A data analysis method, device, computer device, and storage medium

Abstract

Embodiments of the present application provide a data analysis method, apparatus, computer device, and storage medium, which are applied to the field of fintech. The method includes: obtaining in-bank data, external data, and public data of an offshore customer in a target bank; updating an equity penetration map by using the in-bank data, a first risk list included in the external data, and a second risk list included in the public data to obtain an updated equity penetration map; when determining that the offshore customer is a first type of customer according to the updated equity penetration map, obtaining a target transaction behavior feature set of the offshore customer and the holding data of the offshore customer; and invoking an offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data, and the registered business information of the offshore customer included in the public data. By adopting the present application, the accuracy of identifying the category of offshore customers can be improved. The present application relates to blockchain technology, and the data can be obtained through a blockchain network.

Description

Technical Field

[0001] This application relates to the field of data analysis technology, and in particular, to a data analysis method, device, computer device, and storage medium. Background Art

[0002] Cross-border offshore customers have natural advantages such as convenient overseas registration, free flow of funds, and loose supervision. At the same time, it also makes the offshore account (Off shore Account, OSA) an important tool relied on by telecom fraudsters for illegal activities such as transferring funds, seriously affecting the supervision order and financial stability of China's financial market.

[0003] Currently, the risk prevention and control means for cross-border offshore customers mainly identify the categories of offshore customers based on the list of abnormal objects provided by banks. However, the inventor has found that with the continuous expansion of the scale of the offshore customer group and banks, the existing methods for identifying the categories of offshore customers are relatively single, not comprehensive enough, and have low accuracy. Summary of the Invention

[0004] Embodiments of this application provide a data analysis method, device, computer device, and storage medium, which can improve the accuracy of identifying the categories of offshore customers.

[0005] On the one hand, embodiments of this application provide a data analysis method, including:

[0006] Obtain the in-bank data, external data, and public data of the offshore customer in the target bank; the external data includes international service provider data, and the international service provider data includes the first risk list and the equity penetration map provided by the international service provider, and the public data includes the second risk list provided by other institutions and the registered business information of the offshore customer;

[0007] Update the equity penetration map using the in-bank data, the first risk list, and the second risk list to obtain an updated equity penetration map;

[0008] When determining that the offshore customer is a first-type customer based on the updated equity penetration map, obtain the target transaction behavior feature set of the offshore customer and the holding data of the offshore customer;

[0009] Call an offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer.

[0010] On the other hand, embodiments of this application provide a data analysis device, including:

[0011] An acquisition module, configured to acquire in-bank data, external data, and public data of an offshore customer in a target bank; the external data includes international service provider data, and the international service provider data includes a first risk list and an equity penetration map provided by an international service provider; the public data includes a second risk list provided by other institutions and the registered business information of the offshore customer;

[0012] An update module, configured to update the equity penetration map by using the in-bank data, the first risk list, and the second risk list to obtain an updated equity penetration map;

[0013] The acquisition module is further configured to, when determining that the offshore customer is a first type of customer according to the updated equity penetration map, acquire a target transaction behavior feature set of the offshore customer and the holding data of the offshore customer;

[0014] A classification module, configured to call an offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer.

[0015] In another aspect, an embodiment of the present application provides a computer device, including a processor and a memory, the processor and the memory are connected to each other, wherein the memory is used to store computer program instructions, and the processor is configured to execute the program instructions to implement the data analysis method described above.

[0016] In another aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer program instructions are stored, and when the computer program instructions are executed by a processor, they are used to execute the data analysis method described above.

[0017] In summary, a computer device can acquire in-bank data, external data, and public data of an offshore customer in a target bank; the computer device updates an equity penetration map by using the in-bank data, the first risk list included in the external data, and the second risk list included in the public data to obtain an updated equity penetration map; when the computer device determines that the offshore customer is a first type of customer according to the updated equity penetration map, the computer device acquires a target transaction behavior feature set of the offshore customer and the holding data of the offshore customer, and calls an offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer. Compared with the prior art method of identifying the category of an offshore customer solely based on an abnormal object list provided by a bank, the present application combines multi-dimensional data to identify the category of an offshore customer, which can improve the accuracy of identifying the category of an offshore customer. Description of the Drawings

[0018] To more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for the description of the embodiments or the prior art. Obviously, the accompanying drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0019] Figure 1 is a schematic flowchart of a data analysis method provided by an embodiment of the present application;

[0020] Figure 2 is a schematic flowchart of a data analysis method provided by another embodiment of the present application;

[0021] Figure 3 is a schematic structural diagram of a data analysis device provided by an embodiment of the present application;

[0022] Figure 4 is a schematic structural diagram of a computer device provided by an embodiment of the present application. Detailed implementation manners

[0023] The following will describe the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application.

