Identification method, device and equipment, computer readable storage medium and program product

Abstract

The embodiment of the invention provides a recognition method, device and equipment, a computer readable storage medium and a program product, and relates to the fields of artificial intelligence, maps and the like, and application scenes include but are not limited to object recognition scenes. The method comprises the following steps: determining a transaction feature vector of a to-be-identified object, wherein the transaction feature vector of the to-be-identified object is used for representing structured features related to the transaction of the to-be-identified object; based on the transaction feature vector of the to-be-identified object, determining a transaction risk degree of the to-be-identified object, the transaction risk degree of the to-be-identified object being used for representing an initial probability that the to-be-identified object belongs to a risk object type; determining the risk association degree of the to-be-identified object and the risk similarity of the to-be-identified object, wherein each seed object belongs to a risk object type; based on the transaction risk degree of the to-be-identified object, the risk association degree of the to-be-identified object and the risk similarity of the to-be-identified object, determining the probability that the to-be-identified object belongs to the risk object type; therefore, the recognition accuracy of the to-be-recognized object is improved.

Description

Technical Field

[0001] The present application relates to the field of computer technology. Specifically, the present application relates to an identification method, device, equipment, computer-readable storage medium and program product. Background Art

[0002] In the prior art, among the merchants conducting transactions on the platform, some are risk merchants, and it is necessary to identify these risk merchants, that is, it is necessary to identify the merchants conducting transactions on the platform and evaluate whether the merchant belongs to the risk merchant type. For example, to evaluate whether a merchant belongs to the risk merchant type, multiple strategies are summarized based on expert experience and applied to merchant identification. A merchant that hits a certain strategy among the multiple strategies belongs to the risk merchant type. However, for the above-mentioned method of evaluating whether a merchant belongs to the risk merchant type, the relevant features of the merchant used are single, which often leads to low accuracy in identifying merchants. Summary of the Invention

[0003] In view of the disadvantages of the existing method, the present application provides an identification method, device, equipment, computer-readable storage medium and computer program product to solve the problem of how to improve the accuracy of merchant identification.

[0004] In a first aspect, the present application provides an identification method, including:

[0005] Determine the transaction feature vector of the object to be identified, where the transaction feature vector of the object to be identified is used to represent the structured features related to the transaction of the object to be identified;

[0006] Based on the transaction feature vector of the object to be identified, determine the transaction risk degree of the object to be identified, where the transaction risk degree of the object to be identified is used to represent the initial probability that the object to be identified belongs to the risk object type;

[0007] Determine the risk association degree of the object to be identified and the risk similarity degree of the object to be identified. The risk association degree of the object to be identified is used to represent the association relationship between the object to be identified and various sub-objects in the preset seed object set, and the risk similarity degree of the object to be identified is used to represent the similarity between the object to be identified and various sub-objects, and various sub-objects belong to the risk object type;

[0008] Based on the transaction risk degree of the object to be identified, the risk association degree of the object to be identified and the risk similarity degree of the object to be identified, determine the probability that the object to be identified belongs to the risk object type.

[0009] In one embodiment, determining the transaction feature vector of the object to be identified includes:

[0010] Based on a preset set of transaction features, determine the transaction feature values of the object to be identified for each transaction feature in the set of transaction features. The feature type of each transaction feature belongs to any one of the object registration features, object transaction process features, and object transaction detection features;

[0011] Based on the transaction feature values of the object to be identified, determine the transaction feature vector of the object to be identified.

[0012] In one embodiment, determining the transaction risk level of the object to be identified based on the transaction feature vector of the object to be identified includes:

[0013] Input the transaction feature vector of the object to be identified into the first classification model in the target model, and determine the transaction risk level of the object to be identified through classification prediction processing.

[0014] In one embodiment, determining the risk correlation level of the object to be identified includes:

[0015] For each risk correlation feature in the preset set of risk correlation features, determine the correlation relationship between the object to be identified and various sub-objects in the preset set of seed objects;

[0016] Based on the correlation relationships corresponding to the risk correlation features, determine the risk correlation level of the object to be identified.

[0017] In one embodiment, determining the risk similarity of the object to be identified includes:

[0018] For each unstructured feature in the preset set of unstructured features, determine the initial similarity between the object to be identified and various sub-objects;

[0019] Based on the initial similarities, through screening processing, determine the initial risk similarity of the object to be identified for each unstructured feature;

[0020] Based on the initial risk similarities, determine the risk similarity of the object to be identified.

[0021] In one embodiment, for each unstructured feature in the preset set of unstructured features, determining the initial similarity between the object to be identified and various sub-objects includes:

[0022] For each unstructured feature in the set of unstructured features, obtain multiple unstructured feature information of the object to be identified and multiple unstructured feature information of various sub-objects;

[0023] Based on the multiple unstructured feature information of the object to be identified, through screening and splicing processing, determine the first feature embedding vector of the object to be identified;

[0024] Based on the multiple unstructured feature information of various sub-objects, through screening and splicing processing, determine the second feature embedding vectors of various sub-objects;

[0025] Based on the first feature embedding vector of the object to be recognized and the second feature embedding vectors of various sub-objects, determine the initial similarity between the object to be recognized and various sub-objects.

[0026] In one embodiment, based on the transaction risk degree of the object to be recognized, the risk correlation degree of the object to be recognized, and the risk similarity degree of the object to be recognized, determining the probability that the object to be recognized belongs to the risk object type includes:

[0027] Input the transaction risk degree of the object to be recognized, the risk correlation degree of the object to be recognized, and the risk similarity degree of the object to be recognized into the second classification model in the target model, and through classification prediction processing, determine the probability that the object to be recognized belongs to the risk object type.

[0028] In one embodiment, before determining the transaction feature vector of the object to be recognized, it further includes:

[0029] Obtain the initial transaction feature set for multiple object samples;

[0030] For the transaction feature value interval corresponding to any transaction feature in the initial transaction feature set, perform grouping processing to determine multiple grouping intervals;

[0031] Determine the evidence weight corresponding to each grouping interval in the multiple grouping intervals;

[0032] Based on each evidence weight, perform weighted summation processing to determine the information amount corresponding to any transaction feature;

[0033] Based on the information amounts corresponding to the transaction features in the initial transaction feature set, perform sorting processing, and construct the transaction features corresponding to the multiple information amounts ranked ahead into a transaction feature set.

[0034] In one embodiment, before determining the transaction feature vector of the object to be recognized, it further includes:

[0035] Determine the transaction feature vectors of multiple object samples;

[0036] Input the transaction feature vectors of multiple object samples into the first classification model in the target model to be trained, and through classification prediction processing, determine the transaction risk degree of each object sample in the multiple object samples;

[0037] Determine the risk correlation degree of each object sample and the risk similarity degree of each object sample;

[0038] Input the transaction risk degree of each object sample, the risk correlation degree of each object sample, and the risk similarity degree of each object sample into the second classification model in the target model to be trained. Through classification prediction processing, determine the probability that each object sample belongs to the risk object type;

[0039] Based on the probability that each object sample belongs to the risk object type, determine the value of the loss function of the target model to be trained;

[0040] If the value of the loss function of the target model to be trained is greater than the loss threshold, update the network parameters of the target model to be trained;

[0041] Repeat the process of inputting the transaction feature vectors of multiple object samples into the first classification model in the target model to be trained. Through classification prediction processing, determine the transaction risk degree of each object sample among the multiple object samples, input the transaction risk degree of each object sample, the risk correlation degree of each object sample, and the risk similarity degree of each object sample into the second classification model in the target model to be trained. Through classification prediction processing, determine the probability that each object sample belongs to the risk object type, based on the probability that each object sample belongs to the risk object type, determine the value of the loss function of the target model to be trained, and if the value of the loss function of the target model to be trained is greater than the loss threshold, update the network parameters of the target model to be trained, until the value of the loss function of the target model to be trained is equal to the loss threshold to obtain the target model.

[0042] In a second aspect, the present application provides an identification device, including:

[0043] A first processing module for determining the transaction feature vector of the object to be identified, where the transaction feature vector of the object to be identified is used to characterize the structured features related to the transaction of the object to be identified;

[0044] A second processing module for determining the transaction risk degree of the object to be identified based on the transaction feature vector of the object to be identified, where the transaction risk degree of the object to be identified is used to characterize the initial probability that the object to be identified belongs to the risk object type;

[0045] A third processing module for determining the risk correlation degree of the object to be identified and the risk similarity degree of the object to be identified, where the risk correlation degree of the object to be identified is used to characterize the association relationship between the object to be identified and various sub-objects in the preset seed object set, and the risk similarity degree of the object to be identified is used to characterize the similarity between the object to be identified and various sub-objects, and various sub-objects belong to the risk object type;

[0046] A fourth processing module for determining the probability that the object to be identified belongs to the risk object type based on the transaction risk degree of the object to be identified, the risk correlation degree of the object to be identified, and the risk similarity degree of the object to be identified.

[0047] In a third aspect, the present application provides an electronic device, including: a processor, a memory, and a bus;

[0048] The bus is used to connect the processor and the memory;

[0049] The memory is used to store operation instructions;

[0050] The processor is used to execute the recognition method of the first aspect of the present application by calling the operation instructions.

[0051] In a fourth aspect, the present application provides a computer-readable storage medium storing a computer program, and the computer program is used to execute the recognition method of the first aspect of the present application.

