Target object identification method and device, computer equipment and readable storage medium

Abstract

The invention relates to a target object recognition method and device, computer equipment and a readable storage medium. The method comprises the following steps: acquiring historical stock data, wherein the historical stock data comprises a plurality of organizations and attribute information of different dimensions of each organization; determining each organization as a node of the initial relationship network, and determining associated organizations associated with the organizations through the target medium and association relationships of different levels between the organizations and the associated organizations according to the attribute information of each organization; according to the incidence relation of different levels, constructing the initial relation network to obtain a constructed multi-medium relation network; and obtaining medium data of the to-be-identified organization, performing retrieval in the multi-medium relation network based on the medium data, and if a target organization exists in the multi-medium relation network and an association relationship exists between the to-be-identified organization and the target medium, determining an identification result of the to-be-identified organization according to the target organization. By adopting the method, the target object recognition accuracy can be improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a target object recognition method, device, computer equipment and readable storage medium. Background Art

[0002] With the development of Internet technology, more and more cross-regional business processing is taking place. However, the complex cross-regional business processing environment involves diverse organizational roles and complex relationship links, among which the relationship links may include key links such as resource flow, information flow, and logistics. These links may also be associated at the physical level through media such as IP addresses and devices, forming a complex heterogeneous network.

[0003] However, in cross-regional business, as key links such as resource flow, information flow and logistics become more complex, the difficulty of monitoring and managing these links also increases. Target objects (for example, different organizations or individuals, etc.) can use this complexity to hide their illegal activities, such as committing fraud by forging logistics information or tampering with resource flows. Therefore, it is necessary to identify these target objects in business scenarios to ensure the security of business processing.

[0004] In the related art, target objects in business scenarios are identified through IP addresses or device information. This method cannot accurately identify target objects in complex heterogeneous networks. Summary of the invention

[0005] Based on this, it is necessary to provide a target object recognition method, apparatus, computer device, computer-readable storage medium and computer program product that can improve the accuracy of target object recognition in order to solve the above technical problems.

[0006] In a first aspect, the present application provides a target object recognition method, comprising:

[0007] Acquire historical stock data, where the historical stock data includes multiple organizations and attribute information of different dimensions of each of the organizations;

[0008] Determine each of the organizations as a node of the initial relationship network, determine the associated organizations associated with the organization through the target medium, and the different levels of association relationships between the organization and the associated organizations according to the attribute information of each of the organizations; wherein the target medium is determined according to the attribute information;

[0009] According to the association relationships at different levels, the initial relationship network is constructed to obtain a constructed multi-media relationship network;

[0010] The medium data of the organization to be identified is obtained, and a search is performed in the multi-media relationship network based on the medium data. If there is a target organization in the multi-media relationship network and there is an association relationship between the organization to be identified through the target medium, the identification result of the organization to be identified is determined according to the target organization.

[0011] In one embodiment, the attribute information includes static data and historical behavior data of the organization, the historical behavior data includes first historical behavior data and second historical behavior data, the target medium includes first type medium and second type medium, and the determining each of the organizations as a node of the initial relationship network, determining the associated organization associated with the organization through the target medium according to the attribute information of each of the organizations, and the association relationships of different levels between the organization and the associated organizations, includes:

[0012] Determine each of the organizations as a node of an initial relationship network, determine the static data and the first historical behavior data as a first type of medium, and determine the second historical behavior data as a second type of medium;

[0013] For each of the organizations, determining a first associated organization having a first associated relationship through the first type of medium, and determining a second associated organization having a second associated relationship through the second type of medium;

[0014] The correlation degree of the first correlation relationship is greater than the correlation degree of the second correlation relationship.

[0015] In one embodiment, constructing the initial relationship network according to the association relationships at different levels to obtain a constructed multimedia relationship network includes:

[0016] According to the first association relationship, an edge for connecting the nodes corresponding to the first association relationship in the initial relationship network is generated, and according to the second association relationship, an edge for connecting the nodes corresponding to the second association relationship in the initial relationship network is generated to obtain a first multimedia relationship network;

[0017] Performing connectivity detection and multi-media information fusion on the first multi-media relationship network to obtain a second multi-media relationship network;

[0018] The network structure of the second multimedia relationship network is optimized to obtain a constructed multimedia relationship network; the multimedia relationship network includes at least one community structure, and each community structure has a target node.

[0019] In one embodiment, the performing connectivity detection and multi-media information fusion on the first multi-media relationship network to obtain the second multi-media relationship network includes:

[0020] Performing connectivity detection on the first multimedia relationship network to obtain all connected subgraphs in the first multimedia relationship network;

[0021] Determine the number of nodes in all the connected subgraphs, determine a maximum connected subgraph according to the number of nodes, and determine the maximum connected subgraph as a core network of the first multimedia relationship network;

[0022] Determine the weights corresponding to different types of edges between the nodes and the centrality index of each node according to the importance of the association relationship between the nodes in the first multimedia relationship network, and obtain a second multimedia relationship network;

[0023] The nodes are directly or indirectly associated with each other through the first association relationship and / or the second association relationship.

[0024] In one embodiment, the step of optimizing the network structure of the second multimedia relationship network to obtain a constructed multimedia relationship network includes:

[0025] Performing noise filtering on the edges of the second multimedia relationship network according to a preset correlation threshold to obtain a third multimedia relationship network;

[0026] The third multimedia relationship network is optimized to obtain a constructed multimedia relationship network.

[0027] In one embodiment, the step of performing network optimization on the candidate multimedia relationship network to obtain a constructed multimedia relationship network includes:

[0028] If there is a replaceable node in the node of the third multimedia relationship network, merge the node and the replaceable node to obtain a merged node, and the merged node has an association relationship between the node and the replaceable node; and / or

[0029] If two nodes in the fourth multimedia relationship network have different types of first association relationships and / or second association relationships, merge the edges corresponding to the different types of first association relationships and / or second association relationships to obtain merged edges, and determine a candidate multimedia relationship network;

[0030] Wherein, the merged edge includes the first association relationship and / or the second association relationship of the same type, and the fourth multimedia relationship network is the third multimedia relationship network or the third multimedia relationship network after merging nodes;

[0031] The community structure of the candidate multimedia relationship network is identified, and the target node of the community structure is determined according to the centrality index to obtain a constructed multimedia relationship network.

[0032] In one embodiment, the medium data includes attribute information of different dimensions of the organization to be identified, and the retrieval is performed in the multi-media relationship network based on the medium data. If there is a target organization and the organization to be identified in the multi-media relationship network and there is an association relationship through the target medium, then the identification result of the organization to be identified is determined according to the target organization, including:

[0033] Determining a behavior label of each node in the multimedia relationship network and a behavior label value corresponding to the behavior label;

[0034] Based on the medium data, the multi-media relationship network is searched, and if there is a target organization in the multi-media relationship network and there is an association relationship between the target organization and the organization to be identified through the target medium, then the target community structure to which the target organization belongs in the multi-media relationship network is determined; the target medium is the first type of medium, and / or the second type of medium;

[0035] Determine a target node of the target community structure and a first behavior label and a first label value of the target node; the first label value is used to represent a probability value that the organization corresponding to the target node has the first behavior;

[0036] According to the association relationship between the target organization and the nodes in the multimedia relationship network, the probability value of the existence of the first behavior in the organization to be identified is determined, and the identification result of the organization to be identified is obtained.

[0037] In one embodiment, determining the probability value of the existence of the first behavior in the organization to be identified according to the association relationship between the target organization and the nodes in the multimedia relationship network includes:

[0038] Determining a first association weight between the to-be-identified organization and the target organization;

[0039] In the case where an association relationship exists between the organization corresponding to the target node and the target organization, determining a second association weight between the organization corresponding to the target node and the target organization;

[0040] A second label value of the first behavior propagating to the to-be-identified organization is determined according to the first association weight, the second association weight, and the first label value, wherein the second label value is used to characterize a probability value of the to-be-identified organization having the first behavior.

[0041] In one embodiment, the method further comprises:

[0042] If there are other target organizations in the multi-media relationship network that have an association relationship with the organization to be identified through the target medium, and the other target organizations do not belong to the target community structure, determining a third association weight between the organization to be identified and the other target organizations;

[0043] A third label value of the first behavior propagating to the other target tissues is determined according to the third association weight and the second label value, wherein the third label value is used to characterize a probability value of the first behavior existing in the other target tissues.