[0024] Please refer to Figure 1 , which is a schematic flowchart of a data analysis method provided by an embodiment of the present application. This method can be applied to a computer device. The computer device can be a user terminal or a server. The user terminal can be a desktop computer, etc. The server can be an independent server, or 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 communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms, but is not limited thereto. This method can include the following steps:

[0025] S101. Obtain the in-bank data, external data, and public data of the offshore customer in the target bank; the external data includes international service provider data, and the international service provider data includes the first risk list and equity penetration map provided by the international service provider, and the public data includes the second risk list provided by other institutions and the registered business information of the offshore customer.

[0026] Among them, offshore customers refer to companies registered overseas and opening offshore accounts in China. Among them, in-bank data may include historical transaction behavior data. Historical transaction behavior data may include transaction flow data of offshore accounts. Transaction flow data may include data such as account information of both parties to the transaction, transaction time, and transaction amount. In one embodiment, in-bank data may also include due diligence information and account opening information. Due diligence information may include on-site due diligence information and / or remote due diligence information. The due diligence information may specifically include basic enterprise information. The basic enterprise information may include the enterprise registration address, enterprise office address, number of enterprise employees, and annual turnover of the enterprise. The account opening information may include indication information on whether to open an account remotely and indication information on whether to open accounts in batches.

[0027] Among them, external data may include international service provider data. International service provider data may include a first risk list and equity penetration data. The first risk list may include international service providers, such as a list of defaulters and abnormal objects provided by international authoritative credit rating agencies (including information on defaulter enterprises, etc., which is a type of abnormal object list), and a downgrade watch list (which is a type of suspicious list). The equity penetration data may include an equity penetration diagram. The equity penetration diagram can reflect information such as the equity situation of an enterprise. In one embodiment, the external data may also include customs data. The customs data may include data such as the tax refund information of offshore customers.

[0028] Among them, public data may include a second risk list and the registered business information of offshore customers. In one embodiment, the second risk list may include a list of abnormal objects under international sanctions (including enterprise identification information of enterprises sanctioned by the target country, which is a type of abnormal object list), and an international suspicious list (including enterprise identification information of enterprises involved in legal cases in the target country, such as designated legal cases or any legal cases, which is a type of suspicious list). The registered business information may include the business duration and the country of registration. In one embodiment, the public data may include the industrial and commercial information of offshore customers. The registered business information may be obtained based on the industrial and commercial information. The industrial and commercial information may include enterprise identification information (such as enterprise name), enterprise legal person information, enterprise registration address, enterprise establishment time, enterprise registered capital, enterprise registration years, and other information. The legal person referred to in the embodiments of this application is the legal representative.

[0029] S102. Update the equity penetration diagram by using the in-bank data, the first risk list, and the second risk list to obtain an updated equity penetration diagram.

[0030] Among them, the equity penetration diagram may include the enterprise identification information of offshore customers, the identification information of individuals having a holding relationship with offshore customers (such as the names of individuals) or the enterprise identification information of the first enterprise (referring to the enterprise having a holding relationship with offshore customers), and the holding relationship corresponding to offshore customers. In one embodiment, the equity penetration diagram may further include the identification information of individuals having a holding relationship with the first enterprise or the enterprise identification information of the second enterprise (referring to the enterprise having a holding relationship with the first enterprise) having a holding relationship with the first enterprise, the holding relationship corresponding to the first enterprise, and so on.

[0031] In one embodiment, the updated equity penetration diagram may further include the enterprise identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, and the enterprise attribute information of the offshore customer. The target enterprise refers to the enterprise having a transaction relationship with the offshore customer. The transaction relationship can be understood as a resource transaction relationship, such as a fund transaction relationship. The transaction attribute information may include information such as the transaction time, transaction amount, and number of transaction times with the target enterprise.

[0032] In one embodiment, when the in-line data includes historical transaction behavior data, due diligence information, and account opening information, the updated equity penetration map can be obtained in the following manner: The computer device performs entity recognition on the historical transaction behavior data to obtain the identification information of the target enterprise having a transaction relationship with the offshore customer; the computer device performs relationship extraction on the historical transaction behavior data to obtain the transaction relationship between the offshore customer and the target enterprise; the computer device performs attribute extraction on the historical transaction behavior data to obtain the transaction attribute information of the offshore customer; the computer device performs attribute extraction on the due diligence information and the account opening information to obtain the first enterprise attribute information of the offshore customer; the computer device generates labels indicating whether the offshore customer is a customer on the abnormal object list and indicating whether the offshore customer is a suspicious customer based on the first risk list and the second risk list as the second enterprise attribute information of the offshore customer; the computer device updates the equity penetration map according to the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer to obtain the updated equity penetration map. In one embodiment, the computer device can obtain the label indicating that the offshore customer is a customer on the abnormal object list when determining that the abnormal object list provided by the international service provider includes the enterprise identification information of the offshore customer and / or when determining that the abnormal object list provided by other institutions includes the enterprise identification information of the offshore customer. The fact that the offshore customer has this label indicates that the offshore customer is a customer on the abnormal object list. The computer device can also obtain the label indicating that the offshore customer is a suspicious customer when determining that the suspicious list provided by the international service provider includes the enterprise identification information of the offshore customer and / or when determining that the suspicious list provided by other institutions includes the enterprise identification information of the offshore customer. The fact that the offshore customer has this label indicates that the offshore customer is suspicious. Among them, the above-mentioned updated equity penetration map can also include the enterprise identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer. The first enterprise attribute information can include the indication information of whether the enterprise has an office in another location, the number of employees, the annual turnover, the indication information of whether it has an account opened in another location, and the indication information of whether it has opened accounts in batches.

