Entity evaluation methods, apparatus, computer-readable media, and electronic devices
By combining the characteristics of the entity itself and related entities to construct a target entity evaluation model, the problem of low evaluation accuracy in existing technologies is solved, and higher evaluation accuracy and effectiveness are achieved.
Patent Information
- Application Number
- CN202210582121.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-05-26
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2042-05-26
AI Technical Summary
Existing entity evaluation methods rely on the characteristics of the entity itself to build models, resulting in low evaluation accuracy and an inability to effectively distinguish different types of entities.
By constructing the original features of the entity and the target features of the associated entities, additional feature values are generated. Combined with the original features, an evaluation model for the target entity is established, which increases the dimensionality and accuracy of the evaluation.
It improves the accuracy and effectiveness of entity evaluation, and can better reflect the characteristics and changing trends of entities.
Smart Images

Figure CN117194932B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of machine learning technology, and more specifically, to a method, apparatus, computer-readable medium, and electronic device for evaluating entities. Background Technology
[0002] Existing entity evaluation methods utilize the entity's own characteristics to construct an entity evaluation model, and then use the entity evaluation model to evaluate the entity. This method has limited evaluation effectiveness and low evaluation accuracy. Summary of the Invention
[0003] The embodiments of this application provide a method, apparatus, computer-readable medium, and electronic device for evaluating entities, thereby improving the effectiveness and accuracy of entity evaluation to at least a certain extent.
[0004] Other features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.
[0005] According to one aspect of the embodiments of this application, an entity evaluation method is provided, the method comprising: constructing an original entity sample set, the original entity sample set including original entity samples corresponding to multiple entities respectively, the original entity samples including entity labels and feature values corresponding to multiple original features respectively; determining at least one associated entity that has an association relationship with the entity, and obtaining target feature values corresponding to multiple target features of the associated entity respectively; for each entity, determining additional feature values corresponding to additional features of the entity based on the target feature values of each associated entity, wherein the additional features are generated based on the target features; adding the additional feature values corresponding to the additional features of the entity to the original entity sample set to obtain a target entity sample set; and establishing a target entity evaluation model based on the target entity sample set to perform entity evaluation based on the target entity evaluation model.
[0006] According to one aspect of the embodiments of this application, an entity evaluation apparatus is provided, the apparatus comprising: a construction unit, configured to construct an original entity sample set, the original entity sample set including original entity samples corresponding to multiple entities respectively, the original entity samples including entity labels and feature values corresponding to multiple original features respectively; a determination and acquisition unit, configured to determine at least one associated entity that has an association relationship with the entity, and acquire target feature values corresponding to multiple target features of the associated entity respectively; an additional feature value determination unit, configured to, for each entity, determine an additional feature value corresponding to an additional feature of the entity based on the target feature values of each associated entity, wherein the additional feature is generated based on the target feature; an addition unit, configured to add the additional feature value corresponding to the additional feature of the entity to the original entity sample set to obtain a target entity sample set; and a modeling unit, configured to establish a target entity evaluation model based on the target entity sample set, so as to perform entity evaluation based on the target entity evaluation model.
[0007] In some embodiments of this application, based on the foregoing scheme, the determining and obtaining unit is configured to: for each of a plurality of preset dimensions, determine the associated entities that are associated with the entity in the preset dimension, so as to obtain a plurality of associated entities that are associated with the entity.
[0008] In some embodiments of this application, based on the foregoing scheme, the plurality of original features are the plurality of target features.
[0009] In some embodiments of this application, based on the foregoing scheme, the additional features are multiple, and the adding unit is configured to: determine the information value of each additional feature according to the additional feature value corresponding to the additional feature of each entity; select target additional features from each additional feature according to the information value of each additional feature; add the additional feature value corresponding to the target additional feature of the entity to the original entity sample set to obtain the target entity sample set.
[0010] In some embodiments of this application, based on the aforementioned scheme, the adding unit is configured to: select additional features from each additional feature whose corresponding information value is greater than a predetermined information value threshold, and use them as target additional features.
[0011] In some embodiments of this application, based on the foregoing scheme, before determining the additional feature value corresponding to the additional feature of the entity according to the target feature value of each associated entity that is associated with the entity, the additional feature value determining unit is further configured to: select a plurality of derivative means sets from the derivative means included in a plurality of derivative dimensions, wherein each derivative dimension includes at least one derivative means, the derivative means set includes at least one derivative means, and the derivative means in the derivative means set belong to different derivative dimensions respectively; and derive additional features corresponding to the target feature of the entity according to each derivative means set respectively.
[0012] In some embodiments of this application, based on the foregoing scheme, the modeling unit is configured to: establish multiple candidate entity evaluation models according to the target entity sample set; evaluate the multiple candidate entity evaluation models according to predetermined model evaluation indicators to obtain the target entity evaluation model with the optimal predetermined model evaluation indicators.
[0013] In some embodiments of this application, based on the foregoing scheme, the modeling unit is configured to: evaluate the plurality of candidate entity evaluation models and obtain the KS value of each candidate entity evaluation model; and determine the candidate entity evaluation model with the largest KS value as the target entity evaluation model.
[0014] In some embodiments of this application, based on the foregoing scheme, the construction unit is configured to: select entities that are abnormal and normal within a predetermined time period before the current time to obtain multiple entities; and construct original entity samples corresponding to the multiple entities to obtain an original entity sample set.
[0015] According to one aspect of the embodiments of this application, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the entity evaluation method as described in the above embodiments.
[0016] According to one aspect of the embodiments of this application, an electronic device is provided, including: one or more processors; and a storage device for storing one or more programs, which, when executed by the one or more processors, cause the one or more processors to implement the entity evaluation method as described in the above embodiments.
[0017] According to one aspect of the embodiments of this application, a computer program product is provided, the computer program product including computer instructions stored in a computer-readable storage medium, a processor of a computer device reading the computer instructions from the computer-readable storage medium, and the processor executing the computer instructions, causing the computer device to perform the entity evaluation method as described in the above embodiments.