[0052] In a fifth aspect, the present application provides a computer program product including a computer program, and when the computer program is executed by a processor, it implements the steps of the recognition method in the first aspect of the present application.

[0053] The technical solutions provided by the embodiments of the present application have at least the following beneficial effects:

[0054] Determine the transaction feature vector of the object to be recognized, where the transaction feature vector of the object to be recognized is used to characterize the structured features related to the transaction of the object to be recognized; based on the transaction feature vector of the object to be recognized, determine the transaction risk degree of the object to be recognized, where the transaction risk degree of the object to be recognized is used to characterize the initial probability that the object to be recognized belongs to the risk object type; determine the risk association degree of the object to be recognized and the risk similarity degree of the object to be recognized, where the risk association degree of the object to be recognized is used to characterize the association relationship between the object to be recognized and various sub-objects in the preset seed object set, and the risk similarity degree of the object to be recognized is used to characterize the similarity between the object to be recognized and various sub-objects, and various sub-objects belong to the risk object type; based on the transaction risk degree of the object to be recognized, the risk association degree of the object to be recognized, and the risk similarity degree of the object to be recognized, determine the probability that the object to be recognized belongs to the risk object type; thus, through the structured features (i.e., the transaction risk degree), the comprehensiveness of the features of the object to be recognized is considered; through the association features (i.e., the risk association degree) and the unstructured features (i.e., the risk similarity degree), the accuracy of the features of the object to be recognized is considered; based on the above-mentioned structured features, association features, and unstructured features, the accuracy of determining the probability that the object to be recognized belongs to the risk object type is improved, that is, the accuracy of recognizing the object to be recognized (such as a merchant) is improved. Description of the Drawings

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for the description in the embodiments of the present application.

[0056] Figure 1 It is a schematic diagram of the network structure of the XGB model;

[0057] Figure 2 It is a schematic architecture diagram of the recognition system provided by the embodiment of the present application;

[0058] Figure 3 It is a schematic flowchart of a recognition method provided by the embodiment of the present application;

[0059] Figure 4 It is a schematic diagram of the recognition provided by the embodiment of the present application;

[0060] Figure 5 It is a schematic diagram of the recognition provided by the embodiment of the present application;

[0061] Figure 6 It is a schematic diagram of the recognition provided by the embodiment of the present application;

[0062] Figure 7 It is a schematic flowchart of a recognition method provided by the embodiment of the present application;

[0063] Figure 8 It is a schematic structural diagram of a recognition device provided by the embodiment of the present application;

[0064] Figure 9 It is a schematic structural diagram of an electronic device provided by the embodiment of the present application. Detailed implementation manners

[0065] The embodiments of the present application will be described below with reference to the accompanying drawings in the present application. It should be understood that the embodiments described below in conjunction with the accompanying drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present application, and do not constitute limitations on the technical solutions of the embodiments of the present application.

[0066] Those skilled in the art of the present technology can understand that unless specifically stated, the singular forms "a", "an", "the" and "said" used herein may also include the plural forms. It should be further understood that the terms "including" and "comprising" used in the embodiments of the present application mean that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements and / or components, but do not exclude implementation as other features, information, data, steps, operations, elements, components and / or their combinations supported by the art of the present technology. It should be understood that when we say an element is "connected" or "coupled" to another element, the one element can be directly connected or coupled to the other element, or it can mean that the one element and the other element establish a connection relationship through an intermediate element. In addition, the "connection" or "coupling" used here can include wireless connection or wireless coupling. The term "and / or" used here indicates at least one of the items defined by the term, for example, "A and / or B" indicates being implemented as "A", or being implemented as "B", or being implemented as "A and B".

[0067] It can be understood that in the specific embodiments of the present application, when it comes to data related to recognition, when the above embodiments of the present application are applied to specific products or technologies, user permission or consent needs to be obtained, and the collection, use, and processing of relevant data need to comply with relevant laws, regulations, and standards in relevant countries and regions.

[0068] To make the objectives, technical solutions, and advantages of the present application clearer, the following will further describe the embodiments of the present application in detail with reference to the accompanying drawings.

[0069] The embodiment of the present application is a recognition method provided by a recognition system, and this recognition method involves fields such as artificial intelligence and maps.

[0070] Artificial Intelligence (AI) is to use a digital computer or a machine controlled by a digital computer to simulate, extend, and expand human intelligence, a theory, method, technology, and application system that can perceive the environment, acquire knowledge, and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science. It attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a way similar to human intelligence. Artificial intelligence is also to study the design principles and implementation methods of various intelligent machines, so that the machines have the functions of perception, reasoning, and decision-making.

[0071] Artificial intelligence technology is an interdisciplinary subject, involving a wide range of fields, including both hardware-level technologies and software-level technologies. The basic technologies of artificial intelligence generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, and mechatronics. The software technologies of artificial intelligence mainly include several major directions such as computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning, autonomous driving, and intelligent transportation.

[0072] Intelligent Traffic System (ITS), also known as Intelligent Transportation System, is to effectively integrate advanced scientific and technological means (information technology, computer technology, data communication technology, sensor technology, electronic control technology, automatic control theory, operations research, artificial intelligence, etc.) into transportation, service control, and vehicle manufacturing, strengthening the connection among vehicles, roads, and users, thereby forming a comprehensive transportation system that ensures safety, improves efficiency, improves the environment, and saves energy.

[0073] To better understand and illustrate the solutions of the embodiments of the present application, the following briefly explains some technical terms involved in the embodiments of the present application.

[0074] WOE: WOE (Weight of Evidence) is a form of encoding for the original independent variable.

[0075] IV: IV (information value) can be used to represent the contribution degree of a feature to target prediction, that is, the prediction ability of the feature; the higher the IV, the stronger the prediction ability of the feature and the higher the information contribution degree.

[0076] XGB model: The network structure of the XGB (eXtreme Gradient Boosting) model is as Figure 1 shown; the XGB model is a forest of trees composed of many trees, and each tree generates the residual of the result by fitting the previous tree. The XGB model is a way of gradient boosting.

[0077] The solution provided in the embodiments of this application relates to artificial intelligence technology. The technical solution of this application will be described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of this application will be described below with reference to the accompanying drawings.

[0078] To better understand the solution provided in the embodiments of this application, the solution will be described below with a specific application scenario.

[0079] In one embodiment, Figure 2 shows a schematic diagram of the architecture of an identification system applicable to the embodiments of this application. It can be understood that the identification method provided in the embodiments of this application can be applicable to but not limited to the application scenarios such as Figure 2 shown.

[0080] In this example, as Figure 2As shown, the architecture of the recognition system in this example may include, but is not limited to, server 10, terminal 20, and database 30. Server 10, terminal 20, and database 30 can interact with each other through network 40. Server 10 determines the transaction feature vector of the object to be recognized, and the transaction feature vector of the object to be recognized is used to characterize the structured features related to the transaction of the object to be recognized; based on the transaction feature vector of the object to be recognized, server 10 determines the transaction risk degree of the object to be recognized, and the transaction risk degree of the object to be recognized is used to characterize the initial probability that the object to be recognized belongs to the risk object type; server 10 determines the risk association degree of the object to be recognized and the risk similarity degree of the object to be recognized. The risk association degree of the object to be recognized is used to characterize the association relationship between the object to be recognized and various sub-objects in the preset seed object set, and the risk similarity degree of the object to be recognized is used to characterize the similarity between the object to be recognized and various sub-objects. Various sub-objects belong to the risk object type; server 10 determines the probability that the object to be recognized belongs to the risk object type based on the transaction risk degree of the object to be recognized, the risk association degree of the object to be recognized, and the risk similarity degree of the object to be recognized; server 10 sends the probability that the object to be recognized belongs to the risk object type to terminal 20 for display, and server 10 sends the probability that the object to be recognized belongs to the risk object type to database 30 for storage.

[0081] It can be understood that the above is only an example, and this embodiment is not limited herein.

[0082] Among them, the terminal includes, but is not limited to, smart phones (such as Android phones, iOS phones, etc.), mobile phone emulators, tablet computers, laptop computers, digital broadcast receivers, MIDs (Mobile Internet Devices), PDAs (Personal Digital Assistants), intelligent voice interaction devices, smart home appliances, vehicle-mounted terminals, etc.

[0083] The server can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server or server cluster that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0084] Cloud computing is a computing model that distributes computing tasks across a resource pool composed of a large number of computing devices, enabling various application systems to obtain computing power, storage space, and information services as needed. The network that provides resources is called the "cloud". The resources in the "cloud" seem to users to be infinitely scalable, and can be obtained at any time, used on demand, expanded at any time, and paid according to usage.

[0085] As a basic capability provider of cloud computing, a cloud computing resource pool (abbreviated as cloud platform, generally called IaaS (Infrastructure as a Service) platform) will be established, and various types of virtual resources will be deployed in the resource pool for external customers to select and use. The cloud computing resource pool mainly includes: computing devices (virtual machines, including operating systems), storage devices, and network devices.

[0086] Logically divided, the PaaS (Platform as a Service) layer can be deployed on the IaaS (Infrastructure as a Service) layer, and the SaaS (Software as a Service) layer can be deployed on top of the PaaS layer. The SaaS layer can also be directly deployed on the IaaS. PaaS is a platform for software operation, such as databases, web containers, etc. SaaS is various business software, such as web portals, SMS mass senders, etc. Generally speaking, SaaS and PaaS are upper layers relative to IaaS.