[0033] In one embodiment, when the external data further includes the aforementioned customs data and the public data further includes the aforementioned industrial and commercial information, the computer device may further determine the customs data and the industrial and commercial information as the third enterprise attribute information of the offshore customer. Accordingly, the way for the computer device to update the equity penetration diagram based on the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer to obtain the updated equity penetration diagram may be: The computer device updates the equity penetration diagram based on the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, the second enterprise attribute information of the offshore customer, and the third enterprise attribute information of the offshore customer to obtain the updated equity penetration diagram. Among them, the updated equity penetration diagram may include the enterprise identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, the second enterprise attribute information of the offshore customer, and the third enterprise attribute information of the offshore customer.

[0034] In one embodiment, when the in-house data includes historical transaction behavior data, the updated equity penetration diagram may be obtained in the following manner: The computer device performs entity recognition on the historical transaction behavior data to obtain the identification information of the target enterprise having a transaction relationship with the offshore customer; The computer device performs relationship extraction on the historical transaction behavior data to obtain the transaction relationship between the offshore customer and the target enterprise; The computer device performs attribute extraction on the historical transaction behavior data to obtain the transaction attribute information of the offshore customer; The computer device generates labels indicating whether the offshore customer is a customer on the abnormal object list and indicating whether the offshore customer is a suspicious customer based on the first risk list and the second risk list as the second enterprise attribute information of the offshore customer; The computer device updates the equity penetration diagram based on the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer to obtain the updated equity penetration diagram. Among them, the updated equity penetration diagram further includes the enterprise identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer.

[0035] In one embodiment, when the external data further includes the aforementioned customs data and the public data further includes the aforementioned industrial and commercial information, the computer device may further determine the customs data and the industrial and commercial information as the third enterprise attribute information of the offshore customer. Accordingly, the manner in which the computer device updates the equity penetration diagram according to the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer to obtain the updated equity penetration diagram may be: The computer device updates the equity penetration diagram according to the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the second enterprise attribute information of the offshore customer, and the third enterprise attribute information of the offshore customer. The updated equity penetration diagram may further include the third enterprise attribute information of the offshore customer. Among them, the updated equity penetration diagram further includes the enterprise identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the second enterprise attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer.

[0036] S103. When determining that the offshore customer is a first type of customer according to the updated equity penetration diagram, obtain the target transaction behavior feature set of the offshore customer and the holding data of the offshore customer.

[0037] In the embodiment of the present application, when determining that the offshore customer is a first type of customer according to the updated equity penetration diagram, it indicates that the specific classification of the offshore customer cannot be determined temporarily according to the updated equity penetration diagram. Therefore, the computer device may further obtain the target transaction behavior feature set of the offshore customer and the holding data of the offshore customer for further determining the category to which the offshore customer belongs. Among them, the target transaction behavior feature set is obtained according to the historical transaction data. The holding data is obtained according to the updated equity penetration diagram.

[0038] In one embodiment, the computer device may determine whether the offshore customer is a second type of customer according to the updated equity penetration diagram; the computer device may also determine whether the offshore customer is a third type of customer according to the updated equity penetration diagram; the computer device may also determine that the offshore customer is a first type of customer when determining that the offshore customer is neither a second type of customer nor a third type of customer according to the updated equity penetration diagram. In an application scenario, the first type of customer may be a customer to be observed, the second type of customer may be a customer for enhanced due diligence, and the third type of customer may be a high-quality customer. Among them, when the offshore customer is determined to be a customer for enhanced due diligence, it means that due diligence on the offshore customer should be strengthened. By strengthening due diligence on such offshore customers, key information affecting transactions of the company or individual can be discovered, thereby avoiding potential risks.

[0039] In one embodiment, the way for the computer device to determine whether an offshore customer is a second - type customer can be as follows: If the computer device determines that the offshore customer is a customer on the abnormal object list according to the updated equity penetration graph, the relationship between the offshore customer and the enterprise on the abnormal object list meets the preset conditions, the transaction behavior of the offshore customer does not conform to the regulations, or the offshore customer is suspicious, then it is determined that the offshore customer is a second - type customer.

[0040] In one embodiment, the way for the computer device to determine whether an offshore customer is a customer on the abnormal object list can be as follows: The computer device determines whether the second enterprise attribute information of the offshore customer in the updated equity penetration graph includes a label indicating that the offshore customer is a customer on the abnormal object list; if it includes, the computer device determines that the offshore customer is a customer on the abnormal object list.