[0018] In some embodiments of this application, the technical solutions involve constructing a set of original entity samples, first identifying associated entities with relationships to the entity, obtaining target feature values for these associated entities, then determining additional feature values based on the target feature values of the associated entities, and finally adding these additional feature values to the original entity sample set to obtain the target entity sample set. Therefore, the target entity sample set used in the embodiments of this application not only includes feature values corresponding to the original features of the entity but also additional feature values determined based on the target feature values of the entity's associated entities. This adds additional feature values from other dimensions that can be used for modeling from a unique perspective, enabling the target entity evaluation model built based on the target entity sample set to consider the influence of more relevant factors, thereby improving the accuracy of the entity evaluation model and achieving better results in entity evaluation.
[0019] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description
[0020] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings: Figure 1 A schematic diagram of an exemplary system architecture to which the technical solutions of the embodiments of this application can be applied is shown; Figure 2 A flowchart of an entity evaluation method according to an embodiment of this application is shown; Figure 3 A schematic diagram illustrating the original features according to an embodiment of this application is shown; Figure 4 An embodiment according to this application is shown. Figure 2 A flowchart detailing step 220 in the embodiment; Figure 5 A schematic diagram of multiple preset dimensions according to an embodiment of this application is shown; Figure 6 An embodiment according to this application is shown. Figure 2 A flowchart of the steps preceding step 250 in the embodiment; Figure 7 A schematic diagram of the principle derived from a feature of one embodiment of this application is shown; Figure 8An embodiment according to this application is shown. Figure 2 A flowchart detailing step 260 in the embodiment; Figure 9 A schematic diagram illustrating the principle of feature filtering according to an embodiment of this application is shown; Figure 10 An embodiment according to this application is shown. Figure 2 A flowchart detailing step 270 in the embodiment; Figure 11 A block diagram of an entity evaluation apparatus according to one embodiment of this application is shown; Figure 12 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown. Detailed Implementation
[0021] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this application more comprehensive and complete, and to fully convey the concept of the exemplary embodiments to those skilled in the art.
[0022] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.
[0023] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.
[0024] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily need to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.
[0025] With the development and application of machine learning and artificial intelligence, the classification or evaluation of entities is required in many fields. For example, in the field of risk control, it may be necessary to identify the risk category of an account or the probability of risk; in the distribution of fresh produce, it may be necessary to identify the quality grade of fresh produce or the probability that fresh produce is substandard.
[0026] As described in the background section of this application, related technologies typically construct entity evaluation models based on the inherent characteristics of the entity itself, and then use these models to evaluate the entity. For example, to differentiate the trustworthiness of merchants, a merchant trustworthiness differentiation model is constructed using only the merchant's own characteristics. This model can output the probability that a merchant is trustworthy. However, such a merchant trustworthiness differentiation model has low accuracy and cannot accurately distinguish between trustworthy and untrustworthy merchants.
[0027] Therefore, the accuracy of entity evaluation using existing methods is not high, and good evaluation results cannot be achieved.
[0028] Therefore, this application first provides an entity evaluation method. The entity evaluation method provided by the embodiments of this application can overcome the above-mentioned deficiencies. Based on the original features of the entity, it further determines the entity's additional features, and simultaneously constructs an entity evaluation model based on both the original and additional features. Furthermore, unlike the entity's original features, the entity's additional features are determined using the target features of related entities that have a relationship with the entity. Therefore, compared to the original features, the additional features can carry additional information. Consequently, the constructed entity evaluation model has higher accuracy, further improving the effect of entity evaluation.
[0029] Figure 1 A schematic diagram of an exemplary system architecture to which the technical solutions of the embodiments of this application can be applied is shown. For example... Figure 1 As shown, the system architecture 100 may include a first user terminal 111, a second user terminal 112, a third user terminal 113, a cloud platform 120, and an approval terminal 130. Each user terminal can communicate with the cloud platform 120 via a network, and the approval terminal 130 can also access the cloud platform 120 via a network. Each user terminal has a client-side software system deployed on it, and the cloud platform 120 has a server-side software system deployed on it. The cloud platform 120 can be the execution entity of the scheme in this application embodiment. Based on Figure 1The entity evaluation process of the system architecture shown in the embodiment is as follows: First, users on each user terminal access the cloud 120 using a client to register in the software system of the cloud 120 and obtain a user account; then, each user uses the services and functions provided by the software system based on their user account, and can perform operations such as fund transfers between user accounts, thereby generating corresponding data records in the cloud; next, the cloud 120 summarizes the data records of each user account and sends the summary results to the approval terminal 130. On the approval terminal 130, the approver marks each user account with corresponding tags based on the data records of each user account and sends the tags of each user account to the cloud 120; next, taking the user account as an entity, the cloud 120 generates feature values corresponding to multiple original features based on the data records of each user account, and uses the feature values and corresponding tags of the user account as a raw sample, thereby obtaining the corresponding features of multiple user accounts. The cloud platform 120 first generates multiple original samples; then, it identifies associated user accounts related to each user account and generates corresponding additional features based on the original features of the associated user accounts, and calculates the additional feature values corresponding to the additional features; next, it adds the additional feature values to the original samples corresponding to the user accounts, thereby obtaining target samples corresponding to the user accounts, and constructing a target sample set; next, it uses the target sample set to train a user account risk assessment model; finally, when it is necessary to evaluate a specific user account, the cloud platform 120 can use the constructed user account risk assessment model to perform the evaluation and obtain the corresponding evaluation results. The cloud platform 120 can also push the data records of the specified user account and the corresponding evaluation results to the approval terminal 130, and the approval terminal 130 will then conduct a second review. The cloud platform 120 can use the review results obtained from the approval terminal 130 as the final evaluation result.
[0030] In some embodiments of this application, the user account risk assessment model is constructed based on a logistic regression model or an XGBoost (eXtreme Gradient Boosting) model.
[0031] In some embodiments of this application, the software system is a payment tool or a shopping platform.
[0032] It should be understood that Figure 1 The number of approval terminals, cloud terminals, and user terminals shown is merely illustrative. Depending on implementation needs, any number of approval terminals, cloud terminals, and user terminals can be included; that is, there can be multiple approval terminals, and the number of user terminals can be less than or more than three. Furthermore, the cloud terminal 120 can be a server cluster composed of multiple servers, etc.
