching of input text blocks. Associated feature vectors refer to one or more feature vectors in the preset feature vector library that are semantically related to the input text block. Retrieval-Augmented Generation (RAG) technology is used to retrieve these vectors from the preset feature vector library. The retrieval process is achieved by calculating the similarity between the text block and each vector in the vector library; vectors with a similarity exceeding a preset similarity threshold are considered associated feature vectors.
[0054] For example, firstly, the raw text data of multiple trading entities within a preset time period after the transaction is completed is acquired. Then, the raw text data is divided into blocks according to multiple time points within the preset time period to obtain text blocks corresponding to each time point. For example, the raw text data within 12 months after the transaction is completed can be divided into 12 text blocks corresponding to 12 time points, using a monthly granularity. Next, the text block corresponding to each time point is input into a preset feature vector library for retrieval (i.e., the retrieval step in the retrieval enhancement generation) to obtain multiple associated feature vectors semantically related to the text block. Finally, the retrieved multiple associated feature vectors are fused, for example, using a weighted average or concatenation, to obtain the feature vector corresponding to that time point. This feature vector serves as the feature vector for the subsequent trading entity at that time point.
[0055] The above steps involve segmenting the original text data according to time points, retrieving and fusing related feature vectors from a pre-defined feature vector library to obtain feature vectors corresponding to each time point. This allows for subsequent curvature calculations and relationship graph construction based on semantic-level feature vectors, eliminating reliance on pre-defined keyword matching. Semantic retrieval of text blocks and the feature vector library can identify risk descriptions that are different from keywords but semantically identical (i.e., generalization ability in retrieval enhancement generation). Fusing multiple related feature vectors can integrate semantic information from multiple sources, thereby improving the accuracy of feature vector representation and ultimately increasing the accuracy of identifying abnormal transaction objects at the overall level.
[0056] According to the method for determining abnormal transaction objects in this application, the target curvature of the change vector sequence is calculated, including: obtaining multiple continuous change vector groups in the change vector sequence, each continuous change vector group including at least three continuous change vectors; determining multiple initial curvature values based on the multiple continuous change vector groups, and using the average of the multiple initial curvature values as the target curvature of the change vector sequence.
[0057] For example, a continuous change vector group refers to a combination of at least three change vectors arranged consecutively in chronological order within a change vector sequence. The initial curvature value is a numerical value calculated based on a continuous change vector group, reflecting the degree of directional change of the change vectors within that group; the more drastic the directional change, the larger the initial curvature value. The target curvature refers to the curvature value obtained by averaging multiple initial curvature values. This average value serves as the final curvature representation of the change vector sequence, indicating the overall degree of curvature of the state change trajectory of adjacent trading objects within a preset time period.
[0058] For example, firstly, multiple continuous change vector groups are obtained from the change vector sequence. Each continuous change vector group includes at least three continuous change vectors. For instance, if the change vector sequence contains five change vectors (V1, V2, V3, V4, V5), then three continuous change vector groups can be obtained: (V1, V2, V3), (V2, V3, V4), and (V3, V4, V5). Then, the corresponding initial curvature value is determined for each continuous change vector group. Since the degree of directional change between the change vectors in each group is different, the calculated initial curvature values will also be different. Finally, the average of the multiple initial curvature values is taken to obtain the target curvature of the change vector sequence. For example, if the three initial curvature values are 0.5, 0.7, and 0.6, then the target curvature of the change vector sequence is (0.5 + 0.7 + 0.6) / 3 = 0.6.
[0059] The above steps divide the changing vector sequence into multiple continuous changing vector groups, calculate the initial curvature value of each group, and take the average value as the target curvature. This makes the calculation of the target curvature no longer dependent on a single group of continuous changing vectors, but integrates the curvature information of multiple continuous changing vector groups. This can smooth out the curvature calculation deviation caused by data fluctuations in individual continuous changing vector groups, thereby improving the stability and accuracy of the target curvature in representing the overall bending degree of the changing vector sequence.
[0060] According to the method for determining abnormal transaction objects in this application, the target curvature of the change vector sequence is calculated, and multiple initial curvature values are determined based on multiple continuous change vector groups, including: constructing a target arc based on three continuous change vectors; and determining the reciprocal of the radius of the target arc as the initial curvature value corresponding to the continuous change vector group.