[0087] The so-called artificial intelligence cloud service is generally also called AIaaS (AI as a Service). This is a current mainstream service method for artificial intelligence platforms. Specifically, the AIaaS platform will split several common AI services and provide independent or packaged services in the cloud. This service model is similar to opening an AI-themed mall: all developers can access and use one or more artificial intelligence services provided by the platform through the API interface. Some senior developers can also use the AI frameworks and AI infrastructure provided by the platform to deploy and operate their own exclusive cloud artificial intelligence services.

[0088] The above network can include but is not limited to: wired networks, wireless networks. Among them, the wired network includes: local area networks, metropolitan area networks, and wide area networks. The wireless network includes: Bluetooth, Wi-Fi, and other networks that implement wireless communication. Specifically, it can also be determined based on the actual application scenario requirements and is not limited here.

[0089] See Figure 3 ,Figure 3 The flowchart of an identification method provided by an embodiment of the present application is shown. Among them, this method can be executed by any electronic device, such as a server, etc. As an alternative embodiment, this method can be executed by a server. For the convenience of description, in the description of some alternative embodiments below, the server will be taken as an example of the execution subject of this method. As Figure 3 shown, the identification method provided by the embodiment of the present application includes the following steps:

[0090] S201. Determine the transaction feature vector of the object to be identified. The transaction feature vector of the object to be identified is used to characterize the structured features related to the transaction of the object to be identified.

[0091] Specifically, the object to be identified is, for example, a merchant. The feature types of the structured features related to the transaction of the object to be identified include pre-entry, in-transaction, and post-audit, etc.

[0092] For example, as Figure 4 shown, pre-entry refers to the information submitted during the merchant entry stage. Pre-entry includes the entry industry, entry time, etc. The entry industry and entry time are both structured features related to the transaction. For example, the entry time can be the time when the merchant accesses the platform, and the entry time can be derived as the number of days from the time point when the merchant starts to be detected. For example, the entry industry can be the industry to which the business engaged by the merchant belongs. The industry is, for example, the automotive industry. For example, the industry is mapped to the corresponding digital identifier in a mapping manner. For example, the automotive industry is mapped to the number 1, the food industry is mapped to the number 2, etc.

[0093] For example, as Figure 4 shown, in-transaction refers to the clue information attached when the merchant conducts a transaction. In-transaction includes large transactions (large transaction amounts), high-frequency transactions (high transaction frequencies), special transactions, transaction fluctuations, transaction patterns, transaction counterparts, etc. Large transactions, high-frequency transactions, special transactions, transaction fluctuations, transaction patterns, transaction counterparts, etc. are all structured features related to the transaction. For example, the transaction pattern can be the payment scenario of the merchant's transaction order. The payment scenario is, for example, code scanning payment, card swiping payment, balance payment, etc. The transaction pattern can be derived as the ratio between the number of payment scenario pens and the amount. For example, the transaction fluctuation can be the change trend of the merchant's transaction amount, order number, user number, etc. The change trend is, for example, a sudden increase type trend. The transaction fluctuation can be derived as the change ratio of the amount, the change ratio of the number of pens, and the change ratio of the user number compared to the previous X months, where X is a positive integer.

[0094] For example, as Figure 4As shown in the figure, post-event review refers to the clue information attached to a merchant after a transaction is investigated or complained about. Post-event review includes the number of investigations, control status, risk labels, complaints, etc. The number of investigations, control status, risk labels, complaints, etc. are all structured features related to the transaction. For example, the number of investigations can be the number of times a merchant has been pushed for investigation historically, and the number of investigations can be derived into whether the merchant has been investigated, the intervals corresponding to different numbers of investigations, etc. For example, a complaint can be the direct feedback of a user to a merchant, and the complaint can be derived into whether the merchant has been complained about, the number of complaints, the proportion of important complaints, etc.

[0095] S202. Based on the transaction feature vector of the object to be identified, determine the transaction risk level of the object to be identified. The transaction risk level of the object to be identified is used to represent the initial probability that the object to be identified belongs to the risk object type.

[0096] Specifically, input the transaction feature vector of the object to be identified into the first classification model in the target model, and through classification prediction processing, determine the transaction risk level of the object to be identified. For example, as Figure 4 shown in the figure, the target model includes two layers of models, which are the first layer model and the second layer model; the first layer model is the first classification model, and the second layer model is the second classification model.

[0097] For example, as Figure 4 shown in the figure, input the transaction feature vector of the object to be identified, which is composed of the structured features related to the transaction corresponding to pre-event settlement, in-event transaction, and post-event review, into the first classification model in the target model for model prediction, that is, through classification prediction processing, determine the transaction risk level of the object to be identified.

[0098] S203. Determine the risk association degree of the object to be identified and the risk similarity degree of the object to be identified. The risk association degree of the object to be identified is used to represent the association relationship between the object to be identified and various sub-objects in the preset seed object set. The risk similarity degree of the object to be identified is used to represent the similarity between the object to be identified and various sub-objects. Various sub-objects belong to the risk object type.

[0099] Specifically, the object to be identified is, for example, the merchant to be identified, and the seed objects in the seed object set are, for example, the seed merchants. For example, as Figure 4 shown in the figure, the risk association features (risk association) corresponding to the risk association degree of the object to be identified include full name / abbreviation, contact information, legal person / supervisor (super administrator), etc.; the risk similarity features (risk similarity) corresponding to the risk similarity degree of the object to be identified include complaints, transaction remarks, business descriptions, etc. The risk similarity features corresponding to the risk similarity degree of the object to be identified are unstructured features.

[0100] For example, the risk association features can be some attributes of a merchant to be identified that are strongly associated with seed merchants. The risk association features include, for example, having the same name. Having the same name can mean that the full name or short name of the merchant to be identified is the same as that of any one of the multiple seed merchants. The risk association features can be derived into whether they have the same name, the number of same names, etc. For example, if the full name and short name of the merchant to be identified are the same as those of any one of the seed merchants, the feature value of the risk association feature corresponding to having the same name is 2; if the full name or short name of the merchant to be identified is the same as that of any one of the seed merchants, the feature value of the risk association feature corresponding to having the same name is 1; if the full name and short name of the merchant to be identified are not the same as those of any one of the seed merchants, the feature value of the risk association feature corresponding to having the same name is 0; that is, the number of same names is 2, 1, or 0. The risk association features include, for example, having the same contact information. Having the same contact information can mean that the contact information of the merchant to be identified is the same as that of any one of the multiple seed merchants. The risk association features can be derived into whether they have the same contact information, the number of same contact information, etc. The risk association features include, for example, having the same legal person / same supervisor. Having the same legal person / same supervisor can mean that the name of the legal person or the name of the supervisor of the merchant to be identified is the same as that of any one of the multiple seed merchants. The risk association features can be derived into whether they have the same legal person, whether they have the same supervisor, etc.

[0101] For example, unstructured features can be unstructured texts, pictures, etc. related to the merchant to be identified; for example, a complaint can be the complaint text of a user regarding a transaction order of the merchant to be identified, a transaction note can be the text note of a transaction order of the merchant to be identified, and an operation description can be the business scope description attached to the merchant to be identified during the settlement and transaction processes.

[0102] S204. Determine the probability that the object to be identified belongs to the risk object type based on the transaction risk degree, risk association degree, and risk similarity degree of the object to be identified.

[0103] Specifically, input the transaction risk degree, risk association degree, and risk similarity degree of the object to be identified into the second classification model in the target model, and through classification prediction processing, determine the probability that the object to be identified belongs to the risk object type. For example, as Figure 4 shown, the target model includes two layers of models, namely the first layer model and the second layer model; the first layer model is the first classification model, and the second layer model is the second classification model.

[0104] For example, as Figure 4 shown, input the transaction risk degree, risk association degree (risk association), and risk similarity degree (risk similarity) of the object to be identified into the second classification model in the target model for model prediction, that is, through classification prediction processing, determine the probability that the object to be identified belongs to the risk object type.

[0105] In the embodiments of the present application, a transaction feature vector of an object to be recognized is determined, and the transaction feature vector of the object to be recognized is used to characterize the structured features related to the transaction of the object to be recognized; based on the transaction feature vector of the object to be recognized, a transaction risk degree of the object to be recognized is determined, and the transaction risk degree of the object to be recognized is used to characterize the initial probability that the object to be recognized belongs to the risk object type; a risk association degree of the object to be recognized and a risk similarity degree of the object to be recognized are determined, the risk association degree of the object to be recognized is used to characterize the association relationship between the object to be recognized and various sub-objects in a preset seed object set, and the risk similarity degree of the object to be recognized is used to characterize the similarity between the object to be recognized and various sub-objects, and various sub-objects belong to the risk object type; based on the transaction risk degree of the object to be recognized, the risk association degree of the object to be recognized, and the risk similarity degree of the object to be recognized, the probability that the object to be recognized belongs to the risk object type is determined; thus, through the structured features (i.e., the transaction risk degree), the comprehensiveness of the features of the object to be recognized is considered; through the association features (i.e., the risk association degree) and the unstructured features (i.e., the risk similarity degree), the accuracy of the features of the object to be recognized is considered; based on the above-mentioned structured features, association features, and unstructured features, the accuracy of determining the probability that the object to be recognized belongs to the risk object type is improved, that is, the accuracy of recognizing the object to be recognized (such as a merchant) is improved.