[0033] It should be noted that, Figure 1 The illustration shown is merely one embodiment of this application. Although in Figure 1 In the embodiment, the approval terminal is a laptop computer and the user terminal is a smartphone. However, in other embodiments of this application, the approval terminal and user terminal can also be various types of terminal devices such as desktop computers, workstations, tablets, and vehicle terminals, and the types of user terminals can be different. Figure 1 In this embodiment, the entity evaluation method is applied to the scenario of risk assessment of user accounts; however, in other embodiments of this application, it can also be applied to the evaluation scenarios of various other types of entities such as merchants and fruits; although in Figure 1 In the embodiments described above, labeling is performed on the approval terminal. However, in other embodiments of this application, labeling can be performed on various terminal devices other than the approval terminal. This application does not limit the scope of protection of this application in any way.
[0034] As is readily understood, the entity evaluation method provided in this application embodiment is generally executed by a server, and correspondingly, the entity evaluation device is generally located in the server. However, in other embodiments of this application, the terminal device may also have similar functions to the server, thereby executing the entity evaluation scheme provided in this application embodiment.
[0035] Therefore, the embodiments of this application can be applied to terminals or servers. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, etc., but is not limited to these. The terminal and server can be directly or indirectly connected via wired or wireless communication, and this application does not impose any restrictions.
[0036] The embodiments of this application can be applied to blockchain. For example, entity data can be obtained from the blockchain for entity evaluation, entity evaluation results can be stored in the blockchain, and sample data used to build entity evaluation models can even be stored in the blockchain.
[0037] Blockchain is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and cryptographic algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying platform, a platform product and service layer, and an application service layer.
[0038] The underlying blockchain platform can include modules for user management, basic services, smart contracts, and operations. The user management module is responsible for managing the identity information of all blockchain participants, including maintaining public and private key generation (account management), key management, and maintaining the correspondence between user identities and blockchain addresses (access management). Under authorization, it also monitors and audits transactions of certain real identities and provides risk control rule configuration (risk control audit). The basic services module is deployed on all blockchain node devices to verify the validity of business requests. After consensus is reached on valid requests, they are recorded in storage. For a new business request, the basic services first perform interface adaptation parsing and authentication (interface adaptation), and then encrypt the business information using a consensus algorithm (consensus). The management module (network communication) transmits encrypted data to the shared ledger in a complete and consistent manner, and records and stores the data. The smart contract module is responsible for contract registration, issuance, triggering, and execution. Developers can define contract logic using a programming language and publish it to the blockchain (contract registration). According to the contract terms, the module calls keys or other events to trigger execution and complete the contract logic. It also provides functions for contract upgrades and cancellations. The operations module is mainly responsible for deployment, configuration modification, contract settings, cloud adaptation, and real-time visualization of the product's running status during the product release process, such as alarms, network status updates, and node device health status updates.
[0039] The platform's product service layer provides the basic capabilities and implementation frameworks for typical applications. Developers can leverage these basic capabilities, along with the specific characteristics of their business needs, to implement blockchain-based business logic. The application service layer provides blockchain-based application services to business stakeholders.
[0040] The implementation details of the technical solutions in the embodiments of this application are described in detail below: Figure 2A flowchart illustrating an entity evaluation method according to an embodiment of this application is shown. This entity evaluation method can be executed by various computing and processing devices, such as user terminals or cloud servers. User terminals include, but are not limited to, mobile phones, computers, smart voice interaction devices, smart home appliances, vehicle terminals, and aircraft. Embodiments of this application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, and assisted driving. Please refer to... Figure 2 As shown, the evaluation method for this entity includes at least the following steps: In step 210, an original entity sample set is constructed. The original entity sample set includes original entity samples corresponding to multiple entities. The original entity samples include entity labels and feature values corresponding to multiple original features.
[0041] First, it is necessary to determine the entities used to construct the original entity sample set. Then, a corresponding original entity sample is generated for each entity, thereby obtaining an original entity sample set that includes multiple original entity samples.
[0042] An entity can be any object with certain characteristics or attributes. For example, it can be a physical entity that humans can directly perceive, such as animals, plants, fruits, or household appliances. It can also be a virtual entity such as accounts, merchants, or users in websites, online platforms, or software systems. Original characteristics, also known as variables or attributes, reflect the entity's properties in a certain aspect. For example, the original characteristics of fruit could be its color, smell, weight, and shape. A banana can be an entity; its color can be considered an original characteristic, and whether the banana is dark yellow, light yellow, or bluish-green is the feature value corresponding to this original characteristic.
[0043] Entity labels are variable values assigned to entities that correspond to the target variable. Entity labels are pre-defined and reflect the entity's category. For example, an entity label can be 0 or 1, and 0 and 1 can have different meanings depending on the entity and the target variable; when the entity is a watermelon and the target variable is the quality of the watermelon, 1 can represent a bad watermelon, and 0 can represent a good watermelon.
[0044] In one embodiment of this application, the entity is a merchant.
[0045] Merchants can be merchants on e-commerce platforms or payment platforms, such as merchants using payment tools to receive and make payments. When the entity is a merchant, the target variable can be whether the merchant is abnormal. Accordingly, when the entity label is 0, it can represent that the merchant is not abnormal; when the entity label is 1, it can represent that the merchant is abnormal. Accordingly, the entity evaluation method is used to evaluate whether the merchant is abnormal.
[0046] A merchant being flagged as abnormal can mean that the online platform has imposed certain restrictions on the merchant. These restrictions can take many forms, including adding the merchant to a blacklist, freezing payments, or disabling payment functionality. This information about the restrictions is stored in the platform's database. By retrieving this database, the platform can obtain the details of these restrictions and generate corresponding tags.
[0047] In one embodiment of this application, constructing an original entity sample set includes: selecting entities that are abnormal and normal within a predetermined time period before the current time to obtain multiple entities; and constructing original entity samples corresponding to the multiple entities to obtain an original entity sample set.
[0048] Specifically, taking merchants as an example, an abnormal entity is a merchant that has an abnormality within a predetermined time period before the current time; a normal entity is a merchant that does not have an abnormality within a predetermined time period before the current time; the original entity sample set includes good samples and bad samples, with bad samples being samples corresponding to abnormal entities and good samples being samples corresponding to normal entities.
[0049] Therefore, an original entity sample can be classified as a good sample or a bad sample. Since merchants' risk behavior is often dynamic, this embodiment of the application uses merchants with recent anomalies as corresponding bad samples, which allows the final target entity evaluation model to accurately reflect the latest trends in merchants' risk behavior, thereby ensuring the model's recognition accuracy.
[0050] Of course, in other embodiments of this application, normal entities can also be set as merchants that never have any anomalies to improve accuracy.