[0061] For example, in vector space, three consecutive changing vectors correspond to three directed segments in space, and these three directed segments can be fitted into a unique target arc. The radius of this target arc, determined by the three consecutive changing vectors, reflects the degree of drastic change in the directions corresponding to these three vectors. That is, the smaller the radius, the more abrupt the change in direction among the three vectors, and the larger the corresponding initial curvature value. Conversely, the larger the radius, the gentler the change in direction, and the smaller the corresponding initial curvature value.
[0062] It should be noted that in the actual calculation process, three consecutive change vectors can be selected from the continuous change vector group to construct the target arc and determine the initial curvature value. Alternatively, more consecutive change vectors can be selected to determine the initial curvature value by fitting a circle or polynomial fitting. This embodiment is not limited to this.
[0063] The above steps define an arc in vector space using three consecutive change vectors. The radius of this arc directly reflects the drastic change in direction between the change vectors. The reciprocal of the radius is used as the initial curvature value, thus quantifying the degree of geometric curvature into a calculable numerical value. When the direction of state change between trading objects changes drastically within a certain time interval, the radius of the target arc determined by the corresponding three change vectors is smaller, and the curvature value is larger. This indicates a higher degree of anomaly in the adjacent trading object pair, thus providing a quantitative basis for the determination of abnormal trading objects at the overall method level.
[0064] The method for determining abnormal transaction objects according to the embodiments of this application determines the target distance between adjacent nodes in the curvature relationship graph based on the target curvature corresponding to adjacent transaction objects, including: determining the repulsion strength between adjacent nodes corresponding to adjacent transaction objects based on the target curvature corresponding to adjacent transaction objects, wherein the target curvature and the repulsion strength are positively correlated; and taking the product of the repulsion strength between adjacent nodes and the preset distance as the target distance between adjacent nodes.
[0065] For example, such as Figure 3 As shown, repulsion strength refers to the degree of distance between adjacent nodes of adjacent trading objects in the curvature relationship graph. This repulsion strength is determined by the target curvature of the adjacent trading objects. The larger the target curvature, the greater the repulsion strength, indicating that the state change trajectory of the adjacent trading object pair is more drastically curved, and therefore should be placed further apart in the curvature relationship graph. The preset distance is a pre-defined constant used to quantify the repulsion strength into an actual spatial distance value.
[0066] For example, firstly, the target curvature corresponding to adjacent trading objects is mapped to a preset numerical range, such as [0, 1], to determine the repulsion strength between adjacent nodes corresponding to the adjacent trading objects. Then, the repulsion strength is multiplied by a preset distance to obtain the target distance between adjacent nodes. Assuming the preset distance is 1 unit, when the repulsion strength corresponding to adjacent trading objects is 0.8, the target distance is 0.8 × 1 = 0.8 units. Trading object pairs with larger target curvatures are spaced further apart in the curvature relationship graph, while trading object pairs with smaller target curvatures are closer together in the curvature relationship graph.
[0067] The above steps enable the spatial positions of nodes in the curvature graph to intuitively reflect the differences in the state change trajectories between trading objects. Trading object pairs with smaller target curvature cluster together in the graph, while those with larger target curvature are pushed away from the group and placed in isolated positions. This allows subsequent steps to accurately identify trading objects whose state behavior deviates from the group through local neighborhood density, improving the accuracy of identifying abnormal trading objects.
[0068] According to the method for determining abnormal transaction objects according to the embodiments of this application, after determining the target node from multiple nodes included in the curvature relationship graph, the method further includes: determining at least one node adjacent to the target node in the curvature relationship graph; determining the target curvature corresponding to the target node and the at least one adjacent node as the abnormal curvature; determining three consecutive change vectors corresponding to the abnormal curvature, and determining the change vector located in the middle position among the three consecutive change vectors as the abnormal vector; and determining the time point corresponding to the abnormal vector as the time point when the abnormal transaction object exhibits abnormal behavior within a preset time period.
[0069] For example, a target node is determined from the curvature graph based on its local neighborhood density, and the transaction object corresponding to this node is identified as an abnormal transaction object. Adjacent nodes are those adjacent to the target node in the curvature graph. Anomaly curvature refers to the target curvature shared by the target node and its adjacent nodes; this curvature value causes the target node to be pushed out of the group in the curvature graph. Anomaly vectors are the change vectors located in the middle of the continuous change vector group used to calculate the anomaly curvature; the time point corresponding to this change vector is the time point at which the abnormal behavior occurred.
[0070] It should be noted that the continuous change vector group used when calculating abnormal curvature may include three continuous change vectors or more continuous change vectors. In actual calculation, the length of the continuous change vector group can be flexibly selected according to the data situation. This embodiment is not limited to this.