[0044] In one embodiment, the method further comprises:

[0045] Obtaining a preset update strategy for the multi-media relationship network; the preset update strategy includes at least one of a time window mechanism, a preset event trigger update, and a historical association retrospective update;

[0046] The multimedia relationship network is updated according to the preset update strategy to obtain an updated multimedia relationship network.

[0047] In a second aspect, the present application further provides a target object recognition device, comprising:

[0048] A data acquisition module, used to acquire historical stock data, wherein the historical stock data includes multiple organizations and attribute information of different dimensions of each organization;

[0049] A data analysis module, configured to determine each of the organizations as a node of an initial relationship network, and determine, based on the attribute information of each of the organizations, associated organizations associated with the organization through a target medium, and association relationships of different levels between the organization and the associated organizations; wherein the target medium is determined based on the attribute information;

[0050] A relationship network construction module, used to construct the initial relationship network according to the association relationships at different levels to obtain a constructed multimedia relationship network;

[0051] The identification module is used to obtain the media data of the organization to be identified, and to search the multi-media relationship network based on the media data. If there is a target organization in the multi-media relationship network and there is an association relationship between the organization to be identified through the target medium, the identification result of the organization to be identified is determined based on the target organization.

[0052] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0053] Acquire historical stock data, where the historical stock data includes multiple organizations and attribute information of different dimensions of each of the organizations;

[0054] Determine each of the organizations as a node of the initial relationship network, determine the associated organizations associated with the organization through the target medium, and the different levels of association relationships between the organization and the associated organizations according to the attribute information of each of the organizations; wherein the target medium is determined according to the attribute information;

[0055] According to the association relationships at different levels, the initial relationship network is constructed to obtain a constructed multi-media relationship network;

[0056] The medium data of the organization to be identified is obtained, and a search is performed in the multi-media relationship network based on the medium data. If there is a target organization in the multi-media relationship network and there is an association relationship between the organization to be identified through the target medium, the identification result of the organization to be identified is determined according to the target organization.

[0057] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0058] Acquire historical stock data, where the historical stock data includes multiple organizations and attribute information of different dimensions of each of the organizations;

[0059] Determine each of the organizations as a node of the initial relationship network, determine the associated organizations associated with the organization through the target medium, and the different levels of association relationships between the organization and the associated organizations according to the attribute information of each of the organizations; wherein the target medium is determined according to the attribute information;

[0060] According to the association relationships at different levels, the initial relationship network is constructed to obtain a constructed multi-media relationship network;

[0061] The medium data of the organization to be identified is obtained, and a search is performed in the multi-media relationship network based on the medium data. If there is a target organization in the multi-media relationship network and there is an association relationship between the organization to be identified through the target medium, the identification result of the organization to be identified is determined according to the target organization.

[0062] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0063] Acquire historical stock data, where the historical stock data includes multiple organizations and attribute information of different dimensions of each of the organizations;

[0064] Determine each of the organizations as a node of the initial relationship network, determine the associated organizations associated with the organization through the target medium, and the different levels of association relationships between the organization and the associated organizations according to the attribute information of each of the organizations; wherein the target medium is determined according to the attribute information;

[0065] According to the association relationships at different levels, the initial relationship network is constructed to obtain a constructed multi-media relationship network;

[0066] The medium data of the organization to be identified is obtained, and a search is performed in the multi-media relationship network based on the medium data. If there is a target organization in the multi-media relationship network and there is an association relationship between the organization to be identified through the target medium, the identification result of the organization to be identified is determined according to the target organization.

[0067] The target object identification method, device, computer equipment, computer readable storage medium and computer program product, by acquiring attribute information of multiple organizations and different dimensions of each organization, use attribute data of different dimensions as the medium of association relationship between organizations, obtain multiple different types of media, and can determine the association relationship of different levels between organizations through different types of media. Based on different types of media, the complex interactive relationship between different media can be considered. On this basis, each organization is determined as a node of the initial relationship network, and the association relationship of different levels between organizations through different types of media is constructed to obtain a constructed multi-media relationship network. According to the multi-media relationship network constructed in this way, when the organization to be identified is obtained, the medium data of the organization to be identified is retrieved in the multi-media relationship network, and the relationship between the organization to be identified and each node in the multi-media relationship network can be determined based on different types of media. Compared with the single type of medium identification in the prior art, the complex interactive relationship between different media and the association hidden in the multi-layer relationship are considered. In addition, by dividing the association relationship into levels, the refinement of the identification granularity is achieved, the accuracy and identification ability of the identification are improved, and the reliability of the identification is ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0069] Figure 1 A diagram of an application environment of a target object recognition method in an embodiment;

[0070] Figure 2 A schematic diagram of a target object identification method in one embodiment;

[0071] Figure 3 is a schematic flow chart of step 204 in one embodiment;

[0072] Figure 4 A schematic diagram of a flow chart of a method for constructing a multimedia relationship network in one embodiment;

[0073] Figure 5 is a flow chart of a method for determining a second multimedia relationship network in one embodiment;

[0074] Figure 6 It is a flowchart of a target object identification method based on a multimedia relationship network in one embodiment;

[0075] Figure 7 is a flow chart of a target object identification method in another embodiment;

[0076] Figure 8 is a structural block diagram of a target object identification device in one embodiment;

[0077] Fig. 9 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0078] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0079] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0080] With the acceleration of globalization, cross-regional resource transfer (such as cross-border payments) has become an indispensable part of international trade. However, in this complex trading environment, the roles involved are diverse and the relationship links are complex, mainly including three key links: capital flow, information flow and logistics. These links may also be associated through media such as IP addresses and devices at the physical level, forming a complex heterogeneous network. Moreover, cross-border payment fraudsters have gradually evolved from single individual behavior to clusters and gangs. In such a network environment, traditional detection methods often rely on single-dimensional relationship analysis, which makes it difficult to fully cover and identify potential fraudulent behaviors.

[0081] For example, in business scenarios, for the actual business of cross-border payments, all parties need to interact in a highly complex and dynamically changing network. For example, the flow of funds involves banks, payment platforms, and user accounts; the flow of information covers transaction records, user identity verification and other data; and logistics involves the transportation and delivery of goods. At the same time, fraud gangs use these complex relationship networks to conduct fraudulent activities by integrating multiple relationships. Since the flow of funds, information flow, and logistics are usually represented as directed graphs, while the medium relationship is an undirected graph, traditional methods are difficult to effectively handle the complexity of such heterogeneous graph networks.

[0082] Therefore, in order to address the technical problem that traditional methods cannot accurately identify target objects in complex heterogeneous networks, a target object identification method is proposed. By obtaining attribute information of multiple organizations and different dimensions of each organization, the attribute data of different dimensions are used as the medium of the association relationship between organizations, and multiple different types of media are obtained, as well as the ability to determine the different levels of association between organizations through different types of media. Based on the different types of media, the complex interactive relationship between different media can be considered, each organization is determined as a node of the initial relationship network, and the different levels of association between organizations through different types of media are used to construct the initial relationship network to obtain a constructed multi-media relationship network. The multi-media relationship network is searched according to the medium data of the organization to be identified to identify the organization to be identified.

[0083] The target object recognition method provided in the embodiment of the present application can be applied to Figure 1In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The terminal obtains historical stock data from the server, and the historical stock data includes multiple organizations and attribute information of different dimensions of each organization; each organization is determined as a node of the initial relationship network, and the associated organization associated with the organization through the target medium is determined according to the attribute information of each organization, as well as the different levels of association between the organization and the associated organization; wherein the target medium is determined according to the attribute information; according to the associations at different levels, the initial relationship network is constructed to obtain a constructed multi-media relationship network; the medium data of the organization to be identified is obtained, and the multi-media relationship network is searched based on the medium data. If there is an association relationship between the target organization and the organization to be identified through the target medium in the multi-media relationship network, the identification result of the organization to be identified is determined according to the target organization. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices, and portable wearable devices. The server 104 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services.

[0084] It is understandable that in response to the processing needs of large-scale enterprise data, a distributed computing framework and efficient data storage solutions are adopted on the terminal. Specific measures include: Distributed computing framework: Hadoop ecosystem: Use Hadoop's distributed storage and computing capabilities to process large-scale data sets. Spark framework: Use Spark's memory computing advantages to achieve fast data processing and analysis. Efficient data storage: Graph databases (such as Neo4j, Nebula): Specialized for storing and querying graph structured data, supporting efficient graph traversal and community discovery operations. NoSQL databases: Such as MongoDB, suitable for storing semi-structured data, and flexibly responding to the diversity and dynamic changes of data.