[0051] In one embodiment of this application, the good sample refers to the merchant that has no anomalies in the predetermined time period before the current time period and whose trusted variables are preset values.
[0052] Credible variables are those that can reflect to a large extent whether a merchant is likely to be abnormal.
[0053] Specifically, the reliable variable can be a credit rating, and the reliable variable is a preset value, i.e., the credit rating is one of the following: S+, A. A+, A-, B+.
[0054] Credit ratings can be obtained by analyzing and statistically analyzing the merchant's behavior on the online platform; the rules for analysis and statistics can be set as needed.
[0055] Therefore, bad samples can be merchants with abnormal activity within the previous 6 months, while good samples can be merchants with credit ratings of S+ or A. Merchants that are one of A+, A-, or B+ and have no abnormalities in the 6 months prior to the current moment.
[0056] The number of good and bad samples in the original entity sample set can be arbitrary. The number of good and bad samples can be the same or different. The original entity sample set should include multiple good samples and multiple bad samples.
[0057] Specifically, the number of good and bad samples in the original entity sample set can be shown in Table 1:
[0058] Table 1 As shown in Table 1, the original entity sample set includes 10,233 good samples and 4,984 bad samples. Therefore, the original entity sample set includes 15,217 original entity samples, of which the bad sample rate is 32.75%. The number of good samples in the original entity sample set is greater than the number of bad samples.
[0059] Multiple raw feature dimensions can be set, and one or more raw features can be selected under each raw feature dimension.
[0060] In one embodiment of this application, the plurality of original features include at least one of the following: original features related to transactions, original features related to credit ratings, and original features related to risk operation information.
[0061] A primitive feature is a primitive feature dimension, which is equivalent to a class of primitive features.
[0062] Figure 3 A schematic diagram illustrating the original features according to an embodiment of this application is shown. Please refer to... Figure 3 As shown, multiple variable dimensions are set, which are also the original feature dimensions. Each variable under a variable dimension can be used as an original feature. By summing up the variables under each variable dimension, multiple original features can be obtained.
[0063] Specifically, Figure 3 The variables shown represent transaction variables, credit ratings, and risk labels and risk operations. Transaction variables reflect transaction activity over the past 180 days and include four variables: amount, number of transactions, number of accounts, and credit card amount. Amount represents the total transaction amount, number of transactions represents the number of transactions made by the merchant, number of accounts represents the number of accounts that made payments to the merchant, and credit card amount represents the amount paid to the merchant via credit card. Credit rating represents the payment instrument's assessment of the merchant's creditworthiness. This variable includes S+, A, and other ratings. The four variables are A-, B+, and A-, each taking a value of 0 or 1. Of course, the credit rating variable dimension can also include other variables such as A+. In risk tags and risk operations, the risk tags are called risk tags (frisk_tags), and the risk operations are called risk types (frisk_tag_type). Risk tags and risk operations consist of four pairs of risk tag (frisk_tags) and risk type (frisk_tag_type). These pairs can be pre-marked for merchants on the payment platform, and each pair can be considered a variable.
[0064] You can set the variable dimensions and the variables under each dimension as needed. By summing up the variables under each dimension, you can obtain multiple original features, and then calculate the feature values corresponding to each of the original features.
[0065] In step 220, at least one associated entity that has a relationship with the entity is determined, and the target feature values corresponding to multiple target features of the associated entity are obtained respectively.
[0066] For each entity, one or more associated entities can be identified. Associated entities are other entities of the same type as the entity. The relationships between an entity and its associated entities can be varied.
[0067] The inventors discovered that the concept of "birds of a feather flock together" can be applied to the feature engineering stage of machine learning. Therefore, the inventors believe that the features of related entities can also reflect the characteristics of the entities themselves, and thus creatively introduced related entities. By using the features of related entities to assist in the evaluation of corresponding entities, better evaluation results can be achieved.
[0068] The associated entities can be other entities in the original entity sample set that correspond to other original entity samples, or they can be entities that do not belong to the original entity samples.
[0069] Figure 4 An embodiment according to this application is shown. Figure 2 A flowchart detailing step 220 in the embodiment. (See attached flowchart.) Figure 4 As shown, determining at least one associated entity that has a relationship with an entity may specifically include the following steps: In step 220', for each of the multiple preset dimensions, the associated entities that have a relationship with the entity in the preset dimension are determined, so as to obtain multiple associated entities that have a relationship with the entity.
[0070] Each preset dimension represents a relationship between an entity and its associated entities. Multiple preset dimensions represent various different relationships. For each entity, one or more associated entities can be found through each preset dimension. The associated entities found for this entity through each preset dimension may be the same or different. Ultimately, the multiple associated entities identified for the entity are distinct entities, and the relationships between each entity and the entity can be the same or different types of relationships. The number of preset dimensions can be designed as needed.
[0071] In one embodiment of this application, the preset dimensions can be set according to the user's actual needs, and are not limited here. If the entity refers to a merchant, the preset dimensions can refer to the merchant's legal representative, bank card number, merchant name, business entity, shareholders, etc., and multiple preset dimensions mean at least two preset dimensions.
[0072] Any two or more of the above preset dimensions can be used as multiple preset dimensions.
[0073] Figure 5 A schematic diagram illustrating multiple preset dimensions according to an embodiment of this application is shown. Please refer to... Figure 5 As shown, taking an entity as a merchant and related entities as related merchants as an example, a merchant can be matched with corresponding related merchants through preset dimensions designed based on business background information, such as legal representative, bank card, merchant name, business entity, and shareholders. Specifically, using the preset dimension of legal representative, merchants with the same legal representative as the merchant can be found as related merchants; using the preset dimension of bank card, merchants with the same bank card number as the merchant can be found as related merchants; using the preset dimension of merchant name, merchants with the same name as the merchant can be found as related merchants; using the preset dimension of business entity, merchants with the same business registration number as the merchant can be found as related merchants; and using the preset dimension of shareholders, merchants with one or more shareholders as the merchant can be found as related merchants.
[0074] When the entity is a fruit, other fruits related to this fruit can be determined based on the preset dimension of other fruits placed near this fruit in the fruit stand; other fruits related to this fruit can also be determined based on the preset dimension of other fruits liked by customers who like this fruit.