[0106] In one embodiment, determining the transaction feature vector of the object to be recognized includes steps A1 - A2:

[0107] Step A1, based on a preset transaction feature set, determine the transaction feature value of the object to be recognized for each transaction feature in the transaction feature set, and the feature type of each transaction feature belongs to any one of the object settlement-in features, object transaction process features, and object transaction detection features.

[0108] Specifically, the preset transaction feature set includes multiple transaction features, and each transaction feature in the multiple transaction features is a structured feature related to the transaction of the object to be recognized. The feature types of the transaction features are, for example, object settlement-in features, object transaction process features, object transaction detection features, etc.; among them, the object settlement-in features are, for example, Figure 4 the pre-settlement-in shown in Figure 4 the in-transaction shown in Figure 4 the post-audit shown in

[0109] The transaction feature value of the object to be recognized for each transaction feature in the transaction feature set can be a number. For example, if the transaction feature is the settlement-in time, the transaction feature value of the object to be recognized for the settlement-in time is 30, and 30 indicates that the settlement-in time of the object to be recognized is 30 days.

[0110] Step A2: Determine the transaction feature vector of the object to be recognized based on the transaction feature values of the object to be recognized.

[0111] Specifically, for example, as Figure 5 shown in the feature matrix, F1, F2... Fn represent different transaction features, B1, B2... Bn represent different objects to be recognized. The transaction feature values of object B1 for transaction features F1, F2... Fn are respectively R11, T12... F1n, the transaction feature values of object B2 for transaction features F1, F2... Fn are respectively R21, T22... F2n, and so on. The transaction feature values of object Bn for transaction features F1, F2... Fn are respectively Rn1, Tn2... Fnn; based on the transaction feature values R11, T12... F1n of object B1 to be recognized, determine the transaction feature vector of object B1 to be recognized. The transaction feature vector of object B1 to be recognized includes the transaction feature values R11, T12... F1n; based on the transaction feature values R21, T22... F2n of object B2 to be recognized, determine the transaction feature vector of object B2 to be recognized. The transaction feature vector of object B2 to be recognized includes the transaction feature values R21, T22... F2n; and so on. Based on the transaction feature values Rn1, Tn2... Fnn of object Bn to be recognized, determine the transaction feature vector of object Bn to be recognized. The transaction feature vector of object Bn to be recognized includes the transaction feature values Rn1, Tn2... Fnn.

[0112] In one embodiment, determining the transaction risk level of the object to be recognized includes:

[0113] Input the transaction feature vector of the object to be recognized into the first classification model in the target model, and determine the transaction risk level of the object to be recognized through classification prediction processing.

[0114] Specifically, for example, as Figure 4 shown, the target model includes two-layer models, which are the first-layer model and the second-layer model; the first-layer model is the first classification model, and the second-layer model is the second classification model. The first classification model is, for example, an XGB model, and the second classification model is, for example, another XGB model.

[0115] For example, input the transaction feature vector of the object to be recognized into the first classification model in the target model, and through classification prediction processing, determine the initial probability that the object to be recognized belongs to the risk object type. The initial probability that the object to be recognized belongs to the risk object type is, for example, 5%, that is, the transaction risk level of the object to be recognized is 5%.

[0116] In one embodiment, determining the risk correlation degree of the object to be recognized includes steps B1 - B2:

[0117] Step B1: For each risk association feature in the preset risk association feature set, determine the association relationships between the object to be identified and various sub-objects in the preset seed object set.

[0118] Specifically, for example, as Figure 4 shown, the risk association features (risk associations) corresponding to the risk association degree of the object to be identified include full name / abbreviation, contact information, legal person / super administrator (supervisor), etc. For example, the risk association features can be some attributes of strong association between the merchant to be identified and the seed merchants. Risk association features such as having the same name, having the same name can be that the full name or abbreviation of the merchant to be identified is the same as that of any one of the multiple seed merchants; the risk association feature can be derived into whether having the same name, the number of same names, etc. For example, if the full name and abbreviation of the merchant to be identified are the same as those of any one of the seed merchants, the feature value of the risk association feature corresponding to having the same name is 2; if the full name or abbreviation of the merchant to be identified is the same as that of any one of the seed merchants, the feature value of the risk association feature corresponding to having the same name is 1; if the full name and abbreviation of the merchant to be identified are not the same as those of any one of the seed merchants, the feature value of the risk association feature corresponding to having the same name is 0; that is, the number of same names is 2, 1, or 0. Risk association features such as having the same contact information, having the same contact information can be that the contact information of the merchant to be identified is the same as that of any one of the multiple seed merchants; the risk association feature can be derived into whether having the same contact information, the number of same contact information, etc. Risk association features such as having the same legal person / same supervisor, having the same legal person / same supervisor can be that the legal person name or supervisor name of the merchant to be identified is the same as that of any one of the multiple seed merchants; the risk association feature can be derived into whether having the same legal person, whether having the same supervisor, etc.

[0119] For example, the risk association degree of the object to be identified corresponds to a feature matrix n*m. In the feature matrix n*m, n (rows) represents n risk association features, and m (columns) represents m seed merchants, where both n and m are positive integers; the values of the elements in the feature matrix n*m can be 0, 1, etc., and the values of the elements are the feature values of the risk association features.

[0120] For example, the risk association feature is having the same contact information. For having the same contact information, determine the association relationships between the object to be identified and seed object 1, seed object 2, and seed object 3 respectively; if the contact information between the object to be identified and seed object 1 is the same, the value of the corresponding element in the feature matrix n*m is 1, where 1 represents the same, and 1 represents the association relationship between the object to be identified and seed object 1 respectively; if the contact information between the object to be identified and seed object 2 is different, the value of the corresponding element in the feature matrix n*m is 0, where 0 represents different, and 0 represents the association relationship between the object to be identified and seed object 1 respectively; if the contact information between the object to be identified and seed object 3 is the same, the value of the corresponding element in the feature matrix n*m is 1, where 1 represents the same, and 1 represents the association relationship between the object to be identified and seed object 1 respectively.

[0121] Step B2: Determine the risk correlation degree of the object to be identified based on the correlation relationships corresponding to each risk correlation feature.

[0122] Specifically, the risk correlation degree of the object to be identified corresponds to a feature matrix of n*m; based on the correlation relationships corresponding to each risk correlation feature, that is, based on the values of the elements in the feature matrix of n*m, determine the risk correlation degree of the object to be identified, that is, determine the feature matrix of n*m; the values of the elements in the feature matrix of n*m are, for example, 0, 1, etc.

[0123] In one embodiment, the risk correlation feature can be an unstructured feature, and the unstructured feature can be characterized by similarity.

[0124] In one embodiment, determining the risk similarity of the object to be identified includes steps C1-C3:

[0125] Step C1: For each unstructured feature in the preset unstructured feature set, determine the initial similarity between the object to be identified and various sub-objects.

[0126] Specifically, for example, as Figure 6 shown, a certain unstructured feature in the preset unstructured feature set is a transaction note; the transaction note list of the object to be identified includes multiple transaction notes, and the multiple transaction notes can be the transaction notes in the past six months. The multiple transaction notes correspond to N texts, which are text 1, text 2... text N respectively, and N is a positive integer; based on these N texts, perform denoising processing to remove the notes automatically associated with the transaction in these N texts (such as order numbers, merchant names, etc.) to obtain the N texts after denoising processing; based on the N texts after denoising processing, perform text deduplication processing to obtain multiple texts after deduplication; sort based on the text occurrence frequency, and obtain the top 10 texts sorted from the multiple texts after deduplication, that is, TOP10 texts; splice the top 10 texts sorted to obtain a spliced text; perform vectorization representation on the spliced text, and the semantic representation of the spliced text is the mean of the semantic representations of the words included in the spliced text, that is, tokenize the sentences in the spliced text, perform an embedding (word vector) query on each word in the sentence (for example, based on a preset dictionary, obtain the word vector corresponding to the word) to obtain the word vector of each word, and the embedding (feature embedding vector) of the sentence is the sum of the embeddings (word vectors) of all the words in the sentence; the feature embedding vector of the spliced text corresponding to the object to be identified is calculated through formulas (1) and (2), and formulas (1) and (2) are as follows:

[0127]

[0128]

[0129] Where, V in formula (1) i Represents the concatenated text S i The feature embedding vector of ij Indicates S i The word vector of the jth word in the sentence; m represents S i There are m words in total after word segmentation; Formula (2) represents the i Perform regularization.

[0130] For example, Figure 6 As shown, the transaction note list of any seed object among the M seed objects includes multiple transaction notes, and the multiple transaction notes can be transaction notes in the past six months. The multiple transaction notes correspond to N texts, and the N texts are text 1, text 2...text N, respectively, where N is a positive integer; based on the N texts, denoising is performed to remove the notes automatically associated with the transaction in the N texts (such as order number, merchant name, etc.) to obtain the N texts after denoising; based on the N texts after denoising, text deduplication is performed to obtain multiple texts after deduplication; based on the text occurrence frequency, the texts are sorted to obtain the top 10 texts from the multiple texts after deduplication, namely, T OP10 text; the top 10 texts are concatenated to obtain a concatenated text; the concatenated text is vectorized, and the semantic representation of the concatenated text is the mean of the semantic representations of the words contained in the concatenated text, that is, the sentences in the concatenated text are segmented, and each word in the sentence is queried for embedding (word vector) (for example, based on a preset dictionary, the word vector corresponding to the word is obtained), and the word vector of each word is obtained. The embedding (feature embedding vector) of the sentence is the sum of the embeddings (word vectors) of all the words in the sentence; the feature embedding vector of the concatenated text corresponding to the seed object is calculated by formula (1) and formula (2).