[0071] For example, when determining abnormal curvature, this abnormal curvature corresponds to the target distance between the target node and its adjacent nodes, and corresponds to a specific continuous change vector group (this continuous change vector group contains at least three continuous change vectors). The change vector located in the middle position of this continuous change vector group is determined as the abnormal vector, and the time point corresponding to this abnormal vector is the time point within the preset time period when the abnormal transaction object exhibits abnormal behavior. For example, the continuous change vector group corresponding to the abnormal curvature is (V2, V3, V4), where the change vector located in the middle position is V3, and the time point corresponding to V3 is the time point when the abnormal transaction object exhibits abnormal behavior.
[0072] For example, after determining the time point when the abnormal transaction object exhibits abnormal behavior, the text block in the original text data corresponding to that time point is obtained, and key statements related to the abnormal behavior are extracted from the text block. This allows the abnormality determination result to be directly associated with the original text data, making the determination result traceable and verifiable.
[0073] The above steps, after identifying the abnormal transaction object, further pinpoint the exact time point when the abnormal behavior occurred. Since curvature is calculated from the changes in the feature vectors of adjacent transaction objects along the time axis, the magnitude of the curvature value directly reflects the degree of bending in the trajectory of the transaction object's state change. Therefore, the time point corresponding to the abnormal curvature is the time point when the state change trajectory experiences maximum bending, which is also the time point when the abnormal transaction object exhibits abnormal behavior within a preset time period. When an abnormal transaction object is identified in a monitoring scenario after a transaction has occurred, this method can accurately pinpoint the specific time point when the abnormal transaction object exhibits abnormal behavior within a preset time period, thus providing a clear time indication for subsequent risk investigation and handling, further improving the efficiency of risk investigation.
[0074] In some embodiments, the method further includes: retrieving text fragments most relevant to a specific risk event type corresponding to the abnormal trading behavior of the abnormal trading object from a preset feature vector library; combining the retrieved text fragments with risk indicator instructions to form prompt words and inputting them into a Large Language Model (LLM). The LLM performs risk identification and classification based on the prompt words, determining whether the text fragments describe a risk event. It then summarizes the risk event corresponding to the abnormal trading object in natural language and explicitly identifies the corresponding text block. If the risk level of the risk event is higher than a risk threshold, the abnormal trading object, risk event, and abnormal time point identified in the above steps are reported to a human operator in the form of a report.
[0075] Based on the above-mentioned method for determining abnormal transaction objects, this application also provides a device for determining abnormal transaction objects. The following will be combined with... Figure 4 The device is described in detail.
[0076] Figure 4 A schematic block diagram of a device for determining abnormal transaction objects according to an embodiment of this application is shown.
[0077] like Figure 4 The device 400 for determining abnormal transaction objects shown includes an acquisition module 401, a processing module 402, a construction module 403, a first determination module 404, and a second determination module 405.
[0078] The acquisition module 401 is used to acquire multiple feature vectors corresponding to multiple time points within a preset time period after multiple trading objects have performed the same trading action. In one embodiment, the acquisition module 401 can be used to perform the operation S210 described above, which will not be repeated here.
[0079] The processing module 402 is used to sort multiple transaction objects according to time, take the sum of the feature vectors of adjacent transaction objects at each time point as the change vector corresponding to the time point, and concatenate the multiple change vectors corresponding to multiple time points in sequence according to time to obtain a change vector sequence, and calculate the target curvature of the change vector sequence. In one embodiment, the processing module 402 can be used to perform the operation S220 described above, which will not be repeated here.
[0080] The construction module 403 is used to construct a curvature relationship graph based on multiple transaction objects and target curvatures. The curvature relationship graph uses transaction objects as nodes, and the target distance between adjacent nodes in the curvature relationship graph is determined based on the target curvatures corresponding to adjacent transaction objects. The target curvature and target distance are positively correlated. In one embodiment, the construction module 403 can be used to perform the operation S230 described above, which will not be repeated here.
[0081] The first determining module 404 is used to determine a search area with nodes in the curvature graph as centers and a preset search radius, and to use the number of other nodes within the search area as the local neighborhood density of the node. In one embodiment, the first determining module 404 can be used to perform the operation S240 described above, which will not be repeated here.