[0075] Similar to entities, associated entities also have corresponding features or attributes. One or more target features can be the same as one or more original features.
[0076] In one embodiment of this application, the multiple original features are multiple target features.
[0077] In this embodiment of the application, multiple target features are the same as multiple original features.
[0078] Specifically, taking the merchant as an example, if the merchant's original characteristics are variables under three dimensions: transaction variables, credit rating, and risk labels and risk operations, then the target characteristics of the associated merchant can also be variables under these three dimensions. Specifically, the target characteristics of the associated merchant can include: under the transaction variable dimension, amounts, number of transactions, number of accounts, and credit card amounts; and under the credit rating dimension, S+, A, etc. The four variables are A-, B+, and risk labels and risk operations, including [CX], [DB], [CX], [DJFX], and [DPJY].
[0079] Please continue reading Figure 2 As shown, in step 250, for each entity, based on the target feature values of each associated entity that has a relationship with the entity, the additional feature value corresponding to the additional feature of the entity is determined, wherein the additional feature is generated based on the target feature.
[0080] There can be one or more additional features. Specifically, corresponding additional features can be generated based on each target feature, thereby obtaining multiple additional features corresponding to multiple target features respectively; after obtaining the additional features, the corresponding additional feature values can be calculated based on the target feature values.
[0081] Figure 6 An embodiment according to this application is shown. Figure 2 A flowchart of the steps preceding step 250 in this embodiment. Please refer to [link / reference]. Figure 6 As shown, the steps prior to step 250 may specifically include: In step 230, multiple sets of derivative means are selected from the derivative means included in multiple derivative dimensions, wherein each derivative dimension includes at least one derivative means, the set of derivative means includes at least one derivative means, and the derivative means in the set of derivative means belong to different derivative dimensions.
[0082] The fact that the derivative methods in the derivative method set belong to different derivative dimensions means that when the derivative method set includes multiple derivative dimensions, each derivative method is selected from a different derivative dimension, and only one derivative method is selected from each derivative dimension. The selection of the derivative method from each derivative dimension can be determined according to predetermined rules; alternatively, derivative dimensions can be randomly selected, and derivative methods can be further randomly selected from those randomly selected dimensions; furthermore, expert experience can be used to determine which derivative dimension to choose and which derivative method to select. The selection method for derivative dimensions and derivative methods is not limited here. The derivative methods are selected separately from those under different derivative dimensions, and the combination of each selected derivative method constitutes the derivative method set. It is possible to select only one derivative method from the derivative methods included in a single derivative dimension to form the derivative method set, or to select corresponding derivative methods from at least two derivative dimensions to form the derivative method set.
[0083] In step 240, additional features corresponding to the target features of the entity are derived according to each set of derivative means.
[0084] By using each set of derivation methods, additional features corresponding to a target feature of the entity can be derived; by using multiple sets of derivation methods, multiple additional features corresponding to a target feature can be derived respectively. For details on how to derive additional features based on derivation methods, please refer to the following section. Figure 7 The explanatory content.
[0085] Figure 7 A schematic diagram illustrating a feature derived from an embodiment of this application is shown. Please refer to... Figure 7As shown, two derivation dimensions are set: variable processing and derivation method. The variable processing dimension includes two derivation tools: absolute value and same-type proportion. The same-type proportion is the percentage of related merchants of the same type among all related merchants. The derivation method dimension includes three derivation tools: mean, maximum, and minimum. Each related merchant has multiple variables under three variable dimensions: transaction variables, credit rating, and risk label and risk operation, and also has variable values corresponding to each variable. By selecting one derivation tool under each of the variable processing and derivation method dimensions, forming a set of derivation tools, and using the derivation tools in this set to derive from each variable, the corresponding variable can be derived. For example, if we choose absolute value as the derivation method under the variable treatment dimension and mean as the derivation method under the derivation mode dimension, then for a specific transaction variable under the transaction variable dimension, based on the set of derivation methods composed of these two methods, we can derive the additional feature of the average transaction value of related merchants over the past 180 days. If we choose the same type percentage as the derivation method under the variable treatment dimension and maximum value as the derivation mode dimension, then for the variable under the credit rating dimension, we can derive the additional feature of the trustworthy percentage of related merchants, which represents the proportion of related merchants with a predetermined trustworthy credit rating.
[0086] Therefore, the number of transaction variables under the variable dimension can be multiplied, significantly increasing the number of additional features. Simultaneously, for the same variable value across various related merchants, the variables derived from that variable have only one value, thus achieving variable aggregation.
[0087] Please continue reading Figure 2 As shown, in step 260, the additional feature values corresponding to the additional features of the entity are added to the original entity sample set to obtain the target entity sample set.
[0088] For each entity corresponding to an original entity sample in the original entity sample set, the additional feature value corresponding to one or more additional features of the entity can be added to the original entity sample corresponding to the entity to obtain the target entity sample; by adding the corresponding additional feature value of the entity to each original entity sample, a target entity sample set including multiple target entity samples can be obtained.
[0089] Figure 8 An embodiment according to this application is shown. Figure 2 A flowchart detailing step 260 in the embodiment is provided. Please refer to [link / reference]. Figure 8 As shown, step 260 may specifically include the following steps: In step 261, the information value of each additional feature is determined based on the additional feature value corresponding to the additional features of each entity.
[0090] Information Value (IV) of Additional Features: , in, IV For information value, i This refers to the box's serial number. The boxes are divided based on the additional feature values corresponding to the additional features. This represents the ratio of the number of good samples currently in the container to the total number of good samples. This is the ratio of the number of bad samples in the current bin to the total number of bad samples.
[0091] In step 262, target additional features are selected from the additional features based on the information value of each additional feature.
[0092] In one embodiment of this application, selecting target additional features from among the additional features based on the information value of each additional feature includes: Among the various additional features, those with information value greater than a predetermined information value threshold are selected as target additional features.
[0093] In step 263, the additional feature values corresponding to the target additional features of the entity are added to the original entity sample set to obtain the target entity sample set.
[0094] Of course, in other embodiments of this application, target additional features can also be screened based on other indicators.
[0095] Although the filtering in this application embodiment is based on additional features, in other embodiments of this application, the filtering can also be performed on the whole consisting of additional features and original features.
[0096] In this embodiment, by further filtering the additional features of entities based on information value, the effectiveness of the filtered target additional features is ensured, thereby improving the performance of the established target entity evaluation model.