[0131] For example, the preset seed object set is denoted as A, and the text representation set corresponding to the preset seed object set is denoted as V A , V Ai is denoted as the feature embedding vector of the concatenated text of the i-th seed object; the set of objects to be identified is denoted as B, and the set of text representations corresponding to the set of objects to be identified is denoted as V B , V Bj V is the feature embedding vector of the concatenated text of the jth object to be identified; V is the feature embedding vector of the i-th seed object Ai and the object to be identified V Bj The similarity matrix W between ij The calculation formula (3) is as follows:

[0132]

[0133] Among them, W ijDenote the i-th seed object as V Ai and the object V to be recognized Bj The similarity score therebetween, i.e., the initial similarity between the i-th seed object V Ai and the object V to be recognized Bj For example, as shown in Figure 6 the similarity score in the similarity score list is W ij .

[0134] Step C2: Based on each initial similarity, through screening processing, determine the initial risk similarity of the object to be recognized for each unstructured feature.

[0135] Specifically, for example, as shown in Figure 6 Based on each initial similarity (similarity score list), determine the initial risk similarity T of the object to be recognized for each unstructured feature Bj is the maximum initial similarity among the initial similarities W ij The calculation formula (4) of the initial risk similarity T Bj is as follows:[[]]

[0136] T Bj = max(∑ i∈A W ij ) Formula (4)

[0137] where W ij represents the similarity score between the i-th seed object V Ai and the object V to be recognized Bj i.e., the initial similarity between the i-th seed object V Ai and the object V to be recognized Bj For example, as shown in Figure 6 the maximum similarity score is T Bj .

[0138] Step C3: Based on each initial risk similarity, determine the risk similarity of the object to be recognized.

[0139] Specifically, the risk similarity of the object to be recognized includes each initial risk similarity. For example, the risk similarity of the object to be recognized is a feature matrix, each initial risk similarity is an element in the feature matrix, and the value of the initial risk similarity is an eigenvalue.

[0140] In one embodiment, for each unstructured feature in the preset unstructured feature set, determining the initial similarity between the object to be recognized and various sub-objects includes steps D1 - D4:

[0141] Step D1: For each unstructured feature in the unstructured feature set, obtain multiple unstructured feature information for the object to be recognized and multiple unstructured feature information for various sub-objects.

[0142] Specifically, the unstructured feature information is, for example, text, pictures, etc.

[0143] For example, as Figure 6 shown, a certain unstructured feature in the unstructured feature set is a transaction note; the transaction note list of the object to be recognized includes multiple transaction notes, and the multiple transaction notes can be transaction notes in the recent half year. The multiple transaction notes correspond to N texts, which are text 1, text 2... text N respectively, where N is a positive integer. The multiple unstructured feature information of the object to be recognized is these N texts.

[0144] For example, as Figure 6 shown, the transaction note list of any one of the M seed objects includes multiple transaction notes, and the multiple transaction notes can be transaction notes in the recent half year. The multiple transaction notes correspond to N texts, which are text 1, text 2... text N respectively, where N is a positive integer. The multiple unstructured feature information of any one of the seed objects is these N texts.

[0145] Step D2: Based on the multiple unstructured feature information of the object to be recognized, through screening and splicing processing, determine the first feature embedding vector of the object to be recognized.

[0146] Specifically, as Figure 6 shown, based on the N texts, perform denoising processing to remove the remarks (such as order numbers, merchant names, etc.) automatically associated with the transactions in these N texts, and obtain the N texts after denoising processing; based on the N texts after denoising processing, perform text deduplication processing to obtain the multiple texts after deduplication; sort based on the text occurrence frequency, and obtain the top 10 texts ranked in the front from the multiple texts after deduplication, that is, the TOP10 texts; splice the top 10 texts ranked in the front to obtain a spliced text; perform vector representation on the spliced text, and calculate the feature embedding vector of the spliced text corresponding to the object to be recognized through formula (1) and formula (2), that is, the first feature embedding vector of the object to be recognized.

[0147] Step D3: Based on the multiple unstructured feature information of various sub-objects, through screening and splicing processing, determine the second feature embedding vector of various sub-objects.

[0148] Specifically, for example, as Figure 6As shown in the figure, based on N texts, denoising processing is performed to remove the remarks (such as order numbers, merchant names, etc.) that are automatically associated with transactions in these N texts, obtaining N denoised texts; based on the N denoised texts, text deduplication processing is performed to obtain multiple deduplicated texts; sorting is performed based on the text occurrence frequency, and the top 10 texts, that is, the TOP10 texts, are obtained from the multiple deduplicated texts; the top 10 sorted texts are concatenated to obtain a concatenated text; vector representation is performed on the concatenated text, and the feature embedding vector of the concatenated text corresponding to any seed object is calculated through formula (1) and formula (2), that is, the second feature embedding vector of any seed object.

[0149] Step D4, based on the first feature embedding vector of the object to be recognized and the second feature embedding vectors of various seed objects, determine the initial similarity between the object to be recognized and various seed objects.

[0150] Specifically, for example, the preset seed object set is denoted as A, and the text representation set corresponding to the preset seed object set is denoted as V A , V Ai is denoted as the feature embedding vector of the concatenated text of the i-th seed object; the object set to be recognized is denoted as B, and the text representation set corresponding to the object set to be recognized is denoted as V B , V Bj is denoted as the feature embedding vector of the concatenated text of the j-th object to be recognized; the similarity matrix W between the i-th seed object V Ai and the object to be recognized V Bj is calculated through formula (3); W ij , W ij represents the similarity score between the i-th seed object V Ai and the object to be recognized V Bj , that is, the initial similarity between the i-th seed object V Ai and the object to be recognized V Bj . For example, as Figure 6 shown, the similarity score in the similarity score list is W ij .

[0151] In one embodiment, based on the transaction risk degree, risk association degree, and risk similarity of the object to be recognized, determining the probability that the object to be recognized belongs to the risk object type includes:

[0152] Input the transaction risk degree, risk association degree, and risk similarity of the object to be recognized into the second classification model in the target model, and through classification prediction processing, determine the probability that the object to be recognized belongs to the risk object type.

[0153] Specifically, based on the transaction risk level, risk correlation degree, and risk similarity degree of the object to be recognized, splicing and fusion processing is performed to obtain a vector after splicing and fusion processing; the vector after splicing and fusion processing is input into the second classification model in the target model, and through classification prediction processing, the probability that the object to be recognized belongs to the risk object type is determined.

[0154] For example, the target model includes two layers of models. The first layer of model is the first classification model, and the second layer of model is the second classification model. The first classification model is, for example, an XGB model, and the second classification model is, for example, another XGB model; the probability that the object to be recognized belongs to the risk object type is, for example, 95%.

[0155] In one embodiment, before determining the transaction feature vector of the object to be recognized, steps E1 - E5 are further included:

[0156] Step E1, obtain the initial transaction feature set for multiple object samples.

[0157] Specifically, the object sample is, for example, a merchant sample. For example, the initial transaction feature set includes several hundred transaction features.

[0158] Step E2, for the transaction feature value interval corresponding to any transaction feature in the initial transaction feature set, perform grouping processing to determine multiple grouping intervals.

[0159] Specifically, the transaction feature value interval corresponding to any transaction feature includes multiple transaction feature values. For example, as shown in Table 1, the transaction feature value interval corresponding to transaction feature 1 is (0.999, 403.0]. For the transaction feature value interval (0.999, 403.0] corresponding to transaction feature 1, grouping (binning) is performed to obtain multiple grouping intervals. The grouping intervals are, for example, (16.3, 30.6], (38.9, 43.6], etc.

[0160] Table 1: Grouping (binning) for transaction feature 1

[0161]

[0162]

[0163] Step E3, determine the evidence weight corresponding to each grouping interval among the multiple grouping intervals.

[0164] Specifically, as shown in Table 1 for example, based on the good sample concentration and bad sample concentration, determine the evidence weight corresponding to each grouping interval of transaction feature 1; for example, the evidence weight WOE corresponding to the grouping interval (16.3, 30.6] is 0.44, and the evidence weight WOE corresponding to the grouping interval (38.9, 43.6] is 2.24.

[0165] Take the grouping interval with the largest weight of evidence WOE as the threshold interval for any transaction feature; for example, the weight of evidence WOE corresponding to (38.9, 43.6] is 2.24, and (38.9, 43.6] is the grouping interval with the largest weight of evidence WOE, so (38.9, 43.6] is taken as the threshold interval for transaction feature 1.

[0166] It should be noted that good samples are object samples belonging to the non-risk object type, and object samples belonging to the non-risk object type are merchant samples belonging to the non-risk merchant type, for example; bad samples are object samples belonging to the risk object type, and object samples belonging to the risk object type are merchant samples belonging to the risk merchant type, for example. For example, the number of good samples in (0.999, 16.3] is 4, and the total number of good samples in (0.999, 403.0] is 100, so the concentration of good samples corresponding to (0.999, 16.3] is 4 / 100 = 0.04. The number of bad samples in (0.999, 16.3] is 15, and the total number of bad samples in (0.999, 403.0] is 100, so the concentration of bad samples corresponding to (0.999, 16.3] is 15 / 100 = 0.15.

[0167] Step E4: Based on each weight of evidence, perform a weighted summation process to determine the amount of information corresponding to any transaction feature.