[0085] In an exemplary embodiment, Figure 2 As shown, a target object recognition method is provided, and the method is applied to Figure 1 The terminal in the example is used to illustrate, including the following steps 202 to 208. Among them:

[0086] Step 202, obtaining historical stock data, where the historical stock data includes multiple organizations and attribute information of different dimensions of each organization.

[0087] Among them, the organization can be but not limited to a business organization, and the business organization can be but not limited to any one of the people in an enterprise, a group and an institution. This embodiment takes the organization as an example for explanation. The historical stock data includes multiple organizations and attribute information of different dimensions of each organization. For the attribute information of different dimensions of each organization, the attribute information may include static data and historical behavior data. Static data includes data that can fully reflect the basic characteristics and structure of the enterprise, for example, it may include the enterprise certification name, unified social credit code, legal person ID number, contact person mobile phone number, registered address, business scope, shareholder information and registered capital, etc. Further, in an exemplary embodiment, the static data may include a unified social credit code, legal person ID number, legal person address, legal person address split, ultimate beneficiary ID number, ultimate beneficiary address, ultimate beneficiary address split, agent ID number, agent address, agent address split, contact person mobile phone number, enterprise email address, enterprise registration address, enterprise registration address split and enterprise website, etc.

[0088] Historical behavior data may be historical transaction data. Historical transaction data includes business-related behavior data of an enterprise, which may include enterprise number, user ID, device identification, transaction amount, transaction time, delivery address, payment method, transaction frequency, etc. Historical behavior data may include at least transaction behavior data and user behavior data. Transaction behavior data may include any one of counterparty name, counterparty card number, counterparty address, counterparty email address, transaction country, transaction bank, logistics order number, contract number, and delivery address. User behavior data may include at least any one of IP address, device fingerprint, event time, and device type.

[0089] It is understandable that in order to ensure the quality and consistency of the data, the collected raw data can be preprocessed, and the preprocessing may include data cleaning, data standardization and data completion. Among them, data cleaning may include: Deduplication processing: Identify and remove duplicate data, such as business records, to avoid redundancy caused by multiple registrations or data entry errors. Outlier detection and processing: Filter out data that does not conform to the preset format, for example, a standardized unified social credit code (should be 18 digits), a mobile phone number (should be 11 digits), etc. Missing value processing: For missing data in key fields, the method of deleting records, interpolating missing values, or inferring missing items through other related information is used for processing. The specific implementation method can be achieved through existing methods and will not be repeated here.

[0090] Data standardization can include format unification, address splitting and time unification. Among them, format unification can be to unify mobile phone numbers into 11-digit pure digital format, remove prefixes "+XX", "XXXX", etc.; unify legal person ID numbers into 18 digits, and ensure their legitimacy through verification algorithms. Complete address format unification can be based on regional divisions and address record formats, unify the format of registered addresses, and ensure the consistency of address information. Address splitting: Standardize and split addresses according to regional divisions. If the customer address is not standard, complete the address and then split it into a standard format according to the regional level. Time format unification: Unify transaction time into a standard time format to facilitate subsequent time series analysis.

[0091] Data completion can be based on preset rules. For example, if a company lacks the mobile phone number of the legal person, it can be inferred through the company's certified mobile phone number or other contact information.

[0092] Exemplarily, historical stock data is obtained from a server, and the obtained historical stock data is preprocessed to obtain processed historical stock data.

[0093] Step 204, determine each organization as a node of the initial relationship network, determine the associated organizations associated with the organization through the target medium, and the different levels of association relationships between the organization and the associated organizations according to the attribute information of each organization; wherein the target medium is determined according to the attribute information.

[0094] Among them, the attribute data includes attributes of different dimensions. By dividing the attribute data, multiple types of attribute data are obtained, and each type of attribute data is used as a medium. Each type of attribute data is a data set, including multiple different attributes, so each type of medium also includes different media. The type of target medium can be multiple, and no specific limitation is made here. Among them, different types of target media have different corresponding levels of association. The target medium can include media used to establish transaction associations between enterprises and media of substantial association.

[0095] The medium used to establish transaction associations between enterprises may include transaction behavior data such as IP, counterparty name, counterparty card number, counterparty address, logistics order number, contract number, delivery address, etc. The media of substantial association may be the same legal person: multiple enterprises have the same legal person ID number; the same contact person mobile phone number: multiple enterprises share the same contact person mobile phone number; the same registered address: multiple enterprises are registered at the same or similar address; the same shareholder: multiple enterprises have the same shareholder information. The same agent: multiple enterprises have the same agent information. The same device fingerprint: multiple enterprises have the same device fingerprint information. The same email: multiple enterprises have the same registered email information.

[0096] Exemplarily, a preset initial relationship is obtained, each organization is determined as a node of the initial relationship network, and the associated organizations associated with the organization through the target medium and the different levels of association relationships between the organization and the associated organizations are determined according to the attribute information of each organization, and the edges between the corresponding associated nodes are generated according to the association relationships. It can be understood that organizations can be directly associated with each other or indirectly associated through other organizations, and there can be an association between organizations based on at least one type of medium.

[0097] Step 206, constructing the initial relationship network according to the association relationships at different levels to obtain a constructed multimedia relationship network.

[0098] The edges generated by the different levels of association here are reflected in the generated multi-media relationship network through the weight of the edge, and the weight of the edge can also be node and node (i.e., the association weight between enterprises). The determination of the edge weight can be based on the preset setting rules. Different weights are assigned to different types of edges according to the importance of different associations. For example, the weight of the same legal person relationship is higher than the transaction behavior association. The edge weight can also be assigned to the medium according to the medium type.

[0099] For example, edges between corresponding associated nodes are generated according to the association relationship, and different weights are assigned to different types of edges according to the importance of different association relationships. On this basis, the centrality index of each node in the relationship network obtained at this time is calculated to obtain a constructed multi-media relationship network. It can be understood that the centrality index can be used to identify key nodes in the network.

[0100] Step 208, obtaining the medium data of the organization to be identified, searching in the multi-media relationship network based on the medium data, if there is an association relationship between the target organization and the organization to be identified through the target medium in the multi-media relationship network, then determining the identification result of the organization to be identified based on the target organization.

[0101] The medium data includes attribute data of the same dimension as the historical stock data. Retrieval based on medium data in a multi-media relationship network can be implemented based on a hierarchical labeling algorithm, and the implementation principle of the hierarchical labeling algorithm can be implemented in an existing manner, which will not be elaborated here.

[0102] Exemplarily, before identifying the enterprise to be identified, a multi-media relationship network is constructed based on historical stock data, and the labels and features of each node in the network have been propagated. The node label here can be the behavior category of each node, and the features include features associated with the behavior category. Feature propagation refers to the process of propagating node features in the network. In this way, each node can obtain information about other nodes in the network, thereby updating its own feature representation. The specific propagation method can be achieved through existing methods, which will not be elaborated here.

[0103] Based on the media data, a search is performed in a multi-media relationship network, that is, a media matching is performed. In the process of media matching, matching can be performed according to the hierarchical relationship. If there is a relationship between the target organization and the organization to be identified through the target medium in the multi-media relationship network, the identification result of the organization to be identified is determined based on the target organization. Among them, the hierarchical relationship includes direct association and indirect association. Different levels of association such as direct association and indirect association can be distinguished by the type of target medium. The identification result of the organization to be identified based on the target organization can be identified according to a preset identification mode, such as multiple enterprises sharing the same legal person and frequently conducting large transactions, identifying potential fraud gangs.

[0104] For example, hierarchical relationship analysis is performed based on a multi-media relationship network. The first level is analyzed first, then the second level and deeper levels, and target organizations associated with different levels are obtained. Through predefined identification patterns, potential abnormal groups are identified, and risk concentration areas and key nodes in the network are identified, reflecting high-risk enterprises or groups. The fraud risk score of each enterprise to be identified is calculated based on factors such as the position of the enterprise in the network, the strength of association, and the abnormality of transaction behavior. Among them, the first level can be directly related enterprises, connected by strong associations, that is, the first association (such as the same legal person). The second level and deeper levels: enterprises connected by weak associations, that is, second associations (such as transaction behavior data), and the indirect associations between enterprises are analyzed layer by layer.