[0082] The second determining module 405 is used to determine a target node from multiple nodes contained in the curvature relationship graph based on the local neighborhood density, and to determine the transaction object corresponding to the target node as an abnormal transaction object. In one embodiment, the second determining module 405 can be used to perform the operation S250 described above, which will not be repeated here.
[0083] According to an embodiment of this application, the acquisition module 401 is further configured to acquire the original text data of multiple trading objects within a preset time period after the transaction; divide the original text data into blocks according to multiple time points within the preset time period to obtain text blocks corresponding to each time point; input the text blocks into a preset feature vector library for retrieval to obtain multiple associated feature vectors related to the semantics of the text blocks; and fuse the multiple associated feature vectors to obtain the feature vector corresponding to the time point.
[0084] According to an embodiment of this application, the processing module 402 is further configured to obtain multiple continuous change vector groups in the change vector sequence, each continuous change vector group including at least three continuous change vectors; determine multiple initial curvature values based on the multiple continuous change vector groups, and use the average of the multiple initial curvature values as the target curvature of the change vector sequence.
[0085] According to an embodiment of this application, the processing module 402 is further configured to construct a target arc based on three consecutive change vectors; and to determine the reciprocal of the radius of the target arc as the initial curvature value corresponding to the group of consecutive change vectors.
[0086] According to an embodiment of this application, the construction module 403 is further configured to determine the repulsion strength between adjacent nodes corresponding to adjacent transaction objects based on the target curvature corresponding to adjacent transaction objects, wherein the target curvature and the repulsion strength are positively correlated; and to use the product of the repulsion strength between adjacent nodes and the preset distance as the target distance between adjacent nodes.
[0087] According to an embodiment of this application, the second determining module 405 is further configured to, after determining the target node from the plurality of nodes contained in the curvature relation graph, determine at least one node adjacent to the target node in the curvature relation graph; determine the target curvature corresponding to the target node and the at least one adjacent node as the abnormal curvature; determine three consecutive change vectors corresponding to the abnormal curvature, and determine the change vector located in the middle position among the three consecutive change vectors as the abnormal vector; and determine the time point corresponding to the abnormal vector as the time point when the abnormal transaction object exhibits abnormal behavior within a preset time period.
[0088] Figure 5 A block diagram schematically illustrates an electronic device suitable for implementing a method for determining abnormal transaction objects according to an embodiment of this application.
[0089] like Figure 5As shown, an electronic device 500 according to an embodiment of this application includes a processor 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage portion 508 into a random access memory (RAM) 503. The processor 501 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor, and / or an associated chipset and / or a special-purpose microprocessor, such as an application-specific integrated circuit (ASIC). The processor 501 may also include onboard memory for caching purposes. The processor 501 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.
[0090] RAM 503 stores various programs and data required for the operation of electronic device 500. Processor 501, ROM 502, and RAM 503 are interconnected via bus 504. Processor 501 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 502 and / or RAM 503. It should be noted that programs may also be stored in one or more memories other than ROM 502 and RAM 503. Processor 501 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in one or more memories.
[0091] According to embodiments of this application, the electronic device 500 may further include an input / output (I / O) interface 505, which is also connected to a bus 504. The electronic device 500 may also include one or more of the following components connected to the input / output (I / O) interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the input / output (I / O) interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 510 as needed so that computer programs read from it can be installed into the storage section 508 as needed.
[0092] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.
[0093] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 502 and / or RAM 503 and / or one or more memories other than ROM 502 and RAM 503 described above.
[0094] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the methods provided in the embodiments of this application.
[0095] When the computer program is executed by the processor 501, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.
[0096] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 509, and / or installed from a removable medium 511. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.
[0097] In such an embodiment, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by processor 501, it performs the functions defined in the system of this application embodiment. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.
[0098] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).
[0099] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0100] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.
[0101] The embodiments of this application have been described above. However, these embodiments are merely illustrative and not intended to limit the scope of this application. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. Without departing from the scope of this application, those skilled in the art can make various substitutions and modifications, all of which should fall within the scope of this application.