[0168] Specifically, for example, perform a weighted summation process on each weight of evidence WOE shown in Table 1 to determine the amount of information IV corresponding to transaction feature 1.

[0169] Step E5: Based on the amount of information corresponding to each transaction feature in the initial transaction feature set, perform a sorting process, and construct a transaction feature set from the transaction features corresponding to the top several amounts of information.

[0170] Specifically, for example, the initial transaction feature set includes 300 transaction features, and these 300 transaction features correspond to 300 amounts of information. Sort these 300 amounts of information from largest to smallest, and construct a transaction feature set from the transaction features corresponding to the top 100 amounts of information, that is, the transaction feature set includes the transaction features corresponding to the top 100 amounts of information. For example, the top 10 amounts of information are shown in Table 2.

[0171] Table 2: The top 10 amounts of information

[0172] Transaction characteristics Information volume IV Transaction characteristic 1 5.167103 Transaction characteristic 2 5.167103 Transaction characteristic 3 4.688601 Transaction characteristic 4 4.538491 Transaction characteristic 5 4.538491 Transaction characteristic 6 3.925617 Transaction characteristic 7 3.721977 Transaction characteristic 8 3.721977 Transaction characteristic 9 3.684654 Transaction characteristic 10 3.64114

[0173] In one embodiment, before determining the transaction feature vector of the object to be identified, it further includes:

[0174] Determine the transaction feature vectors of multiple object samples;

[0175] Input the transaction feature vectors of multiple object samples into the first classification model of the target model to be trained, and determine the transaction risk level of each object sample in the multiple object samples through classification prediction processing;

[0176] Determine the risk correlation degree of each object sample and the risk similarity degree of each object sample;

[0177] Input the transaction risk level of each object sample, the risk correlation degree of each object sample, and the risk similarity degree of each object sample into the second classification model of the target model to be trained, and determine the probability that each object sample belongs to the risk object type through classification prediction processing;

[0178] Based on the probabilities that each object sample belongs to the risk object type, determine the value of the loss function of the target model to be trained;

[0179] If the value of the loss function of the target model to be trained is greater than the loss threshold, update the network parameters of the target model to be trained;

[0180] Repeat the operations of inputting the transaction feature vectors of multiple object samples into the first classification model of the target model to be trained, determining the transaction risk level of each object sample in the multiple object samples through classification prediction processing, inputting the transaction risk level of each object sample, the risk correlation degree of each object sample, and the risk similarity degree of each object sample into the second classification model of the target model to be trained, determining the probability that each object sample belongs to the risk object type through classification prediction processing, based on the probabilities that each object sample belongs to the risk object type, determining the value of the loss function of the target model to be trained, and if the value of the loss function of the target model to be trained is greater than the loss threshold, updating the network parameters of the target model to be trained until the value of the loss function of the target model to be trained is equal to the loss threshold to obtain the target model.

[0181] Specifically, construct a set of transaction features; based on the set of transaction features, determine the transaction feature values of the object samples for each transaction feature in the set of transaction features, where the feature type of each transaction feature belongs to any one of object settlement features, object transaction process features, and object transaction detection features; based on the transaction feature values of the object samples, determine the transaction feature vector of the object samples; input the transaction feature vector of the object samples into the first classification model in the target model to be trained, and through classification prediction processing, determine the transaction risk level of the object samples; for each risk association feature in the risk association feature set, determine the association relationship between the object samples and various sub-objects in the seed object set; based on the association relationships corresponding to each risk association feature, determine the risk association level of the object samples; for each unstructured feature in the unstructured feature set, determine the initial similarity between the object samples and various sub-objects; based on each initial similarity, through screening processing, determine the initial risk similarity of the object samples for each unstructured feature; based on each initial risk similarity, determine the risk similarity of the object samples; input the transaction risk level of the object samples, the risk association level of the object samples, and the risk similarity of the object samples into the second classification model in the target model to be trained, and through classification prediction processing, determine the probability that the object samples belong to the risk object type.

[0182] Applying the embodiments of the present application has at least the following beneficial effects:

[0183] Through the structured features (i.e., transaction risk level), the comprehensiveness of the features of the object to be identified is considered; through the association features (i.e., risk association level) and unstructured features (i.e., risk similarity), the accuracy of the features of the object to be identified is considered; based on the above-mentioned structured features, association features, and unstructured features, the accuracy of determining the probability that the object to be identified belongs to the risk object type is improved, that is, the accuracy of identifying the object to be identified (such as a merchant) is improved.

[0184] To better understand the method provided by the embodiments of the present application, the following further illustrates the solution of the embodiments of the present application with examples of specific application scenarios.

[0185] In an embodiment of a specific application scenario, for example, in an object recognition scenario, refer to Figure 7 , which shows the processing flow of an identification method. As Figure 7 shown, the processing flow of the identification method provided by the embodiments of the present application includes the following steps:

[0186] S701, the server constructs a set of transaction features.

[0187] Specifically, obtain an initial transaction feature set for multiple object samples; perform grouping processing on the transaction feature value intervals corresponding to any transaction feature in the initial transaction feature set to determine multiple grouping intervals; determine the evidence weight corresponding to each grouping interval in the multiple grouping intervals; perform weighted summation processing based on each evidence weight to determine the information amount corresponding to any transaction feature; perform sorting processing based on the information amounts corresponding to each transaction feature in the initial transaction feature set, and construct the transaction features corresponding to the multiple information amounts ranked ahead into a transaction feature set.

[0188] S702, the server trains the target model to be trained based on the transaction feature set, the risk association feature set of the object sample, and the unstructured feature set of the object sample to obtain the target model.

[0189] Specifically, based on the transaction feature set, determine the transaction feature values of the object samples for each transaction feature in the transaction feature set, and the feature type of each transaction feature belongs to any one of the object settlement features, object transaction process features, and object transaction detection features; determine the transaction feature vector of the object sample based on the transaction feature values of the object sample; input the transaction feature vector of the object sample into the first classification model in the target model to be trained, and determine the transaction risk degree of the object sample through classification prediction processing; for each risk association feature in the risk association feature set, determine the association relationship between the object sample and each sub-object in the seed object set; determine the risk association degree of the object sample based on the association relationships corresponding to each risk association feature; for each unstructured feature in the unstructured feature set, determine the initial similarity between the object sample and each sub-object; determine the initial risk similarity of the object sample for each unstructured feature through screening processing based on each initial similarity; determine the risk similarity of the object sample based on each initial risk similarity; input the transaction risk degree of the object sample, the risk association degree of the object sample, and the risk similarity of the object sample into the second classification model in the target model to be trained, and determine the probability that the object sample belongs to the risk object type through classification prediction processing.

[0190] Determine the value of the loss function of the target model to be trained based on the probabilities of each object sample belonging to the risk object type; if the value of the loss function of the target model to be trained is greater than the loss threshold, update the network parameters of the target model to be trained; repeatedly execute the steps of inputting the transaction feature vectors of multiple object samples into the first classification model in the target model to be trained, determining the transaction risk degree of each object sample in the multiple object samples through classification prediction processing, inputting the transaction risk degree of each object sample, the risk association degree of each object sample, and the risk similarity degree of each object sample into the second classification model in the target model to be trained, determining the probability of each object sample belonging to the risk object type through classification prediction processing, determining the value of the loss function of the target model to be trained based on the probabilities of each object sample belonging to the risk object type, and if the value of the loss function of the target model to be trained is greater than the loss threshold, updating the network parameters of the target model to be trained, until the value of the loss function of the target model to be trained is equal to the loss threshold, and obtaining the target model.

[0191] S703, the server obtains the object to be identified.

[0192] Specifically, the object to be identified is, for example, the merchant to be identified.

[0193] S704, the server determines the transaction feature values of the object to be identified for each transaction feature in the transaction feature set based on the transaction feature set, and determines the transaction feature vector of the object to be identified based on the transaction feature values of the object to be identified.

[0194] Specifically, the feature type of each transaction feature belongs to any one of the object settlement features, the object transaction process features, and the object transaction detection features.

[0195] S705, the server inputs the transaction feature vector of the object to be identified into the first classification model in the target model, and determines the transaction risk degree of the object to be identified through classification prediction processing.

[0196] Specifically, the transaction risk degree of the object to be identified is used to represent the initial probability of the object to be identified belonging to the risk object type.

[0197] S706, for each risk association feature in the risk association feature set, the server determines the association relationship between the object to be identified and each sub-object in the preset seed object set; and determines the risk association degree of the object to be identified based on the association relationships corresponding to the risk association features.

[0198] S707, the server determines the risk similarity degree of the object to be identified.

[0199] Specifically, for each unstructured feature in the preset unstructured feature set, determine the initial similarity between the object to be identified and various sub-objects; based on the initial similarities, through screening processing, determine the initial risk similarity of the object to be identified for each unstructured feature; based on the initial risk similarities, determine the risk similarity of the object to be identified.

[0200] S708. The server performs splicing and fusion processing based on the transaction risk degree of the object to be identified, the risk correlation degree of the object to be identified, and the risk similarity of the object to be identified, and obtains a vector after the splicing and fusion processing.

[0201] S709. The server inputs the vector after the splicing and fusion processing into the second classification model in the target model, and through classification prediction processing, determines the probability that the object to be identified belongs to the risk object type.