[0105] In the above target object identification method, by obtaining attribute information of multiple organizations and different dimensions of each organization, the attribute data of different dimensions are used as the medium of the association relationship between organizations, and multiple different types of media are obtained, and the association relationships of different levels between organizations through different types of media can be determined. Based on different types of media, the complex interactive relationship between different media can be considered. On this basis, each organization is determined as a node of the initial relationship network, and the association relationships of different levels between organizations through different types of media are constructed to obtain a constructed multi-media relationship network. According to the multi-media relationship network constructed in this way, when the organization to be identified is obtained, the medium data of the organization to be identified is retrieved in the multi-media relationship network, and the relationship between the organization to be identified and each node in the multi-media relationship network can be determined based on different types of media. Compared with the single type of medium identification in the prior art, the complex interactive relationship between different media and the association hidden in the multi-layer relationship are considered. In addition, by dividing the association relationship into levels, the refinement of the recognition granularity is achieved, the accuracy and recognition ability of the recognition are improved, and the reliability of the recognition is ensured.

[0106] In an exemplary embodiment, in order to ensure that the multi-media relationship network is more comprehensive, different types of heterogeneous data are integrated on the basis of the above-mentioned historical stock data to construct a more comprehensive complex relationship network, including: Social media data: By analyzing the public information of enterprises on social media, their associations and potential risks are identified. Public enterprise information: Utilize government-disclosed enterprise information databases, industry reports, etc. to enrich the association information between enterprises. Industry news and announcements: Analyze industry news and corporate announcements to identify potential associations and risk events. Entity alignment: Entity alignment is performed by matching the unique identifier of the enterprise (such as a unified social credit code) to ensure the consistency of enterprise information in different data sources. Multi-source data integration: Use data fusion technology (such as data warehouses, ETL processes) to integrate enterprise information from multiple data sources and build a unified enterprise relationship database.

[0107] In an exemplary embodiment, Figure 3 As shown, step 204 includes steps 302 to 304. Among them:

[0108] Step 302, each organization is determined as a node of the initial relationship network, static data and first historical behavior data are determined as first type of media, and second historical behavior data is determined as second type of media, the historical behavior data includes first historical behavior data and second historical behavior data.

[0109] The first historical behavior data may be part of the user behavior data in the historical behavior data, and the second historical behavior data may be all the historical behavior data in the historical behavior data except the first historical behavior data. It should be noted that the determination of the first type of medium and the second type of medium may be determined according to actual needs. The first historical behavior data may be a device fingerprint, and the second historical behavior data may be other historical behavior data except the device fingerprint.

[0110] Step 304 : for each organization, determine a first associated organization with which a first association relationship exists through a first type of medium, and determine a second associated organization with which a second association relationship exists through a second type of medium.

[0111] The degree of association of the first association relationship is greater than the degree of association of the second association relationship. The first association relationship may include an association relationship established based on the same legal person, the same contact person's mobile phone number, the same registered address, the same shareholder, the same agent, the same device fingerprint, and the same email address, and the second association relationship includes an association relationship established based on transaction behavior data such as IP, counterparty name, counterparty card number, counterparty address, logistics order number, contract number, and delivery address.

[0112] In this embodiment, by dividing the media between organizations into two types of media, the association relationship between organizations is determined based on the first type of media and the second type of media, and different levels of association relationships are taken into consideration, which can avoid the situation where a single perspective cannot capture the diverse media associations in cross-border payments.

[0113] In an exemplary embodiment, Figure 4 As shown, a method for constructing a multimedia relationship network is provided, including steps 402 to 406, wherein:

[0114] Step 402: Generate an edge for connecting nodes corresponding to the first association relationship in the initial relationship network according to the first association relationship, and generate an edge for connecting nodes corresponding to the second association relationship in the initial relationship network according to the second association relationship, to obtain a first multimedia relationship network.

[0115] Exemplarily, based on the first association relationship, an edge is generated for connecting the nodes corresponding to the first association relationship in the initial relationship network; based on the second association relationship, an edge is generated for connecting the nodes corresponding to the second association relationship in the initial relationship network; the corresponding nodes are connected according to the generated edges to obtain a first multimedia relationship network.

[0116] Step 404: Perform connectivity detection and multi-media information fusion on the first multi-media relationship network to obtain a second multi-media relationship network.

[0117] Among them, multi-media fusion includes association strength calculation (i.e. association weight calculation) and network topology optimization. Association strength calculation can be to assign different weights to different types of edges according to the importance of different association relationships. Network topology optimization includes edge weight normalization and node centrality analysis. Node centrality analysis includes calculating the centrality index of each node, and the centrality index includes degree centrality and betweenness centrality.

[0118] Optionally, in an exemplary embodiment, a method for determining a second multimedia relationship network is provided, such as Figure 5 As shown, the following steps are included:

[0119] Step 502: Perform connectivity detection on the first multimedia relationship network to obtain all connected subgraphs in the first multimedia relationship network.

[0120] Exemplarily, a connectivity detection algorithm is used to identify all connected subgraphs in the first multimedia relationship network, thereby obtaining all connected subgraphs in the first multimedia relationship network.

[0121] Step 504, determining the number of nodes in all connected subgraphs, determining the largest connected subgraph according to the number of nodes, and determining the largest connected subgraph as the core network of the first multimedia relationship network.

[0122] Step 506, according to the importance of the association relationship between the nodes in the first multimedia relationship network, determine the weights corresponding to the different types of edges between the nodes and the centrality index of each node to obtain the second multimedia relationship network.

[0123] Among them, the centrality index can be used to determine the target node of the community structure, such as the central node. The nodes are directly or indirectly associated through the first association relationship and / or the second association relationship. The centrality indicators include degree centrality, betweenness centrality and PageRank. When determining the target node, the node corresponding to the largest value of degree centrality, betweenness centrality and PageRank can be used as the target node. Degree centrality: The degree centrality of a node refers to the number of edges directly connected to the node. In an undirected graph, the degree centrality of a node is the degree of the node, that is, the number of edges directly connected to the node, which can be expressed as: C_D(v) = deg(v), where deg(v) is the degree of node v. In a directed graph, the degree centrality of a node can be the sum of the in-degree and out-degree, or the in-degree centrality and out-degree centrality can be calculated separately. Degree centrality and can also include normalized degree centrality, which is the degree centrality divided by the total number of nodes in the network minus one.

[0124] Betweenness centrality can be used to measure the frequency of a node being an intermediate node in the shortest path between other pairs of nodes. The calculation method includes: for each pair of nodes (s, t) in the network, calculate all the shortest paths between them. For each shortest path, if node v is on the path, the count is increased by 1. All counts are added together to obtain the betweenness centrality of node v. The formula is: C_B(v)=Σ_(s≠v≠t)(σ_st(v) / σ_st); where σ_st is the number of shortest paths from s to t, and σ_st(v) is the number of shortest paths from s to t that pass through node v.

[0125] Normalized betweenness centrality is to divide betweenness centrality by the number of possible node pairs in the network, so that the betweenness centrality value can be scaled to between 0 and 1, which is convenient for comparing the centrality of nodes in networks of different sizes. The calculation formula can be expressed as: C'_B(v)=C_B(v) / ((N-1)(N-2) / 2).

[0126] Furthermore, the centrality index can also include PageRank, a recursively defined centrality index that takes into account the importance of connected nodes. The calculation method can be: initialize the PageRank value of each node, iteratively update the PageRank of each node until convergence, and the formula can be expressed as: PR(v) = (1-d) + d*Σ_(u∈M(v))(PR(u) / L(u)), where d is the damping factor (usually 0.85), M(v) is the set of nodes pointing to node v, and L(u) is the out-degree of node u.

[0127] The above-mentioned method for determining the second multimedia relationship network, multimedia information fusion can improve the accuracy of network analysis and enhance the robustness and adaptability of the network through association strength calculation and network topology optimization.

[0128] Step 406, optimizing the network structure of the second multimedia relationship network to obtain a constructed multimedia relationship network; the multimedia relationship network includes at least one community structure, and each community structure has a target node.

[0129] It is understandable that the constructed multi-media complex relationship network may contain noise and redundancy, and further optimization is needed to improve the network quality and analysis effect. Network structure optimization includes noise filtering, network simplification and community discovery, and network simplification includes node merging and edge merging. Community discovery includes applying community discovery methods (for example, Louvain algorithm, etc.) to identify community structures existing in the network structure. Furthermore, in order to improve the effect of relationship mining, community division can be achieved through a combination of community discovery algorithms and algorithm fusion strategies, where the community discovery algorithm combination includes: Girvan-Newman (GN) algorithm: a community discovery algorithm based on edge betweenness, suitable for detecting smaller-scale community structures. Infomap algorithm: a community discovery algorithm based on information flow, which can efficiently identify complex community structures.