[0041] In one embodiment, the way for the computer device to determine whether the relationship between an offshore customer and an enterprise on the abnormal object list meets the preset conditions can be as follows: When the enterprise identification information of any enterprise in the updated equity penetration graph appears in the abnormal object list provided by an international service provider or in the abnormal object list provided by other institutions, it is determined that the enterprise is an enterprise on the abnormal object list, and it is determined that the relationship between the offshore customer and the enterprise on the abnormal object list meets the preset conditions. Since there is a certain correlation between multiple enterprises in the updated equity penetration graph, when it is determined that there is an enterprise on the abnormal object list in the updated equity penetration graph, it can indicate that there is a certain correlation between the offshore customer and the enterprise on the abnormal object list in the updated equity penetration graph. Therefore, it can be determined that the relationship between the offshore customer and the enterprise on the abnormal object list meets the preset conditions. In one embodiment, the way for the computer device to determine whether the relationship between an offshore customer and an enterprise on the abnormal object list meets the preset conditions can be as follows: When the enterprise identification information of the target enterprise in the equity penetration graph appears in the abnormal object list provided by an international service provider or in the abnormal object list provided by other institutions, it is determined that the target enterprise is an enterprise on the abnormal object list, and it is determined that the relationship between the offshore customer and the enterprise on the abnormal object list meets the preset conditions. The target enterprise can include, for example, an enterprise with a primary - level relationship with the offshore customer. The primary - level correlation relationship indicates that the node corresponding to the target enterprise in the updated equity penetration graph is directly connected to the node corresponding to the offshore customer in the updated equity penetration graph.

[0042] In one embodiment, the computer device may invoke transaction - type rules to determine whether the transaction behavior of an offshore customer complies with regulations based on the updated equity penetration map; the transaction - type rules include judgment rules set according to the transaction attribute information of the offshore customer; and / or, the computer device may invoke static - information - type rules to determine whether the offshore customer is suspicious based on the updated equity penetration map, and the static - information - type rules include judgment rules set according to the enterprise attribute information of the offshore customer; and / or, the computer device may invoke suspicious - list - type rules to determine whether the offshore customer is suspicious based on the updated equity penetration map.

[0043] In one embodiment, the way for the computer device to invoke transaction - type rules to determine whether the transaction behavior of an offshore customer complies with regulations based on the updated equity penetration map may be as follows: The computer device determines whether the transaction behavior of the offshore customer complies with regulations according to the transaction attribute information of the offshore customer. If it does not comply with regulations, the offshore customer is determined to be a type - two customer. The updated equity penetration map includes the transaction attribute information of the offshore customer. In one embodiment, the way for the computer device to determine whether the transaction behavior of an offshore customer complies with regulations according to the transaction data information of the offshore customer is as follows:

[0044] ① When it is determined according to the transaction attribute information of the offshore customer that the transaction turnover of the offshore customer and the business scale of the offshore customer do not conform to the norm, it is determined that the transaction behavior of the offshore customer does not comply with regulations. For example, for a startup that has been established for only one or two years, if the monthly transaction turnover is too large, it indicates that there may be transaction risks for this startup. Therefore, it can be determined that the transaction behavior of this startup does not comply with regulations.

[0045] ② When it is determined according to the transaction attribute information of the offshore customer that the offshore customer has transitional fund transactions, it is determined that the transaction behavior of the offshore customer does not comply with regulations. Transitional fund transactions have the following characteristics: The amount transferred in and out daily is basically the same, and the transactions show obvious scattered transfers in.

[0046] In one embodiment, in addition to using the above - mentioned method to determine whether the transaction behavior of an offshore customer complies with regulations, the computer device may also determine whether the transaction behavior of an offshore customer complies with regulations through the following methods: Whether private - to - public and public - to - private transfers are frequent (for example, the frequency of private - to - public and public - to - private transfers is greater than a first preset frequency); and / or, whether large - amount transactions are frequent (for example, the frequency of large - amount transactions is greater than a second preset frequency); and / or, whether funds flow in and out quickly (for example, the frequency of fund transfers in and out is greater than a third preset frequency) and the remaining resource amount of the offshore account is less than a preset value (such as the balance of the offshore account is less than the preset value).

[0047] In one embodiment, the manner in which the computer device invokes the static information type rules to determine whether an offshore customer is suspicious based on the updated equity penetration map may be as follows: when the number of enterprise employees included in the first enterprise attribute information of the offshore customer is less than the preset number, the computer device determines that the offshore customer is suspicious. And / or, when the annual turnover of the enterprise included in the first enterprise attribute information of the offshore customer is less than the preset value, the computer device determines that the offshore customer is suspicious. And / or, when the first enterprise attribute information of the offshore customer includes indication information of remote work, the computer device determines that the offshore customer is suspicious. And / or, when the first enterprise attribute information of the offshore customer includes indication information of opening an account in a different location, the computer device determines that the offshore customer is suspicious. And / or, when the first enterprise attribute information of the offshore customer includes indication information of batch account opening, the computer device determines that the offshore customer is suspicious. The updated knowledge graph includes the first enterprise attribute information of the offshore customer.