[0097] Figure 9 A schematic diagram illustrating the principle of feature filtering according to an embodiment of this application is shown. Please refer to... Figure 9As shown, when the entity is a merchant, 866 dimensions of variables can be obtained as additional features, including the average transaction value of associated merchants over the past 180 days and the trustworthiness ratio of associated merchants. The average transaction value of associated merchants over the past 180 days can be either the average transaction amount or the average number of transactions over the past 180 days. The IV value of the additional features is calculated using the formula above. Then, based on the rule that IV > 0.025, 196 dimensions of variables are selected from the 866 dimensions as target additional features. The types of these variables correspond to the types of variables used to construct the original features: 95 dimensions related to associated merchant transactions, 66 dimensions related to associated merchant risk labels, and 8 dimensions related to associated merchant credit ratings.
[0098] In step 270, a target entity evaluation model is established based on the target entity sample set, so as to evaluate the entity based on the target entity evaluation model.
[0099] By training the model using a sample set of target entities, a target entity evaluation model can be obtained. The target entity evaluation model can be a machine learning model based on various algorithms, such as decision tree algorithm, random forest algorithm, logistic regression (LR) algorithm, XGBoost (eXtreme Gradient Boosting) algorithm, etc., or it can be a model based on deep neural network algorithm. The embodiments of this application do not impose any limitations on this.
[0100] For an entity to be evaluated, a target entity evaluation model can be used to evaluate it. The target entity evaluation model can output probability values or evaluation classification results. When the entity is a merchant, the target entity evaluation model can be an abnormal merchant identification model, which can be used to output the probability that the merchant is abnormal or whether the merchant is abnormal.
[0101] The specific steps of step 270 can be as follows: Figure 10 As shown. Figure 10 An embodiment according to this application is shown. Figure 2 A flowchart detailing step 270 in the embodiment is provided. Please refer to [link / reference]. Figure 10 As shown, establishing a target entity evaluation model based on a target entity sample set may specifically include the following steps: In step 271, multiple candidate entity evaluation models are established based on the target entity sample set.
[0102] Multiple candidate entity evaluation models can be trained using the target entity sample set, and the algorithms used for each candidate entity evaluation model can be different.
[0103] Specifically, logistic regression and XGBoost models can be established as two candidate entity evaluation models.
[0104] Logistic regression maps the result of a linear function to a sigmoid function. The sigmoid function takes values between 0 and 1, a property that allows the model's final result to be interpreted probabilistically. The functional expression of the logistic regression model is as follows:
[0105] in, X As a sample, For output, W and b These are the parameters of the model.
[0106] The XGBoost model is an ensemble tree model, and its objective function expression is as follows:
[0107] in, Let be the predictions of the first t-1 ensemble learners for the samples. This represents the current learner's prediction of the sample. Let be the regularization term for the t-th learner.
[0108] The XGBoost model reduces the loss function by continuously adding new trees and fitting pseudo-residuals, making the model's predictions closer to the true values.
[0109] In step 272, multiple candidate entity evaluation models are evaluated based on predetermined model evaluation indicators to obtain the target entity evaluation model with the optimal predetermined model evaluation indicators.
[0110] The predetermined model evaluation metric can be any type of evaluation metric that can be used to evaluate model performance, such as the KS (Kolmogorov-Smirnov) value. Of course, other evaluation metrics can also be used in other embodiments of this application.
[0111] In this embodiment, multiple models are first constructed using the target entity sample set, and then further screened based on predetermined model evaluation metrics to obtain the model with the optimal predetermined model evaluation metrics, thus ensuring the performance of the model.
[0112] In one embodiment of this application, multiple candidate entity evaluation models are evaluated based on predetermined model evaluation indicators to obtain the target entity evaluation model with the best predetermined model evaluation indicators. This includes: evaluating multiple candidate entity evaluation models to obtain the KS value of each candidate entity evaluation model; and determining the candidate entity evaluation model with the largest KS value as the target entity evaluation model.
[0113]
[0114] Table 1 Table 1 shows the KS values of the trained logistic regression model and XGBoost model when the entity is a merchant. It can be seen that the KS value of the XGBoost model is greater than that of the logistic regression model. Therefore, the performance index of the XGBoost model is better, and the XGBoost model can be selected as the target entity evaluation model.
[0115] When the entity is a merchant, the inventors obtained the following conclusions through experiments: the original model trained based on the merchant's own information had a KS of 61%, while after introducing the associated merchant information in the embodiments of this application, the model's KS was 91%. The model improvement was 148.7%, thus the model's discrimination effect was significantly improved.
[0116] To compare with the average level of evaluations conducted without using the model, the inventors also randomly selected a portion of merchants with lower scores (i.e., higher reliability) based on the model scores and submitted them to the auditors for review. The review results are shown in Table 2.
[0117] Table 2 According to the manual review results in Table 2, out of 184 merchants in the 0-0.014 score range output by the model, 88 merchants were selected for review. The review results showed that 2 merchants were suspected of fraudulent cash-out schemes, resulting in a trustworthy merchant ratio of 97.73%. Out of 3256 merchants in the 0.014-0.1 score range output by the model, 99 merchants were selected for review. The review results showed that 1 merchant was suspected of fraudulent cash-out schemes, resulting in a trustworthy merchant ratio of 98.99%. Statistically, the average trustworthy merchant ratio for all merchants is 15%. Therefore, the model's trustworthy merchant ratio in this score range increased by more than 6 times compared to the average, demonstrating a significant improvement in the trustworthy merchant ratio selected by the model.
[0118] In summary, the entity evaluation method provided in this application, by introducing target features of related entities on top of the original features of the entity and determining additional features and corresponding additional feature values, enables the target entity evaluation model built based on the target entity sample set to consider the influence of more relevant factors, significantly improving the performance of the entity evaluation model and the accuracy of entity evaluation. When this entity evaluation method is applied to evaluate merchants, by introducing information such as transaction characteristics, credit ratings, and risk operations of related merchants, the differentiation effect on the credibility of the merchants themselves is significantly improved.
[0119] The following describes an embodiment of the apparatus described in this application, which can be used to execute the entity evaluation method described in the above embodiments of this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the entity evaluation method described above.
[0120] Figure 11 A block diagram of an entity evaluation apparatus according to an embodiment of this application is shown.