[0130] The algorithm fusion strategy includes: Hierarchical aggregation: First use the GN algorithm to identify primary communities, then use the Infomap algorithm to refine the community structure and improve the accuracy of community discovery. Result integration: Run multiple community discovery algorithms in parallel, merge their community discovery results, and combine the advantages of multiple algorithms to obtain more accurate community divisions.

[0131] The target node is determined based on the centrality index of each node in the multimedia relationship network. The target node can be a key node of the network or a key node of the community structure. The key node can be but is not limited to a central node.

[0132] Exemplarily, based on the relationship network determined in the above steps, there are corresponding association weights between the nodes and the corresponding edges between the nodes. Further, the edges of the second multimedia relationship network can be noise filtered according to a preset association threshold, and the edges corresponding to values ​​less than the preset association threshold are removed, while the edges corresponding to values ​​greater than the preset association threshold are retained. On this basis, possible false associations can be identified and removed through statistical analysis, such as companies that randomly share a certain contact information, to obtain a third multimedia relationship network; the third multimedia relationship network is optimized to obtain a constructed multimedia relationship network.

[0133] In the above-mentioned method for constructing a multi-media relationship network, the accuracy of network analysis can be improved and the robustness and adaptability of the network can be enhanced through connectivity detection, maximum connected subgraph extraction, multi-media information fusion and network structure optimization.

[0134] In an exemplary embodiment, the network structure of the second multimedia relationship network is optimized to obtain a constructed multimedia relationship network, including: performing noise filtering on the edges of the second multimedia relationship network according to a preset correlation threshold to obtain a third multimedia relationship network; and performing network optimization on the third multimedia relationship network to obtain a constructed multimedia relationship network.

[0135] Further, performing network optimization on the candidate multimedia relationship network to obtain the constructed multimedia relationship network may include: if there is a replaceable node in the node in the third multimedia relationship network, merging the node and the replaceable node to obtain a merged node, the merged node having an association relationship between the node and the replaceable node; and / or

[0136] If two nodes in the fourth multimedia relationship network have different types of first association relationships and / or second association relationships, the edges corresponding to the different types of first association relationships and / or second association relationships are merged to obtain merged edges, and a candidate multimedia relationship network is determined; wherein the merged edges include the same type of first association relationships and / or second association relationships, and the fourth multimedia relationship network is the third multimedia relationship network or the third multimedia relationship network after merging the nodes;

[0137] The community structure of the candidate multi-media relationship network is identified, and the target nodes of the community structure are determined according to the centrality index to obtain the constructed multi-media relationship network.

[0138] The replaceable nodes can be determined based on the instructions inputted in the terminal interface, or can be determined by obtaining the social relationship network associated with the multi-media relationship network and identifying the social relationship characteristics of the nodes in the multi-media relationship network. This node merging merges nodes with highly similar features or multiple associations into one node to reduce the complexity of the network. The high similarity and multiple associations here can be understood as being greater than the corresponding pre-set threshold feature number and association threshold.

[0139] Edge merging can be understood as merging these edges into a comprehensive edge when there are multiple different types of association relationships between two nodes, while integrating the attributes and weights of different relationships.

[0140] In the above methods, the network quality and analysis effect are improved through noise filtering, network simplification and community discovery.

[0141] Based on the above method, the constructed multimedia relationship network is determined, and the following provides an application based on the multimedia relationship network. In an exemplary embodiment, Figure 6 As shown, a target object recognition method based on a multimedia relationship network is provided, including steps 602 to 608, wherein:

[0142] Step 602: Determine the behavior label of each node in the multimedia relationship network and the behavior label value corresponding to the behavior label.

[0143] Among them, the behavior label may include abnormal behavior labels and normal behavior labels of the organization. The abnormal behavior label may be a confirmed fraudulent customer, and the corresponding behavior label value is 1. In this step, the behavior labels of each node in the multi-media relationship network and the behavior label values ​​corresponding to the behavior labels are determined, which can indicate that the labels and features of each node in the network have been propagated. The behavior labels and behavior label values ​​can be determined by existing methods, which will not be elaborated here.

[0144] Step 604, searching the multi-media relationship network based on the media data, if there is an association relationship between the target organization and the organization to be identified through the target medium in the multi-media relationship network, then determining the target community structure to which the target organization belongs in the multi-media relationship network; the target medium is the first type of medium, and / or the second type of medium.

[0145] It is understandable that the search here can be understood as medium matching, that is, in the multi-media relationship network matching, whether the target organization and the organization to be identified have an association relationship through the first type of medium and / or the second type of medium. In the actual search process, the search can be started from the core network (maximum connected subgraph) of the multi-media relationship network. For the core network, the search can be carried out from the community structure in the core network as a unit. For the community structure, the search can be started from the key nodes of the community structure. In other words, the search can be carried out in the manner of the maximum connected subgraph, the community structure and then the key nodes.

[0146] Step 606 , determining a target node of the target community structure and a first behavior label and a first label value of the target node; the first label value is used to represent a probability value that the organization corresponding to the target node has the first behavior.

[0147] The target node here may be the central node of the target community structure. The first behavior may be different types of abnormal behaviors, such as fraudulent behavior and abnormal customer trade behavior, and the label values ​​corresponding to different types of abnormal behaviors may be different.

[0148] Step 608, based on the association relationship between the target organization and the nodes in the multimedia relationship network, determine the probability value of the existence of the first behavior in the organization to be identified, and obtain the identification result of the organization to be identified.

[0149] Among them, determining the probability value of the existence of the first behavior in the organization to be identified can be: by determining the first association weight of the organization to be identified and the target organization; in the case that the organization corresponding to the target node and the target organization have an association relationship, determining the second association weight of the organization corresponding to the target node and the target organization; according to the first association weight, the second association weight and the first label value, determining the second label value of the first behavior propagating to the organization to be identified, the second label value is used to characterize the probability value of the existence of the first behavior in the organization to be identified. According to the first association weight, the second association weight and the first label value, determining the second label value of the first behavior propagating to the organization to be identified can be to multiply the first association weight, the second association weight and the first label value to obtain the second label value.

[0150] According to the association between the target organization and the nodes in the multi-media relationship network, the probability value of the existence of the first behavior of the organization to be identified is determined, and the identification result of the organization to be identified is obtained. The identification result is output, for example, it can include the risk level of the enterprise to be identified, the information of related enterprises and the potential fraud gang structure for subsequent decision-making and response. The decision-making system enters the database for query based on the customer number of the organization to be identified, and finds the probability of the first behavior of the organization to be identified, and it comes from the target community structure. If it is abnormal, the application of the organization to be identified is rejected through comprehensive decision-making. Furthermore, after the propagation is completed, the centrality of each node is recalculated. If the target node is still the center of the target community structure, the result after propagation and the recalculated node centrality is updated and stored in the database.

[0151] For example, if the company C to be identified submits a document and has all the media data mentioned above, the company C to be identified is searched in the multi-media relationship network based on the media data of the company C to be identified, and the company C to be identified is associated with the node company B due to the legal person certificate number, with a weight of 1. The node company B is in the community Z whose central node is company A, and the node company B is associated with company A due to the device fingerprint, with a weight of 0.7. The central node A is a confirmed fraud customer with a fraud probability of 1. When it is propagated to company B, the fraud probability of company B is 1*0.7=0.7. When it is propagated to company C to be identified through B, the fraud probability is 1*0.7*1=0.7. After the propagation is completed, the centrality of each node is recalculated, and node A is still the center of community Z. The results of the propagation and the recalculated node centrality are updated and stored in the database. The decision system enters the database for query based on the customer number of node C, and finds that the fraud probability of node C is 0.7 and it comes from community Z. The application of node C is rejected through comprehensive decision-making.

[0152] In the above embodiment, by comprehensively considering static data and historical transaction data, and also through multi-media correlation analysis, the complex interactive relationships between organizations are revealed, and these relationships are analyzed layer by layer to provide more accurate and comprehensive risk assessment results.