[0202] Applying the embodiments of the present application has at least the following beneficial effects:

[0203] Through the structured feature (i.e., the transaction risk degree), the comprehensiveness of the features of the object to be identified is considered; through the associated feature (i.e., the risk correlation degree) and the unstructured feature (i.e., the risk similarity), the accuracy of the features of the object to be identified is considered; based on the above-mentioned structured feature, associated feature and unstructured feature, the accuracy of determining the probability that the object to be identified belongs to the risk object type is improved, that is, the accuracy of identifying the object to be identified (such as a merchant) is improved.

[0204] The embodiments of the present application further provide an identification device, and the structural schematic diagram of the identification device is as Figure 8 shown. The identification device 80 includes a first processing module 801, a second processing module 802, a third processing module 803, and a fourth processing module 804.

[0205] The first processing module 801 is configured to determine the transaction feature vector of the object to be identified, and the transaction feature vector of the object to be identified is used to characterize the structured features related to the transaction of the object to be identified;

[0206] The second processing module 802 is configured to determine the transaction risk degree of the object to be identified based on the transaction feature vector of the object to be identified, and the transaction risk degree of the object to be identified is used to characterize the initial probability that the object to be identified belongs to the risk object type;

[0207] The third processing module 803 is configured to determine the risk correlation degree of the object to be identified and the risk similarity of the object to be identified. The risk correlation degree of the object to be identified is used to characterize the association relationship between the object to be identified and various sub-objects in the preset seed object set, and the risk similarity of the object to be identified is used to characterize the similarity between the object to be identified and various sub-objects, and various sub-objects belong to the risk object type;

[0208] The fourth processing module 804 is configured to determine the probability that the object to be recognized belongs to the risk object type based on the transaction risk degree of the object to be recognized, the risk correlation degree of the object to be recognized, and the risk similarity degree of the object to be recognized.

[0209] In one embodiment, the first processing module 801 is specifically configured to:

[0210] Based on a preset transaction feature set, determine the transaction feature values of the object to be recognized for each transaction feature in the transaction feature set, and the feature type of each transaction feature belongs to any one of the object settlement-in features, object transaction process features, and object transaction detection features;

[0211] Based on the transaction feature values of the object to be recognized, determine the transaction feature vector of the object to be recognized.

[0212] In one embodiment, the second processing module 802 is specifically configured to:

[0213] Input the transaction feature vector of the object to be recognized into the first classification model in the target model, and determine the transaction risk degree of the object to be recognized through classification prediction processing.

[0214] In one embodiment, the third processing module 803 is specifically configured to:

[0215] For each risk correlation feature in the preset risk correlation feature set, determine the association relationship between the object to be recognized and various sub-objects in the preset seed object set;

[0216] Based on the association relationships corresponding to the risk correlation features, determine the risk correlation degree of the object to be recognized.

[0217] In one embodiment, the third processing module 803 is specifically configured to:

[0218] For each unstructured feature in the preset unstructured feature set, determine the initial similarity between the object to be recognized and various sub-objects;

[0219] Based on the initial similarities, through screening processing, determine the initial risk similarity of the object to be recognized for each unstructured feature;

[0220] Based on the initial risk similarities, determine the risk similarity degree of the object to be recognized.

[0221] In one embodiment, the third processing module 803 is specifically configured to:

[0222] For each unstructured feature in the unstructured feature set, obtain multiple unstructured feature information of the object to be recognized and multiple unstructured feature information of various sub-objects;

[0223] Based on multiple unstructured feature information of the object to be recognized, through screening and splicing processing, determine the first feature embedding vector of the object to be recognized;

[0224] Based on multiple unstructured feature information of various sub-objects, through screening and splicing processing, determine the second feature embedding vector of various sub-objects;

[0225] Based on the first feature embedding vector of the object to be recognized and the second feature embedding vectors of various sub-objects, determine the initial similarity between the object to be recognized and various sub-objects.

[0226] In one embodiment, the fourth processing module 804 is specifically configured to:

[0227] Input the transaction risk degree, risk correlation degree, and risk similarity of the object to be recognized into the second classification model in the target model, and through classification prediction processing, determine the probability that the object to be recognized belongs to the risk object type.

[0228] In one embodiment, the first processing module 801 is further configured to:

[0229] Obtain an initial transaction feature set for multiple object samples;

[0230] For the transaction feature value interval corresponding to any transaction feature in the initial transaction feature set, perform grouping processing to determine multiple grouping intervals;

[0231] Determine the evidence weight corresponding to each grouping interval in the multiple grouping intervals;

[0232] Based on each evidence weight, perform weighted summation processing to determine the information amount corresponding to any transaction feature;

[0233] Based on the information amounts corresponding to each transaction feature in the initial transaction feature set, perform sorting processing, and construct the transaction features corresponding to the multiple information amounts ranked ahead into a transaction feature set.

[0234] In one embodiment, the first processing module 801 is further configured to:

[0235] Determine the transaction feature vectors of multiple object samples;

[0236] Input the transaction feature vectors of multiple object samples into the first classification model in the target model to be trained, and through classification prediction processing, determine the transaction risk degree of each object sample in the multiple object samples;

[0237] Determine the risk correlation degree of each object sample and the risk similarity of each object sample;

[0238] Input the transaction risk degree of each object sample, the risk correlation degree of each object sample, and the risk similarity degree of each object sample into the second classification model of the target model to be trained. Through classification prediction processing, determine the probability that each object sample belongs to the risk object type;

[0239] Based on the probability that each object sample belongs to the risk object type, determine the value of the loss function of the target model to be trained;

[0240] If the value of the loss function of the target model to be trained is greater than the loss threshold, update the network parameters of the target model to be trained;

[0241] Repeat the steps of inputting the transaction feature vectors of multiple object samples into the first classification model of the target model to be trained. Through classification prediction processing, determine the transaction risk degree of each object sample among the multiple object samples, input the transaction risk degree of each object sample, the risk correlation degree of each object sample, and the risk similarity degree of each object sample into the second classification model of the target model to be trained. Through classification prediction processing, determine the probability that each object sample belongs to the risk object type, based on the probability that each object sample belongs to the risk object type, determine the value of the loss function of the target model to be trained, and if the value of the loss function of the target model to be trained is greater than the loss threshold, update the network parameters of the target model to be trained until the value of the loss function of the target model to be trained is equal to the loss threshold to obtain the target model.

[0242] Applying the embodiments of the present application has at least the following beneficial effects:

[0243] Determine the transaction feature vector of the object to be recognized, where the transaction feature vector of the object to be recognized is used to characterize the structured features related to the transaction of the object to be recognized; based on the transaction feature vector of the object to be recognized, determine the transaction risk level of the object to be recognized, where the transaction risk level of the object to be recognized is used to characterize the initial probability that the object to be recognized belongs to the risk object type; determine the risk correlation degree of the object to be recognized and the risk similarity degree of the object to be recognized, where the risk correlation degree of the object to be recognized is used to characterize the association relationship between the object to be recognized and various sub-objects in the preset seed object set, and the risk similarity degree of the object to be recognized is used to characterize the similarity between the object to be recognized and various sub-objects, and various sub-objects belong to the risk object type; based on the transaction risk level of the object to be recognized, the risk correlation degree of the object to be recognized, and the risk similarity degree of the object to be recognized, determine the probability that the object to be recognized belongs to the risk object type; in this way, through the structured features (i.e., the transaction risk level), the comprehensiveness of the features of the object to be recognized is considered; through the association features (i.e., the risk correlation degree) and the unstructured features (i.e., the risk similarity degree), the accuracy of the features of the object to be recognized is considered; based on the above-mentioned structured features, association features, and unstructured features, the accuracy of determining the probability that the object to be recognized belongs to the risk object type is improved, that is, the accuracy of recognizing the object to be recognized (such as a merchant) is improved.

[0244] The embodiment of the present application further provides an electronic device, and the structural schematic diagram of the electronic device is as Figure 9 shown Figure 9 The electronic device 4000 shown includes: a processor 4001 and a memory 4003. Among them, the processor 4001 and the memory 4003 are connected, such as being connected through a bus 4002. Optionally, the electronic device 4000 may further include a transceiver 4004, and the transceiver 4004 may be used for data interaction between the electronic device and other electronic devices, such as data sending and / or data receiving, etc. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation to the embodiment of the present application.

[0245] The processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of this application. The processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0246] The bus 4002 may include a path for transmitting information between the above components. The bus 4002 may be a PCI (Peripheral Component Interconnect) bus, an EISA (Extended Industry Standard Architecture) bus, or the like. The bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 9 only a thick line is shown herein, but it does not mean that there is only one bus or one type of bus.

[0247] The memory 4003 may be a ROM (Read Only Memory) or other type of static storage device that can store static information and instructions, a RAM (Random Access Memory) or other type of dynamic storage device that can store information and instructions, or it may also be an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory), or other optical disc storage, optical disc storage (including compact discs, laser discs, optical discs, digital versatile discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, which is not limited herein.

[0248] The memory 4003 is used to store the computer program for implementing the embodiments of the present application, and is controlled by the processor 4001 to execute. The processor 4001 is used to execute the computer program stored in the memory 4003 to implement the steps shown in the foregoing method embodiments.

[0249] Among them, the electronic device includes but is not limited to: servers, etc.