[0121] Reference Figure 11 As shown, an evaluation device 1100 for an entity according to this application includes: a construction unit 1110, a determination and acquisition unit 1120, an additional feature value determination unit 1130, an addition unit 1140, and a modeling unit 1150. The system includes a construction unit 1110 for constructing an original entity sample set, which includes original entity samples corresponding to multiple entities, each original entity sample including entity labels and feature values corresponding to multiple original features; a determination and acquisition unit 1120 for determining at least one associated entity that is related to the entity, and acquiring target feature values corresponding to multiple target features of the associated entity; an additional feature value determination unit 1130 for determining, for each entity, additional feature values corresponding to the additional features of the entity based on the target feature values of each associated entity, wherein the additional features are generated based on the target features; an addition unit 1140 for adding the additional feature values corresponding to the additional features of the entity to the original entity sample set to obtain a target entity sample set; and a modeling unit 1150 for establishing a target entity evaluation model based on the target entity sample set to perform entity evaluation based on the target entity evaluation model.
[0122] In some embodiments of this application, based on the foregoing scheme, the determination and acquisition unit 1120 is configured to: for each of a plurality of preset dimensions, determine the associated entity that has an association relationship with the entity in the preset dimension, so as to obtain a plurality of associated entities that have an association relationship with the entity.
[0123] In some embodiments of this application, based on the foregoing scheme, the plurality of original features are the plurality of target features.
[0124] In some embodiments of this application, based on the foregoing scheme, the additional features are multiple, and the adding unit 1140 is configured to: determine the information value of each additional feature according to the additional feature value corresponding to the additional feature of each entity; select target additional features from each additional feature according to the information value of each additional feature; add the additional feature value corresponding to the target additional feature of the entity to the original entity sample set to obtain the target entity sample set.
[0125] In some embodiments of this application, based on the aforementioned scheme, the adding unit 1140 is configured to: select additional features whose corresponding information value is greater than a predetermined information value threshold from each additional feature, and use them as target additional features.
[0126] In some embodiments of this application, based on the foregoing scheme, before determining the additional feature value corresponding to the additional feature of the entity according to the target feature value of each associated entity that is associated with the entity, the additional feature value determining unit 1130 is further configured to: select a plurality of derivative means sets from the derivative means included in a plurality of derivative dimensions, wherein each derivative dimension includes at least one derivative means, the derivative means set includes at least one derivative means, and the derivative means in the derivative means set belong to different derivative dimensions respectively; and derive additional features corresponding to the target feature of the entity according to each derivative means set respectively.
[0127] In some embodiments of this application, based on the aforementioned scheme, the modeling unit 1150 is configured to: establish multiple candidate entity evaluation models according to the target entity sample set; evaluate the multiple candidate entity evaluation models according to predetermined model evaluation indicators to obtain the target entity evaluation model with the optimal predetermined model evaluation indicators.
[0128] In some embodiments of this application, based on the foregoing scheme, the modeling unit 1150 is configured to: evaluate the plurality of candidate entity evaluation models and obtain the KS value of each candidate entity evaluation model; and determine the candidate entity evaluation model with the largest KS value as the target entity evaluation model.
[0129] In some embodiments of this application, based on the aforementioned scheme, the construction unit 1110 is configured to: select entities that are abnormal and normal within a predetermined time period before the current time to obtain multiple entities; and construct original entity samples corresponding to the multiple entities to obtain an original entity sample set.
[0130] Figure 12 A schematic diagram of the structure of a computer system suitable for implementing the electronic device of the present application is shown.
[0131] It should be noted that, Figure 12 The computer system 1200 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.
[0132] like Figure 12As shown, the computer system 1200 includes a Central Processing Unit (CPU) 1201, which can perform various appropriate actions and processes based on programs stored in Read-Only Memory (ROM) 1202 or programs loaded from storage portion 1208 into Random Access Memory (RAM) 1203, such as performing the methods described in the above embodiments. Various programs and data required for system operation are also stored in RAM 1203. The CPU 1201, ROM 1202, and RAM 1203 are interconnected via bus 1204. An input / output (I / O) interface 1205 is also connected to bus 1204.
[0133] The following components are connected to I / O interface 1205: an input section 1206 including a keyboard, mouse, etc.; an output section 1207 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 1208 including a hard disk, etc.; and a communication section 1209 including a network interface card such as a LAN (Local Area Network) card, modem, etc. The communication section 1209 performs communication processing via a network such as the Internet. A drive 1210 is also connected to I / O interface 1205 as needed. Removable media 1211, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 1210 as needed so that computer programs read from them can be installed into storage section 1208 as needed.
[0134] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication section 1209, and / or installed from removable medium 1211. When the computer program is executed by central processing unit (CPU) 1201, it performs various functions defined in the system of this application.
[0135] It should be noted that the computer-readable medium shown in the embodiments of this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such transmitted data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.
[0136] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0137] The units described in the embodiments of this application can be implemented in software or hardware, and the described units can also be located in a processor. The names of these units do not necessarily limit the specific unit itself.
[0138] In one aspect, this application also provides a computer-readable medium, which may be included in the electronic device described in the above embodiments; or it may exist independently and not assembled into the electronic device. The computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to perform the methods described in the above embodiments.
[0139] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to the embodiments of this application, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.
[0140] Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein can be implemented by software or by combining software with necessary hardware. Therefore, the technical solutions according to the embodiments of this application can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, external hard drive, etc.) or on a network, including several instructions to cause a computing device (such as a personal computer, server, touch terminal, or network device, etc.) to execute the methods according to the embodiments of this application.
[0141] It is understood that in the specific implementation of this application, data related to entities is involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.
[0142] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the embodiments disclosed herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.
[0143] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this application is limited only by the appended claims.