Claims
1. A method for identifying abnormal transaction objects, characterized in that, include: Obtain multiple feature vectors corresponding to multiple time points within a preset time period after multiple trading objects have engaged in the same trading activity; The multiple transaction objects are sorted according to time, and the sum of the feature vectors of adjacent transaction objects at each time point is taken as the change vector corresponding to the time point. The multiple change vectors corresponding to the multiple time points are concatenated in time to obtain a change vector sequence, and the target curvature of the change vector sequence is calculated. A curvature relationship graph is constructed based on the multiple transaction objects and the target curvature, wherein the curvature relationship graph uses the transaction objects as nodes, and the target distance between adjacent nodes in the curvature relationship graph is determined according to the target curvature corresponding to the adjacent transaction objects, and the target curvature is positively correlated with the target distance; Using the nodes in the curvature graph as centers, a search area is determined according to a preset search radius, and the number of other nodes in the search area is used as the local neighborhood density of the node. Based on the local neighborhood density, a target node is determined from the multiple nodes contained in the curvature relationship graph, and the transaction object corresponding to the target node is identified as an abnormal transaction object.
2. The method for determining abnormal transaction objects according to claim 1, characterized in that, Obtain multiple feature vectors corresponding to multiple time points within a preset time period after multiple trading objects have engaged in the same trading activity, including: Obtain the original text data of the multiple trading objects within a preset time period after the transaction; The original text data is divided into blocks according to the multiple time points within the preset time period to obtain the text block corresponding to each time point; The text block is input into a preset feature vector library for retrieval to obtain multiple associated feature vectors that are semantically related to the text block; The multiple associated feature vectors are fused to obtain the feature vector corresponding to the time point.
3. The method for determining abnormal transaction objects according to claim 1, characterized in that, Calculating the target curvature of the changed vector sequence includes: Obtain multiple consecutive change vector groups from the change vector sequence, each of the consecutive change vector groups including at least three consecutive change vectors; Multiple initial curvature values are determined based on the multiple continuously changing vector groups, and the average value of the multiple initial curvature values is used as the target curvature of the changing vector sequence.
4. The method for determining abnormal transaction objects according to claim 3, characterized in that, Calculate the target curvature of the changed vector sequence, and determine multiple initial curvature values based on the multiple consecutive changed vector groups, including: Construct the target circular arc based on the three continuous change vectors; The reciprocal of the radius of the target arc is determined as the initial curvature value corresponding to the continuously changing vector group.
5. The method for determining abnormal transaction objects according to claim 1, characterized in that, Determining the target distance between adjacent nodes in the curvature relationship graph based on the target curvature corresponding to the adjacent transaction objects includes: The repulsion strength between adjacent nodes corresponding to the adjacent transaction objects is determined based on the target curvature corresponding to the adjacent transaction objects, and the target curvature and the repulsion strength are positively correlated. The product of the repulsion strength between adjacent nodes and the preset distance is taken as the target distance between the adjacent nodes.
6. The method for determining abnormal transaction objects according to claim 3, characterized in that, After determining the target node from the plurality of nodes contained in the curvature relation graph, the method further includes: In the curvature relationship diagram, at least one node adjacent to the target node is determined; The target curvature that is common to the target node and at least one adjacent node is determined as an anomalous curvature; Determine the three consecutive change vectors corresponding to the abnormal curvature, and identify the change vector located in the middle position among the three consecutive change vectors as the abnormal vector; The time point corresponding to the anomaly vector is determined as the time point within the preset time period when the abnormal transaction object exhibits abnormal behavior.
7. A device for identifying abnormal transaction objects, characterized in that, The device includes: The acquisition module is used to acquire multiple feature vectors corresponding to multiple time points within a preset time period after multiple trading objects have engaged in the same trading behavior. The processing module is used to sort the multiple transaction objects according to time, take the sum of the feature vectors of adjacent transaction objects at each time point as the change vector corresponding to the time point, and concatenate the multiple change vectors corresponding to the multiple time points in time to obtain a change vector sequence, and calculate the target curvature of the change vector sequence. A construction module is used to construct a curvature relationship graph based on the multiple transaction objects and the target curvature, wherein the curvature relationship graph uses the transaction objects as nodes, and determines the target distance between adjacent nodes in the curvature relationship graph based on the target curvature corresponding to the adjacent transaction objects, and the target curvature is positively correlated with the target distance; The first determining module is used to determine a search area with the nodes in the curvature relationship graph as the center and according to a preset search radius, and to use the number of other nodes in the search area as the local neighborhood density of the node. The second determining module is used to determine a target node from multiple nodes contained in the curvature relationship graph based on the local neighborhood density, and to determine the transaction object corresponding to the target node as an abnormal transaction object.
8. An electronic device, comprising: One or more processors; Memory, used to store one or more computer programs. The characteristic feature is that the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1 to 6.
9. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 6.
10. A computer program product, comprising a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the steps of the method according to any one of claims 1 to 6.