[0048] In one embodiment, the computer device may also determine that the offshore customer is suspicious when the customs data included in the third enterprise attribute information of the offshore customer does not match the corresponding transaction data. For example, the tax refund amount does not match the transaction amount. For example, the export transaction amount is 10 million, but the export tax refund amount is only 5 million, indicating that the offshore customer has a suspicion of tax evasion. Therefore, it can be determined that the offshore customer is suspicious.

[0049] In one embodiment, the manner in which the computer device invokes the suspicious list type rules to determine whether an offshore customer is suspicious based on the updated equity penetration map may be as follows: when the second enterprise attribute information of the offshore customer includes a label indicating that the offshore customer is a suspicious customer, the computer device determines that the offshore customer is a customer subject to enhanced due diligence. The updated knowledge graph includes the second enterprise attribute information of the offshore customer.

[0050] In one embodiment, the manner in which the computer device determines whether an offshore customer is a third type of customer may be as follows: the computer device may also determine that the offshore customer is a third type of customer when it determines that the enterprise type of the offshore customer is the target type based on the updated equity penetration map. Among them, the target type may be an enterprise type indicating that the enterprise is relatively reliable, such as a listed enterprise, state-owned holding, state-owned enterprise or central enterprise.

[0051] In one embodiment, the manner in which the computer device obtains the holding data of the offshore customer may be that the computer device determines the number of companies held by the offshore customer and / or the number of companies held by the legal representative of the offshore customer based on the updated equity penetration map, and determines the number of companies held by the offshore customer and / or the number of companies held by the legal representative of the offshore customer as the holding data of the offshore customer.

[0052] S104. Invoke the offshore customer classification model to classify the offshore customer based on the set of target transaction behavior characteristics, the holding data of the offshore customer, and the registered business information of the offshore customer, so as to obtain the category of the offshore customer.

[0053] In the embodiment of the present application, the computer device may input the set of target transaction behavior characteristics, the holding data, and the registered business information into the offshore customer classification model for classification processing to obtain the category of the offshore customer. It should be noted that the offshore customer mentioned in step S103 belonging to the first category of customers is not the final classification result of the offshore customer. The category of the offshore customer obtained through classification in step S104 is the final classification result of the offshore customer. The above classification process mainly relies on the set of target transaction behavior characteristics for classification, and the holding data and the registered business information play an auxiliary classification role. The final classification result of the offshore customer can be one of the first category of customers, the second category of customers, and the third category of customers.

[0054] In one embodiment, the computer device may manage the transaction quota for the offshore customer according to the category to which the offshore customer belongs. For example, when the offshore customer is a first category of customer, the upper limit of the transaction quota of the offshore customer's offshore account may be adjusted to a target value. When the transaction quota of the offshore account exceeds the target value, a prompt message indicating insufficient quota is fed back to the offshore customer. When the offshore customer is a second category of customer, when the second category of customer uses the offshore account for transactions, the offshore customer may be prompted to perform identity authentication first or to submit relevant materials to the bank counter for data review. In one embodiment, after the identity authentication of the offshore customer is passed or the data review is passed, a transaction quota may be allocated to the offshore customer so that the offshore customer can conduct transactions based on the allocated transaction quota. When the offshore customer is a third category of customer, the restriction on the transaction quota of the offshore customer's offshore account may be cancelled. Through the above transaction quota management strategy, potential transaction risks can be reduced.

[0055] It can be seen that Figure 1 In the embodiment shown, the computer device may obtain the in-bank data, external data, and public data of the offshore customer in the target bank; the computer device updates the equity penetration map by using the in-bank data, the first risk list included in the external data, and the second risk list included in the public data to obtain an updated equity penetration map; when the computer device determines that the offshore customer is a first category of customer according to the updated equity penetration map, it obtains the set of target transaction behavior characteristics of the offshore customer and the holding data of the offshore customer, and invokes the offshore customer classification model to classify the offshore customer based on the set of target transaction behavior characteristics, the holding data of the offshore customer, and the registered business information of the offshore customer, so as to obtain the category of the offshore customer. This process can improve the accuracy of identifying the category of the offshore customer.

[0056] Please refer to Figure 2, which is a schematic flowchart of a data analysis method provided by another embodiment of the present application. This method can be applied to the computer device mentioned above. Specifically, this method may include the following steps:

[0057] S201. Obtain the historical transaction behavior data of each sample customer among multiple sample customers, the shareholding data of each sample customer, and the registered business information of each sample customer.

[0058] S202. Process the historical transaction behavior data of each sample customer to obtain a transaction behavior feature set of each sample customer, where the transaction behavior feature set includes multiple transaction behavior features.