Claims
1. A method for evaluating an entity, characterized in that, The method includes: A raw entity sample set is constructed, comprising raw entity samples corresponding to multiple merchants. Each raw entity sample includes an entity label and feature values corresponding to multiple raw features. The merchants are those on the online platform. The entity label is the variable value corresponding to the target variable, where the target variable is whether the merchant is abnormal. The multiple raw features include at least one of the following: raw features related to transactions, raw features related to credit rating, and raw features related to risk operation information. The raw entity sample set includes good samples and bad samples. Bad samples are those corresponding to merchants with abnormalities, and good samples are those corresponding to normal merchants. Merchants corresponding to good samples are those that did not exhibit abnormalities within a predetermined time period prior to the current moment and whose trusted variables are preset values. For each of the multiple preset dimensions, identify the associated merchants that are related to the merchant in the preset dimension, so as to obtain multiple associated merchants that are related to the merchant, and obtain the target feature values corresponding to multiple target features of the associated merchants respectively; the multiple preset dimensions include at least two of the following: legal representative, bank card number, merchant name, business entity, and shareholder; the multiple original features are the multiple target features; Multiple sets of derivative methods are selected from the two derivative dimensions of variable processing and derivative method. The sets of derivative methods are obtained by selecting one derivative method under each of the two derivative dimensions of variable processing and derivative method. The derivative dimension of variable processing includes two derivative methods: absolute value and same type proportion. The same type proportion is the proportion of the number of related merchants of the same type among all related merchants. The derivative dimension of derivative method includes three derivative methods: mean, maximum value and minimum value. The same type includes credit rating as a predetermined credible rating. Based on each set of derivative methods, additional features corresponding to the target features of the merchant are derived; For each merchant, based on the target feature values of each associated merchant, an additional feature value corresponding to the additional feature of the merchant is determined, wherein the additional feature is generated based on the target feature; The information value of each additional feature is determined based on the additional feature value corresponding to the additional features of each merchant. Among the various additional features, the additional features whose information value is greater than the predetermined information value threshold are selected as the target additional features; The additional feature values corresponding to the target additional features of the merchant are added to the original entity sample set to obtain the target entity sample set; A target entity evaluation model is established based on the target entity sample set to evaluate merchants for anomalies. The target entity evaluation model is an abnormal merchant identification model.
2. The entity evaluation method according to claim 1, characterized in that, The step of establishing a target entity evaluation model based on the target entity sample set includes: Multiple candidate entity evaluation models are established based on the target entity sample set; The multiple candidate entity evaluation models are evaluated based on predetermined model evaluation indicators to obtain the target entity evaluation model with the optimal predetermined model evaluation indicators.
3. The entity evaluation method according to claim 2, characterized in that, The step of evaluating the multiple candidate entity evaluation models based on predetermined model evaluation indicators to obtain the target entity evaluation model with the optimal predetermined model evaluation indicators includes: The evaluation models of the multiple candidate entities are evaluated, and the KS value of each candidate entity evaluation model is obtained. The candidate entity evaluation model with the largest KS value is determined as the target entity evaluation model.
4. The method for evaluating an entity according to any one of claims 1-2, characterized in that, The construction of the original entity sample set includes: Select merchants that are abnormal or normal within a predetermined time period before the current time to obtain multiple merchants; Construct original entity samples corresponding to the multiple merchants respectively to obtain the original entity sample set.
5. An evaluation device for an entity, characterized in that, The device includes: A construction unit is used to construct an original entity sample set, which includes original entity samples corresponding to multiple merchants. Each original entity sample includes an entity label and feature values corresponding to multiple original features. The merchants are merchants on the online platform. The entity label is the variable value corresponding to the target variable, where the target variable is whether the merchant is abnormal. The multiple original features include at least one of the following: original features related to transactions, original features related to credit rating, and original features related to risk operation information. The original entity sample set includes good samples and bad samples. The bad samples are samples corresponding to merchants with abnormalities, and the good samples are samples corresponding to normal merchants. The merchants corresponding to the good samples are those that did not have abnormalities in a predetermined time period before the current moment and whose reliable variables are preset values. The determination and acquisition unit is used to determine at least one associated merchant that has a relationship with the merchant, and to acquire target feature values corresponding to multiple target features of the associated merchant respectively; the determination and acquisition unit is configured to: for each of multiple preset dimensions, determine the associated merchants that have a relationship with the merchant in the preset dimension, so as to obtain multiple associated merchants that have a relationship with the merchant; the multiple preset dimensions include at least two of the following: legal representative, bank card number, merchant name, business entity, and shareholder; the multiple original features are the multiple target features; An additional feature value determination unit is used to determine, for each merchant, an additional feature value corresponding to the additional feature of the merchant based on the target feature values of each associated merchant that has an association relationship with the merchant, wherein the additional feature is generated based on the target feature; Before determining the additional feature value corresponding to the additional feature of the merchant based on the target feature values of each associated merchant, the additional feature value determining unit is further configured to: select multiple sets of derivative means from the two derivative dimensions of variable processing and derivative method, wherein the sets of derivative means are obtained by selecting one derivative means under each of the two derivative dimensions of variable processing and derivative method. The derivative dimension of variable processing includes two derivative means: absolute value and same type proportion. The same type proportion is the proportion of the number of associated merchants of the same type among all associated merchants. The derivative dimension of derivative method includes three derivative means: mean, maximum value, and minimum value. The same type includes credit rating as a predetermined trust rating. The additional feature corresponding to the target feature of the merchant is derived from each set of derivative means. An adding unit is configured to add the additional feature value corresponding to the additional feature of the merchant to the original entity sample set to obtain a target entity sample set; the adding unit is configured to: determine the information value of each additional feature based on the additional feature value corresponding to the additional feature of each merchant; select additional features whose corresponding information value is greater than a predetermined information value threshold from among the additional features as target additional features; add the additional feature value corresponding to the target additional feature of the merchant to the original entity sample set to obtain a target entity sample set; The modeling unit is used to establish a target entity evaluation model based on the target entity sample set, so as to conduct anomaly assessment of merchants based on the target entity evaluation model, wherein the target entity evaluation model is an abnormal merchant identification model.
6. A computer-readable medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the evaluation method for the entity as described in any one of claims 1 to 4.
7. An electronic device, characterized in that, include: One or more processors; A storage device for storing one or more programs, which, when executed by one or more processors, cause the one or more processors to implement the evaluation method of the entity as described in any one of claims 1 to 4.
8. A computer program product, characterized in that, The computer program product includes computer instructions stored in a computer-readable storage medium, a processor of a computer device reading the computer instructions from the computer-readable storage medium, and the processor executing the computer instructions to cause the computer device to perform the evaluation method for the entity as described in any one of claims 1 to 4.
Citation Information
Patent Citations
Enterprise investment risk assessment method, device and storage medium
CN107909274A
Enterprise credit risk scoring method and device, equipment and storage medium
CN114202223A