[0153] In the related art, it is difficult to fully cover and identify potential fraudulent behaviors by relying on single-dimensional relationship analysis. In an exemplary embodiment, a method for identifying potential abnormal behaviors is provided. Based on the above-mentioned identification method of the organization to be identified, if there are other target organizations in the multi-media relationship network that have an association relationship with the organization to be identified through the target medium, and the other target organizations do not belong to the target community structure, a third association weight between the organization to be identified and the other target organizations is determined;

[0154] A third label value of the first behavior propagating to other target organizations is determined according to the third association weight and the second label value, where the third label value is used to characterize a probability value of the first behavior existing in other target organizations.

[0155] For example, if the company C to be identified submits a document and has all the media data mentioned above, the company C to be identified is searched in the graph data based on the media data of the company C to be identified, and the company C to be identified is associated with the node company B due to the legal person certificate number, with a weight of 1, and is associated with the node company D due to multiple IP addresses, with a comprehensive weight of 0.5. Node company B is in community Z, whose central node is company A, and the node company B is associated with company A due to device fingerprints, with a weight of 0.7. The central node A is a confirmed fraud customer, with a fraud probability of 1. When it is propagated to company B, the fraud probability of company B is 1*0.7=0.7, and the fraud probability of propagating to company C to be identified through B is 1*0.7*1=0.7.

[0156] Node D was not in community Z before the introduction of the to-be-identified enterprise C. After the introduction of the to-be-identified enterprise C, node D was associated with community Z. The fraud probability propagated from node C to node D is 1*0.7*1*0.5=0.35. After the propagation is completed, the centrality of each node is recalculated, and node A is still the center of community Z. The results of the propagation and recalculation of the node centrality are updated and stored in the database. The decision system enters the database for query based on the customer number of node C, and finds that the fraud probability of node C is 0.7 and comes from community Z. Through comprehensive decision-making, the entry of node C is rejected.

[0157] In the above embodiment, by comprehensively considering static data and historical transaction data and also through multi-media correlation analysis, the complex interactive relationships between organizations are revealed, and these relationships are analyzed layer by layer to further identify potential abnormal behaviors.

[0158] Taking into account the dynamic changes in inter-organizational relationships, a dynamic relationship network construction mechanism can be designed to regularly update the multimedia complex relationship network. In an exemplary embodiment, a preset update strategy for the multimedia relationship network is obtained; the preset update strategy includes at least one of a time window mechanism, a preset event-triggered update, and a historical association tracing update; the multimedia relationship network is updated according to the preset update strategy to obtain an updated multimedia relationship network.

[0159] Among them, the time window mechanism includes: rolling time window: set a fixed time window (such as monthly, quarterly), regularly rebuild the relationship network, and capture the latest organizational association changes. Event-driven update: based on specific events (such as changes in corporate registration information, large-scale transaction behaviors, etc.), trigger the instant update of the relationship network. Historical association tracing: maintain the historical records of corporate associations, analyze the evolution trend of inter-enterprise relationships, and identify potential emerging risk patterns.

[0160] Based on the construction of the above multi-media relationship network and the identification of target objects, the iteration of the multi-media relationship network and data storage and real-time query can be further realized. The iteration of the multi-media relationship network includes:

[0161] After completing the fraud risk assessment of the enterprise to be identified, its relevant data (i.e. the media data of the organization to be identified) will be incorporated into the multi-media complex relationship network to ensure dynamic updating and iterative optimization of network data. The media data of the newly added organization to be identified will be associated with the enterprises in the existing network, and the association strength and weight will be recalculated. According to the latest data and fraud behavior patterns, the attributes of the edges and nodes in the network will be dynamically adjusted to ensure the real-time and accuracy of the network structure.

[0162] For the identified potential fraud gang sub-network, in-depth optimization is carried out, such as adding relevant transaction behavior data, adjusting the community structure, etc., to improve the accuracy of the fraud score of the sub-network. The composite fraud score of the sub-network is recalculated through the optimized network structure. The specific steps include: Scoring indicator update: According to the latest network structure and corporate behavior data, various scoring indicators are updated, such as correlation strength, transaction anomaly, etc. Scoring model retraining: Using the optimized network data, the fraud scoring model is retrained to ensure that the model can reflect the latest risk changes. Scoring result verification: The optimized scoring results are verified to ensure the accuracy and reliability of the scoring and reduce false positives and false negatives.

[0163] In this way, after the fraud risk assessment of the enterprise to be identified is completed, its relevant data is added to the multi-media complex relationship network for iterative update, continuously optimizing the network structure and improving the fraud risk identification ability. Through this continuous data feedback and network iteration, it can gradually improve its understanding and analysis capabilities of complex relationships, further improve the detection accuracy and response speed of potential fraudulent behaviors, and ensure that the risk assessment model is always in the optimal state.

[0164] For data storage and real-time query, including:

[0165] Import the optimized multi-media complex relationship network and fraud identification results into the relevant database of Alibaba Cloud. The specific steps include:

[0166] Graph database import: Import the constructed and optimized multi-media complex relational network data into the Alibaba Cloud graph database (such as Alibaba Cloud Graph Compute Service), including node information, edge relationships and their attributes. Import the fraud identification results of the hierarchical label model into the Alibaba Cloud ADB database (AnalyticDB) to support high-performance real-time query and analysis.

[0167] By leveraging the powerful query capabilities of Alibaba Cloud Graph Database and ADB, we can achieve real-time query of fraud gang network relationships and risk identification results. Specific functions include:

[0168] Real-time association query: supports real-time query of all related enterprises and relationship types in a complex relationship network based on the enterprise's unified social credit code or other identifiers. Real-time risk query: real-time query of the fraud risk score and related risk indicators of the enterprise to be identified, and supports filtering and sorting by risk level. Visualization display: using the visualization tools provided by Alibaba Cloud, the complex relationship network and fraud identification results are graphically displayed to help users intuitively understand the complex relationships between enterprises and the distribution of fraud risks.

[0169] Among them, in order to ensure the seamless integration of Alibaba Cloud Graph Database and ADB Database (a high-performance real-time query database), specific measures include: Data synchronization mechanism: Establish a real-time data synchronization mechanism to ensure that the data in the graph database and the real-time query database are always consistent, and support fast data updates and query responses. Performance optimization: For large-scale data sets, optimize the query performance and storage efficiency of the database, such as using distributed storage, index optimization and other technical means to improve the overall performance of the system. Security assurance: Implement security measures for data storage and transmission, such as encrypted storage, access control, etc., to ensure the security and privacy protection of enterprise data.

[0170] In this method, the multi-media relationship network results are imported into the Alibaba Cloud graph database, and the hierarchical label model results are imported into the Alibaba Cloud ADB database. Based on the graph database and the real-time database, real-time query of the network relationships and risk identification of fraud gangs can be achieved.

[0171] In an exemplary embodiment, Figure 7 As shown, a target object recognition method is provided, and the method is applied to Figure 1 The terminal in the example is used to illustrate, including the following steps 702 to 708. Among them:

[0172] Step 702: construct a multimedia relationship network.

[0173] Step 704 , obtaining the medium data of the organization to be identified, and using the hierarchical labeling algorithm and the medium data to identify the organization to be identified based on the multi-media relationship network.

[0174] It should be noted that, according to the characteristics of the nodes in the multimedia relationship network, known labels have been assigned to the corresponding nodes. The labels include normal behavior labels and abnormal behavior labels. The abnormal behavior labels can be fraud labels. Hierarchical labeling algorithms (such as Label Propagation or Label Spreading) are applied to identify fraud risks on the graph. These algorithms propagate label information through an iterative process, so that the label status (fraud or non-fraud) of each node is updated based on the label status of its neighboring nodes.

[0175] Step 706, optimizing the multimedia relationship network.

[0176] Step 708: data storage and real-time query implementation.

[0177] It can be understood that the specific implementation of this example can be achieved by the method described in the above embodiment, which will not be described in detail here.

[0178] The above target object identification method obtains attribute information of multiple organizations and different dimensions of each organization, and uses attribute data of different dimensions as the medium of association between organizations, thereby obtaining multiple different types of media, and being able to determine the association between organizations at different levels through different types of media. Based on different types of media, the complex interactive relationship between different media can be considered. On this basis, each organization is determined as a node of the initial relationship network, and the association between organizations at different levels through different types of media is constructed to obtain a constructed multi-media relationship network. The multi-media relationship network constructed in this way simultaneously handles the relationship between directed graphs and undirected graphs, comprehensively covers potential abnormal behaviors through multi-dimensional comprehensive analysis, improves the accuracy of identification through hierarchical labeling algorithm and multi-dimensional feature extraction, reduces false positives and false negatives, and in addition, has efficient storage and query capabilities, ensures real-time update and efficient response of the system, adapts to the needs of large-scale enterprise data processing, improves the accuracy and ability of identification, and ensures the reliability of identification.