[0059] In an embodiment of the present application, the manner in which the computer device can process the historical transaction behavior data of each sample customer to obtain the transaction behavior feature set of each sample customer is that the computer device statistically analyzes the transaction behavior data of the sample customer in each time window among multiple time windows and the transaction behavior data of the sample customer in each frequency window among multiple frequency windows according to the historical transaction behavior data of each sample customer; the computer device processes the transaction behavior data of the sample customer in each time window through a first discrete function and a first continuous function to obtain the transaction behavior feature corresponding to each time window of the sample customer; the computer device processes the transaction behavior data of the sample customer in each frequency window through a second discrete function and a second continuous function to obtain the transaction behavior feature corresponding to each frequency window of the sample customer; the computer device constructs the transaction behavior feature set of each sample customer, and the transaction behavior feature set includes the transaction behavior features corresponding to each time window of the sample customer and the transaction behavior features corresponding to each frequency window of the sample customer. The above solution generates multi-dimensional transaction behavior features by using feature engineering, enriching the transaction behavior features.

[0060] In one embodiment, the computer device may invoke the RFM (Recency, Frequency, Monetary) algorithm to process the historical transaction behavior data of each sample customer to obtain a set of transaction behavior characteristics for each sample customer. Specifically, the computer device may count the transaction behavior data of each sample customer in each time window among multiple time windows and the transaction behavior data of each sample customer in each frequency window among multiple frequency windows according to the historical transaction behavior data of each sample customer; the computer device processes the transaction behavior data of each sample customer in each time window through a first discrete function and a first continuous function to obtain the transaction behavior characteristics corresponding to each time window of the sample customer, and processes the transaction behavior data of each sample customer in each frequency window through a second discrete function and a second continuous function to obtain the transaction behavior characteristics corresponding to each frequency window of the sample customer; the computer device constructs a set of transaction behavior characteristics for each sample customer, and the set of transaction behavior characteristics includes the transaction behavior characteristics corresponding to each time window and the transaction behavior characteristics corresponding to each frequency window.

[0061] Among them, the multiple time windows may include multiple short-term time windows and multiple long-term time windows, and the time lengths corresponding to each short-term time window are different, and the time lengths corresponding to each long-term time window are different. For example, the short-term time window may be in hours, and the multiple short-term time windows may include 1 hour (last 1 hour), 24 hours (last 24 hours), 48 hours (last 48 hours), and so on. In one embodiment, the maximum time length of the short-term time window does not exceed 48 hours. The long-term time window may be in months, and the multiple long-term time windows may include 1 month (last 1 month), 3 months (last 3 months), 6 months (last 6 months). In one embodiment, the maximum time length of the long-term time window does not exceed 6 months. The transaction behavior data of each time window among the multiple time windows may include the number of transaction records in the time window, the transaction amount of each transaction, the transaction time of each transaction, and so on.

[0062] Among them, the multiple frequency windows may include a first preset frequency, such as 3 times (last 1 time), a second preset frequency, such as 5 times (last 5 times), and in one embodiment, the multiple frequency windows may further include a third preset frequency, such as 10 times (last 10 times), and so on. The transaction behavior data of each frequency window among the multiple frequency windows may include the number of transaction records in the frequency window, the transaction amount of each transaction, the transaction time of each transaction, and so on.

[0063] Among them, the processing methods of the first discrete function include counting statistics, frequency statistics, and ratio calculation. The trading behavior characteristics obtained by specifically adopting this method may include: the number of days with transactions in the time window, the number of trading days when the account balance is less than the preset amount in the time window, and the ratio between the number of trading days when the account balance is less than the preset amount and the total number of trading days in the time window. For example, the long-term time window is 1 month. Among them, the number of days with transactions is 20 days, and the number of trading days when the account balance is less than the preset amount within these 20 days is 10 days. If the number of trading days when the account balance is less than the preset amount within these 20 days is 10 days, the ratio calculated here is 1 / 2.

[0064] Among them, the processing methods of the first continuous function may include maximum value calculation, minimum value calculation, cumulative value calculation, and mean value calculation. The trading behavior characteristics obtained in this way may include the maximum trading amount corresponding to the time window, the minimum trading amount, the cumulative trading amount, and the average trading amount. For example, multiple time windows include 1 hour, 12 hours, 1 month, and 3 months. After processing the trading behavior data of these time windows through the first continuous function, the maximum trading amount, minimum trading amount, cumulative trading amount, and average trading amount within 1 hour (as the trading behavior characteristics of the 1-hour time window) can be obtained. The maximum trading amount, minimum trading amount, cumulative trading amount, and average trading amount within 12 hours (as the trading behavior characteristics of the 12-hour time window) can also be obtained. The maximum trading amount, minimum trading amount, cumulative trading amount, and average trading amount within 1 month (as the trading behavior characteristics of the 1-month time window) can also be obtained, and the maximum trading amount, minimum trading amount, cumulative trading amount, and average trading amount within 3 months (as the trading behavior characteristics of the 3-month time window) can be obtained.