[0250] Applying the embodiments of the present application has at least the following beneficial effects:

[0251] Determine the transaction feature vector of the object to be recognized, where the transaction feature vector of the object to be recognized is used to characterize the structured features related to the transaction of the object to be recognized; based on the transaction feature vector of the object to be recognized, determine the transaction risk degree of the object to be recognized, where the transaction risk degree of the object to be recognized is used to characterize the initial probability that the object to be recognized belongs to the risk object type; determine the risk association degree of the object to be recognized and the risk similarity degree of the object to be recognized, where the risk association degree of the object to be recognized is used to characterize the association relationship between the object to be recognized and various sub-objects in the preset seed object set, and the risk similarity degree of the object to be recognized is used to characterize the similarity between the object to be recognized and various sub-objects, and various sub-objects belong to the risk object type; based on the transaction risk degree of the object to be recognized, the risk association degree of the object to be recognized, and the risk similarity degree of the object to be recognized, determine the probability that the object to be recognized belongs to the risk object type; thus, through the structured features (i.e., the transaction risk degree), the comprehensiveness of the features of the object to be recognized is considered; through the association features (i.e., the risk association degree) and the unstructured features (i.e., the risk similarity degree), the accuracy of the features of the object to be recognized is considered; based on the above-mentioned structured features, association features, and unstructured features, the accuracy of determining the probability that the object to be recognized belongs to the risk object type is improved, that is, the accuracy of recognizing the object to be recognized (such as a merchant) is improved.

[0252] The embodiments of the present application provide a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps and corresponding contents of the foregoing method embodiments can be implemented.

[0253] The embodiments of the present application further provide a computer program product, including a computer program, and when the computer program is executed by a processor, the steps and corresponding contents of the foregoing method embodiments can be implemented.

[0254] Based on the same principle as the method provided in the embodiments of the present application, the embodiments of the present application also provide a computer program product or a computer program. The computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. The processor of the computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the method provided in any optional embodiment of the present application above.

[0255] It should be understood that although the flowchart of the embodiments of the present application indicates each operation step by an arrow, the execution order of these steps is not limited to the order indicated by the arrow. Unless there is a clear description in this article, in some implementation scenarios of the embodiments of the present application, the implementation steps in each flowchart can be executed in other orders according to requirements. In addition, some or all of the steps in each flowchart may include multiple sub-steps or multiple stages based on the actual implementation scenario. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage of these sub-steps or stages can also be executed at different times respectively. In the scenario where the execution times are different, the execution order of these sub-steps or stages can be flexibly configured according to requirements, and the embodiments of the present application do not limit this.

[0256] The above are only optional implementation manners of some implementation scenarios of the present application. It should be noted that for those of ordinary skill in the art in the technical field of the present application, without departing from the technical concept of the solution of the present application, adopting other similar implementation means based on the technical idea of the present application also belongs to the protection scope of the embodiments of the present application.

Claims

1. A recognition method, characterized in that, Including: Determine the transaction feature vector of the object to be identified, where the transaction feature vector of the object to be identified is used to characterize the structured features related to the transaction of the object to be identified; Based on the transaction feature vector of the object to be identified, determine the transaction risk degree of the object to be identified, where the transaction risk degree of the object to be identified is used to characterize the initial probability that the object to be identified belongs to the risk object type; Determine the risk association degree of the object to be identified and the risk similarity degree of the object to be identified. The risk association degree of the object to be identified is used to characterize the association relationship between the object to be identified and various sub-objects in the preset seed object set, and the risk similarity degree of the object to be identified is used to characterize the similarity between the object to be identified and the various sub-objects. The various sub-objects belong to the risk object type; Based on the transaction risk degree of the object to be identified, the risk association degree of the object to be identified, and the risk similarity degree of the object to be identified, determine the probability that the object to be identified belongs to the risk object type.

2. The method according to claim 1, characterized in that, The determining the transaction feature vector of the object to be identified includes: Based on a preset transaction feature set, determine the transaction feature value of the object to be identified for each transaction feature in the transaction feature set. The feature type of each transaction feature belongs to any one of object settlement-in features, object transaction process features, and object transaction detection features; Based on the transaction feature values of the object to be identified, determine the transaction feature vector of the object to be identified.

3. The method according to claim 1, characterized in that, The determining the transaction risk degree of the object to be identified based on the transaction feature vector of the object to be identified includes: Input the transaction feature vector of the object to be identified into the first classification model in the target model, and through classification prediction processing, determine the transaction risk degree of the object to be identified.

4. The method according to claim 1, characterized in that, The determining the risk association degree of the object to be identified includes: For each risk association feature in the preset risk association feature set, determine the association relationship between the object to be identified and various sub-objects in the preset seed object set; Based on the association relationships corresponding to the risk association features, determine the risk association degree of the object to be identified.

5. The method according to claim 1, characterized in that, The determining the risk similarity degree of the object to be identified includes: For each unstructured feature in the preset unstructured feature set, determine the initial similarity between the object to be identified and the various sub-objects; Based on the initial similarities, through screening processing, determine the initial risk similarity degree of the object to be identified for each unstructured feature; Based on the initial risk similarity degrees, determine the risk similarity degree of the object to be identified.

6. The method according to claim 5, characterized in that, The determining the initial similarity between the object to be identified and the various sub-objects for each unstructured feature in the preset unstructured feature set includes: For each unstructured feature in the unstructured feature set, obtain multiple unstructured feature information of the object to be identified and multiple unstructured feature information of the various sub-objects; Based on the multiple unstructured feature information of the object to be identified, through screening and splicing processing, determine the first feature embedding vector of the object to be identified; Based on the multiple unstructured feature information of the various sub-objects, through screening and splicing processing, determine the second feature embedding vector of the various sub-objects; Based on the first feature embedding vector of the object to be recognized and the second feature embedding vectors of the various sub-objects, determine the initial similarity between the object to be recognized and the various sub-objects.

7. The method according to claim 1, characterized in that, The determining the probability that the object to be recognized belongs to the risk object type based on the transaction risk degree of the object to be recognized, the risk association degree of the object to be recognized, and the risk similarity of the object to be recognized includes: Input the transaction risk degree of the object to be recognized, the risk association degree of the object to be recognized, and the risk similarity of the object to be recognized into the second classification model in the target model, and through classification prediction processing, determine the probability that the object to be recognized belongs to the risk object type.

8. The method according to claim 2, characterized in that, Before determining the transaction feature vector of the object to be recognized, it further includes: Obtain the initial transaction feature set for multiple object samples; For the transaction feature value interval corresponding to any transaction feature in the initial transaction feature set, perform grouping processing to determine multiple grouping intervals; Determine the weight of evidence corresponding to each grouping interval in the multiple grouping intervals; Based on each weight of evidence, perform weighted summation processing to determine the amount of information corresponding to the any transaction feature; Based on the amounts of information corresponding to the transaction features in the initial transaction feature set, perform sorting processing, and construct the transaction features corresponding to the multiple amounts of information ranked ahead into the transaction feature set.

9. The method according to claim 1, characterized in that, Before determining the transaction feature vector of the object to be recognized, it further includes: Determine the transaction feature vectors of multiple object samples; Input the transaction feature vectors of multiple object samples into the first classification model in the target model to be trained, and through classification prediction processing, determine the transaction risk degree of each object sample in the multiple object samples; Determine the risk association degree of each object sample and the risk similarity of each object sample; Input the transaction risk degree of each object sample, the risk association degree of each object sample, and the risk similarity of each object sample into the second classification model in the target model to be trained, and through classification prediction processing, determine the probability that each object sample belongs to the risk object type; Based on the probabilities that each object sample belongs to the risk object type, determine the value of the loss function of the target model to be trained; If the value of the loss function of the target model to be trained is greater than the loss threshold, update the network parameters of the target model to be trained; Repeat the process of inputting the transaction feature vectors of multiple object samples into the first classification model of the target model to be trained, and through classification prediction processing, determine the transaction risk level of each object sample among the multiple object samples, input the transaction risk level of each object sample, the risk correlation degree of each object sample, and the risk similarity degree of each object sample into the second classification model of the target model to be trained, and through classification prediction processing, determine the probability that each object sample belongs to the risk object type, determine the value of the loss function of the target model to be trained based on the probabilities that each object sample belongs to the risk object type, and if the value of the loss function of the target model to be trained is greater than the loss threshold, update the network parameters of the target model to be trained until the value of the loss function of the target model to be trained is equal to the loss threshold to obtain the target model.

10. An identification device, characterized in that, Including: A first processing module for determining the transaction feature vector of an object to be identified, where the transaction feature vector of the object to be identified is used to represent the structured features related to the transaction of the object to be identified; A second processing module for determining the transaction risk level of the object to be identified based on the transaction feature vector of the object to be identified, where the transaction risk level of the object to be identified is used to represent the initial probability that the object to be identified belongs to the risk object type; A third processing module for determining the risk correlation degree of the object to be identified and the risk similarity degree of the object to be identified, where the risk correlation degree of the object to be identified is used to represent the association relationship between the object to be identified and various sub-objects in a preset seed object set, and the risk similarity degree of the object to be identified is used to represent the similarity between the object to be identified and the various sub-objects, and the various sub-objects belong to the risk object type; A fourth processing module for determining the probability that the object to be identified belongs to the risk object type based on the transaction risk level of the object to be identified, the risk correlation degree of the object to be identified, and the risk similarity degree of the object to be identified.

11. An electronic device, including a memory, a processor, and a computer program stored on the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-9.

12. A computer-readable storage medium, on which a computer program is stored, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1-9.

13. A computer program product, including a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the method according to any one of claims 1-9.

Images