[0179] Optionally, in an exemplary embodiment, the above-mentioned target object recognition method can be combined with deep learning. In the process of feature extraction and model training, advanced technologies such as deep learning and graph neural network (GNN) are introduced to make full use of the structural information of complex relationship networks to improve the performance of fraud recognition models:

[0180] Graph Neural Network (GNN) Applications:

[0181] Node representation learning: Through GNN, learn the embedded representation of each enterprise node in the network to capture its local and global network characteristics. Graph classification and node classification: Use GNN to classify nodes (such as high risk, medium risk, low risk) and make predictions directly on the graph structure.

[0182] Deep Learning Model Ensemble:

[0183] Multimodal feature fusion: Combine graph structure features with other structured and unstructured features, and perform joint training through deep learning models (such as multi-layer perceptrons, convolutional neural networks, etc.) to improve the model's predictive ability.

[0184] Model optimization:

[0185] Hyperparameter tuning: Use automated tools (such as Optuna, Hyperopt) to optimize the hyperparameters of the GNN model, such as the number of layers, learning rate, number of hidden units, etc. Regularization technology: Apply regularization methods (such as Dropout, L2 regularization) to prevent model overfitting and improve the generalization ability of the model.

[0186] By introducing the above optional steps, the construction process of the multi-media complex relationship network can be further optimized and improved, ensuring the high quality and comprehensiveness of the network structure, and providing more solid technical support for the accurate identification of fraud gangs. These optional steps not only improve the depth and breadth of fraud identification, but also enhance the flexibility and adaptability of the system to meet the risk management needs in different business scenarios.

[0187] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0188] Based on the same inventive concept, the embodiment of the present application also provides a target object recognition device for implementing the target object recognition method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more target object recognition device embodiments provided below can refer to the limitations of the target object recognition method above, and will not be repeated here.

[0189] In an exemplary embodiment, Figure 8 As shown, a target object recognition device is provided, including: a data acquisition module 802, a data analysis module 804, a relationship network construction module 806 and a recognition module 808, wherein:

[0190] The data acquisition module 802 is used to acquire historical stock data, which includes multiple organizations and attribute information of different dimensions of each organization.

[0191] The data analysis module 804 is used to determine each organization as a node of the initial relationship network, and determine the associated organizations associated with the organization through the target medium and the different levels of association relationships between the organization and the associated organizations according to the attribute information of each organization; wherein the target medium is determined according to the attribute information.

[0192] The relationship network construction module 806 is used to construct the initial relationship network according to the association relationships at different levels to obtain a constructed multimedia relationship network.

[0193] The identification module 808 is used to obtain the media data of the organization to be identified, and search the multi-media relationship network based on the media data. If there is an association relationship between the target organization and the organization to be identified through the target medium in the multi-media relationship network, the identification result of the organization to be identified is determined based on the target organization.

[0194] The target object identification device obtains attribute information of multiple organizations and different dimensions of each organization, and uses attribute data of different dimensions as the medium of the association relationship between organizations, thereby obtaining multiple different types of media, and determining the association relationship of different levels between organizations through different types of media. Based on different types of media, the complex interactive relationship between different media can be considered. On this basis, each organization is determined as a node of the initial relationship network, and the association relationship of different levels between organizations through different types of media is constructed to obtain a constructed multi-media relationship network. According to the multi-media relationship network constructed in this way, when the organization to be identified is obtained, the medium data of the organization to be identified is retrieved in the multi-media relationship network, and the relationship between the organization to be identified and each node in the multi-media relationship network can be determined based on different types of media. Compared with the single type of medium identification in the prior art, the complex interactive relationship between different media and the association hidden in the multi-layer relationship are considered. In addition, by dividing the association relationship into levels, the refinement of the recognition granularity is achieved, the recognition accuracy and recognition ability are improved, and the reliability of the recognition is ensured.

[0195] In an exemplary embodiment, the attribute information includes static data and historical behavior data of an organization, the target medium includes a first type of medium and a second type of medium, and the data analysis module 804 is used to determine each organization as a node of the initial relationship network, the static data as the first type of medium, and the historical behavior data as the second type of medium;

[0196] For each organization, determining a first associated organization having a first associated relationship through a first type of medium, and determining a second associated organization having a second associated relationship through a second type of medium;

[0197] The correlation degree of the first correlation relationship is greater than the correlation degree of the second correlation relationship.

[0198] In an exemplary embodiment, the relationship network construction module 806 is further used to generate an edge for connecting the nodes corresponding to the first relationship in the initial relationship network according to the first relationship, and to generate an edge for connecting the nodes corresponding to the second relationship in the initial relationship network according to the second relationship, so as to obtain the first multimedia relationship network;

[0199] Performing connectivity detection and multi-media information fusion on the first multi-media relationship network to obtain a second multi-media relationship network;

[0200] The network structure of the second multimedia relationship network is optimized to obtain a constructed multimedia relationship network; the multimedia relationship network includes at least one community structure, and each community structure has a target node.

[0201] In an exemplary embodiment, the relationship network construction module 806 is further used to perform connectivity detection on the first multimedia relationship network to obtain all connected subgraphs in the first multimedia relationship network;

[0202] Determine the number of nodes in all connected subgraphs, determine the largest connected subgraph according to the number of nodes, and determine the largest connected subgraph as the core network of the first multimedia relationship network;

[0203] According to the importance of the association relationship between the nodes in the first multimedia relationship network, the weights corresponding to the different types of edges between the nodes and the centrality index of each node are determined to obtain a second multimedia relationship network;

[0204] The nodes are directly or indirectly associated with each other through a first association relationship and / or a second association relationship.

[0205] In an exemplary embodiment, the relationship network construction module 806 is further used to perform noise filtering on the edges of the second multimedia relationship network according to a preset correlation threshold to obtain a third multimedia relationship network;

[0206] The third multimedia relationship network is optimized to obtain a constructed multimedia relationship network.

[0207] In an exemplary embodiment, the relationship network construction module 806 is further used to merge the node and the replaceable node to obtain a merged node if there is a replaceable node in the node in the third multimedia relationship network, and the merged node has an association relationship between the node and the replaceable node; and / or

[0208] If two nodes in the fourth multimedia relationship network have different types of first association relationships and / or second association relationships, merge the edges corresponding to the different types of first association relationships and / or second association relationships to obtain merged edges, and determine a candidate multimedia relationship network;

[0209] The merged edge includes the first association relationship and / or the second association relationship of the same type, and the fourth multimedia relationship network is the third multimedia relationship network or the third multimedia relationship network after the nodes are merged;

[0210] The community structure of the candidate multi-media relationship network is identified, and the target nodes of the community structure are determined according to the centrality index to obtain the constructed multi-media relationship network.

[0211] In an exemplary embodiment, the media data includes attribute information of different dimensions of the organization to be identified, and the identification module 808 is further used to determine the behavior label of each node in the multi-media relationship network, and the behavior label value corresponding to the behavior label;

[0212] Based on the medium data, a search is performed in a multi-media relationship network. If there is a target organization and an organization to be identified in the multi-media relationship network that are associated through a target medium, a target community structure to which the target organization belongs in the multi-media relationship network is determined; the target medium is a first type of medium, and / or a second type of medium;

[0213] Determine a target node of a target community structure and a first behavior label and a first label value of the target node; the first label value is used to represent a probability value of the existence of the first behavior in the organization corresponding to the target node;

[0214] According to the association relationship between the target organization and the nodes in the multi-media relationship network, the probability value of the existence of the first behavior in the organization to be identified is determined, and the identification result of the organization to be identified is obtained.

[0215] In an exemplary embodiment, the identification module 808 is further configured to determine a first association weight between the tissue to be identified and the target tissue;

[0216] In the case where an organization corresponding to the target node and the target organization have an association relationship, determining a second association weight between the organization corresponding to the target node and the target organization;

[0217] A second label value of the first behavior propagating to the organization to be identified is determined according to the first association weight, the second association weight and the first label value, where the second label value is used to characterize the probability value of the first behavior existing in the organization to be identified.