[0065] Among them, the processing methods of the second discrete function may include counting statistics, frequency statistics, and ratio calculation. The trading behavior characteristics obtained by specifically adopting this method may include: the number of days with transactions in the frequency window, the number of trading days when the account balance is less than the preset amount in the frequency window, and the ratio between the number of trading days when the account balance is less than the preset amount and the total number of trading days in the frequency window.

[0066] Among them, the processing methods of the second continuous function can include maximum value calculation, minimum value calculation, cumulative value calculation, and mean value calculation. The transaction behavior characteristics obtained in this way can include the maximum transaction amount corresponding to the frequency window, the minimum transaction amount, the cumulative transaction amount, and the average transaction amount. For example, if multiple frequency windows include 3 times, 5 times, and 10 times, after processing the transaction behavior data of these frequency windows through the second continuous function, the maximum transaction amount, the minimum transaction amount, the cumulative transaction amount, and the average transaction amount of the last 3 transactions (transaction behavior characteristics corresponding to 3 times) can be obtained, and the maximum transaction amount, the minimum transaction amount, the cumulative transaction amount, and the average transaction amount of the last 5 transactions (transaction behavior characteristics corresponding to 5 times) can be obtained, and the maximum transaction amount, the minimum transaction amount, the cumulative transaction amount, and the average transaction amount of 10 transactions (transaction behavior characteristics corresponding to 10 times) can be obtained.

[0067] It should be noted that feature engineering is the core of the artificial intelligence AI algorithm model, and the quality of feature engineering directly determines the performance and effect of the algorithm model. To find more potential risk customers and accurately perceive risks, the embodiments of this application can construct hundreds of thousands of dimensions of transaction behavior features through an explosive feature derivation scheme based on RFM, accurately depict the target of each account movement transaction, and screen out high-quality indicators for distinguishing good and bad samples from them, and strictly control the dimension of the variables entering the model. Some information on the feature derivation scheme can be seen in the following table:

[0068]

[0069]

[0070] S203. Perform feature screening on the transaction behavior feature set of each sample customer to obtain a sub-transaction behavior feature set of each sample customer, where the sub-transaction behavior feature set includes at least one transaction behavior feature for distinguishing the category to which each sample customer belongs.

[0071] Among them, the transaction behavior feature set of each sample customer corresponds to multiple feature categories. The transaction behavior feature set includes the transaction behavior features corresponding to each feature category in the multiple feature categories. Among them, the multiple feature categories corresponding to the transaction behavior feature set of each sample customer are the same. For example, the transaction behavior feature set of sample customer 1 corresponds to feature categories 1... feature category m. m is a positive integer. Then the transaction behavior feature set of sample customer 2 also corresponds to feature categories 1... feature category m. The sub-transaction behavior feature set of each sample customer corresponds to at least one feature category. The sub-transaction behavior feature set includes the transaction behavior features corresponding to each feature category in the at least one feature category. Among them, the at least one feature category corresponding to the sub-transaction behavior feature set of each sample customer is the same. For example, the sub-transaction behavior feature set of sample customer 1 corresponds to feature categories 1... feature category n. Then the sub-transaction behavior feature set of sample customer 2 also corresponds to feature categories 1... feature category n, where n is a positive integer less than m.

[0072] In one embodiment, the computer device can calculate the Information Value (IV) value of each feature category in the multiple feature categories according to the transaction behavior feature set of each sample customer, and sort the multiple feature categories according to the IV value, so as to obtain the sub-transaction behavior feature set of each sample customer according to the sorting result. Among them, the sub-transaction behavior feature set of the sample customer can include the transaction behavior features corresponding to each feature type among the top K feature types. The sorting method can be sorting from front to back according to the size of the IV value, and the larger the IV value, the more forward the sorting. The value range of the IV value is [0, positive infinity). The above process can realize the selection of variables to be included in the model.

[0073] S204. Use the sub-transaction behavior feature set of each sample customer, the holding data of each sample customer, and the registered business information of each sample customer to train the initial deep learning model to obtain the trained deep learning model as the offshore customer classification model. The target transaction behavior feature set is the sub-transaction behavior feature set of the offshore customer.

[0074] In the embodiments of the present application, a computer device may obtain a plurality of training samples based on the sub-transaction behavior feature sets of each sample customer, the holding data of each sample customer, and the registered business information of each sample customer, and use the plurality of training samples to train an initial deep learning model to obtain a trained deep learning model as an offshore customer classification model. A training sample may include the sub-transaction behavior feature set of a sample customer, the holding data of this sample customer, and the registered business information of this sample customer. The deep learning model may be a boosting model, such as a Light Gradient Boosting Machine (LightGBM) model.