[0218] In an exemplary embodiment, the identification module 808 is further configured to determine a third association weight between the organization to be identified and the other target organization if there are other target organizations in the multi-media relationship network that have an association relationship with the organization to be identified through the target medium, and the other target organizations do not belong to the target community structure;

[0219] A third label value of the first behavior propagating to other target organizations is determined according to the third association weight and the second label value, where the third label value is used to characterize a probability value of the first behavior existing in other target organizations.

[0220] In an exemplary embodiment, the target object identification device further includes an update module, which is used to obtain a preset update strategy of the multimedia relationship network; the preset update strategy includes at least one of a time window mechanism, a preset event trigger update, and a historical association retrospective update;

[0221] The multimedia relationship network is updated according to a preset update strategy to obtain an updated multimedia relationship network.

[0222] Each module in the above-mentioned target object recognition device can be implemented in whole or in part by software, hardware or a combination thereof. Each module can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each module above.

[0223] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Fig. 9As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a target object recognition method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0224] Those skilled in the art will understand that Fig. 9 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0225] In one embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above method embodiments when executing the computer program.

[0226] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0227] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0228] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0229] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0230] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A target object recognition method, characterized in that: The method comprises: Acquire historical stock data, where the historical stock data includes multiple organizations and attribute information of different dimensions of each of the organizations; Determine each of the organizations as a node of the initial relationship network, determine the associated organizations associated with the organization through the target medium, and the different levels of association relationships between the organization and the associated organizations according to the attribute information of each of the organizations; wherein the target medium is determined according to the attribute information; According to the association relationships at different levels, the initial relationship network is constructed to obtain a constructed multi-media relationship network; The medium data of the organization to be identified is obtained, and a search is performed in the multi-media relationship network based on the medium data. If there is a target organization in the multi-media relationship network and there is an association relationship between the organization to be identified through the target medium, the identification result of the organization to be identified is determined according to the target organization.

2. The method according to claim 1, characterized in that The attribute information includes static data and historical behavior data of the organization, the historical behavior data includes first historical behavior data and second historical behavior data, the target medium includes first type medium and second type medium, and the determining each of the organizations as a node of the initial relationship network, determining the associated organization associated with the organization through the target medium according to the attribute information of each of the organizations, and the association relationships of different levels between the organization and the associated organizations, includes: Determine each of the organizations as a node of an initial relationship network, determine the static data and the first historical behavior data as a first type of medium, and determine the second historical behavior data as a second type of medium; For each of the organizations, determining a first associated organization having a first associated relationship through the first type of medium, and determining a second associated organization having a second associated relationship through the second type of medium; The correlation degree of the first correlation relationship is greater than the correlation degree of the second correlation relationship.

3. The method according to claim 2, characterized in that The initial relationship network is constructed according to the association relationships at different levels to obtain a constructed multimedia relationship network, including: According to the first association relationship, an edge for connecting the nodes corresponding to the first association relationship in the initial relationship network is generated, and according to the second association relationship, an edge for connecting the nodes corresponding to the second association relationship in the initial relationship network is generated to obtain a first multimedia relationship network; Performing connectivity detection and multi-media information fusion on the first multi-media relationship network to obtain a second multi-media relationship network; The network structure of the second multimedia relationship network is optimized to obtain a constructed multimedia relationship network; the multimedia relationship network includes at least one community structure, and each community structure has a target node.

4. The method according to claim 3, characterized in that The performing connectivity detection and multi-media information fusion on the first multi-media relationship network to obtain a second multi-media relationship network includes: Performing connectivity detection on the first multimedia relationship network to obtain all connected subgraphs in the first multimedia relationship network; Determine the number of nodes in all the connected subgraphs, determine a maximum connected subgraph according to the number of nodes, and determine the maximum connected subgraph as a core network of the first multimedia relationship network; Determine the weights corresponding to different types of edges between the nodes and the centrality index of each node according to the importance of the association relationship between the nodes in the first multimedia relationship network, and obtain a second multimedia relationship network; The nodes are directly or indirectly associated with each other through the first association relationship and / or the second association relationship.

5. The method according to claim 4, characterized in that The step of optimizing the network structure of the second multimedia relationship network to obtain a constructed multimedia relationship network includes: Performing noise filtering on the edges of the second multimedia relationship network according to a preset correlation threshold to obtain a third multimedia relationship network; The third multimedia relationship network is optimized to obtain a constructed multimedia relationship network.

6. The method according to claim 5, characterized in that The performing of network optimization on the third multimedia relationship network to obtain a constructed multimedia relationship network includes: If there is a replaceable node in the node of the third multimedia relationship network, merge the node and the replaceable node to obtain a merged node, and the merged node has an association relationship between the node and the replaceable node; and / or If two nodes in the fourth multimedia relationship network have different types of first association relationships and / or second association relationships, merge the edges corresponding to the different types of first association relationships and / or second association relationships to obtain merged edges, and determine a candidate multimedia relationship network; Wherein, the merged edge includes the first association relationship and / or the second association relationship of the same type, and the fourth multimedia relationship network is the third multimedia relationship network or the third multimedia relationship network after merging nodes; The community structure of the candidate multimedia relationship network is identified, and the target node of the community structure is determined according to the centrality index to obtain a constructed multimedia relationship network.

7. The method according to claim 5, characterized in that The medium data includes attribute information of different dimensions of the organization to be identified, and the retrieval is performed in the multi-media relationship network based on the medium data. If there is a target organization and the organization to be identified in the multi-media relationship network and there is an association relationship through the target medium, then the identification result of the organization to be identified is determined according to the target organization, including: Determining a behavior label of each node in the multimedia relationship network and a behavior label value corresponding to the behavior label; Based on the medium data, the multi-media relationship network is searched, and if there is a target organization in the multi-media relationship network and there is an association relationship between the target organization and the organization to be identified through the target medium, then the target community structure to which the target organization belongs in the multi-media relationship network is determined; the target medium is the first type of medium, and / or the second type of medium; Determine a target node of the target community structure and a first behavior label and a first label value of the target node; the first label value is used to represent a probability value that the organization corresponding to the target node has the first behavior; According to the association relationship between the target organization and the nodes in the multimedia relationship network, the probability value of the existence of the first behavior in the organization to be identified is determined, and the identification result of the organization to be identified is obtained.

8. The method according to claim 7, characterized in that The determining, based on the association relationship between the target organization and the nodes in the multimedia relationship network, a probability value that the to-be-identified organization has the first behavior includes: Determining a first association weight between the to-be-identified organization and the target organization; In the case where an association relationship exists between the organization corresponding to the target node and the target organization, determining a second association weight between the organization corresponding to the target node and the target organization; A second label value of the first behavior propagating to the to-be-identified organization is determined according to the first association weight, the second association weight, and the first label value, wherein the second label value is used to characterize a probability value of the to-be-identified organization having the first behavior.

9. The method according to claim 8, characterized in that The method further comprises: If there are other target organizations in the multi-media relationship network that have an association relationship with the organization to be identified through the target medium, and the other target organizations do not belong to the target community structure, determining a third association weight between the organization to be identified and the other target organizations; A third label value of the first behavior propagating to the other target tissues is determined according to the third association weight and the second label value, wherein the third label value is used to characterize a probability value of the first behavior existing in the other target tissues.

10. The method according to any one of claims 1 to 8, characterized in that: The method further comprises: Obtaining a preset update strategy for the multi-media relationship network; the preset update strategy includes at least one of a time window mechanism, a preset event trigger update, and a historical association retrospective update; The multimedia relationship network is updated according to the preset update strategy to obtain an updated multimedia relationship network.

11. A target object recognition device, characterized in that: The device comprises: A data acquisition module, used to acquire historical stock data, wherein the historical stock data includes multiple organizations and attribute information of different dimensions of each organization; A data analysis module, configured to determine each of the organizations as a node of an initial relationship network, and determine, based on the attribute information of each of the organizations, associated organizations associated with the organization through a target medium, and association relationships of different levels between the organization and the associated organizations; wherein the target medium is determined based on the attribute information; A relationship network construction module, used to construct the initial relationship network according to the association relationships at different levels to obtain a constructed multimedia relationship network; The identification module is used to obtain the media data of the organization to be identified, and to search the multi-media relationship network based on the media data. If there is a target organization in the multi-media relationship network and there is an association relationship between the organization to be identified through the target medium, the identification result of the organization to be identified is determined based on the target organization.

12. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 10 are implemented.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

14. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 10 are implemented.

Images