[0075] It should be noted that for conventional models such as the Extreme Gradient Boosting (XGBoost) model and the Gradient Boosting Decision Tree (GBDT) model, calculating information gain requires scanning all training samples to find the optimal splitting point. When facing high-dimensional large amounts of data or features, the efficiency and scalability seriously affect the model's performance. The most direct and effective way to solve this problem is to reduce the sample size and the number of features without affecting the accuracy. The LightGBM algorithm is an optimization of the traditional GBDT algorithm, which incorporates algorithms such as Gradient-based One-Side Sampling (GOSS) and Exclusive Feature Bundling (EFB). Among them, the GOSS algorithm speeds up the calculation by sampling samples and using samples with large gradients to calculate information gain, and the EFB algorithm further improves the model training speed through feature sampling, thereby enhancing the learning efficiency.

[0076] The processing flow of the GOSS algorithm may include the following:

[0077] a. Sort the training samples in descending order according to the absolute value of the gradient of each training sample, and take the first a training samples as sample subset A;

[0078] b. Randomly sample b training samples from the training samples other than the first a training samples as sample subset B;

[0079] c. Estimate the information gain based on sample subset A and sample subset B. The formula for estimating the information gain is as follows:

[0080]

[0081] Where:

[0082] A l ={x i ∈A: x ij ≤d}, A r ={xi ∈A: x ij > d}

[0083] B l = {x i ∈B: x ij ≤d}, B r = {x i ∈B: x ij > d}

[0084] where O represents the training samples of the node to be split in the decision tree, gi is the gradient of the training samples, is the number of samples of the left child node of node d, is the number of samples of the right child node of node d, represents the weight coefficient of small gradient samples.

[0085] Among them, the processing flow of the EFB algorithm may include the following:

[0086] a. Sort each transaction behavior feature according to the number of non-zero values;

[0087] b. Calculate the conflict ratio between different transaction behavior features;

[0088] c. Traverse each transaction behavior feature and perform feature merging, calculate the conflict ratio between different merged features, minimize the conflict ratio, and obtain the trained LightGBM model.

[0089] S205. Obtain the in-bank data, external data, and public data of the offshore customer in the target bank; the external data includes international service provider data, and the international service provider data includes the first risk list and the equity penetration map provided by the international service provider, and the public data includes the second risk list provided by other institutions and the registered business information of the offshore customer.

[0090] S206. Update the equity penetration map by using the in-bank data, the first risk list, and the second risk list to obtain the updated equity penetration map.

[0091] S207. When determining that the offshore customer is a first-class customer according to the updated equity penetration map, obtain the target transaction behavior feature set of the offshore customer and the holding data of the offshore customer;

[0092] S208. Call the offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer.

[0093] Among them, for steps S205 - S208, reference can be made toFigure 1 Steps S101 - S104 in the embodiments will not be elaborated here.

[0094] In one embodiment, the way for the computer device to obtain the target transaction behavior feature set of the offshore customer can be that the computer device obtains the historical transaction behavior data of the offshore customer, and processes the historical transaction behavior data of the offshore customer to obtain the sub - transaction behavior feature set of the offshore customer as the target transaction behavior feature set of the offshore customer. Here, since before training the initial deep - learning model, at least one feature type has been determined from multiple feature types through feature screening, for example, the top K feature types are determined from multiple feature types. Therefore, the sub - transaction behavior feature set of the offshore customer can include the at least one feature category, such as the transaction behavior features corresponding to each of the top K feature types. That is to say, in the above process, the computer device can no longer perform the operation of processing the historical transaction behavior data to obtain the transaction behavior feature set of the offshore customer and then performing feature screening based on the transaction behavior feature set of the offshore customer, which can improve the classification efficiency of the offshore customer.

[0095] In one embodiment, the way for the computer device to obtain the target transaction behavior feature set of the offshore customer can be that the computer device obtains the historical transaction behavior data of the offshore customer, and processes the historical transaction behavior data of the offshore customer to obtain the transaction behavior feature set of the offshore customer; the computer device performs feature screening on the transaction behavior feature set of the offshore customer to obtain the sub - transaction behavior feature set of the offshore customer as the target transaction behavior feature set of the offshore customer. Among them, the way for the computer device to process the historical transaction behavior data of the offshore customer to obtain the transaction behavior feature set of the offshore customer can refer to the way for the computer device to process the historical transaction behavior data of the sample customer to obtain the transaction behavior feature set of the sample customer, which will not be elaborated here. The way for the computer device to perform feature screening on the transaction behavior feature set of the offshore customer to obtain the sub - transaction behavior feature set of the offshore customer can refer to the way for the computer device to perform feature screening on the transaction behavior feature set of the sample customer to obtain the sub - transaction behavior feature set of the sample customer, which will not be elaborated here.

[0096] According to the cross-border offshore transaction mode and compliance requirements, the embodiments of the present application introduce in-bank data, external data, and public data and provide a combined solution of a rule model and an AI algorithm model. First, it can automatically classify the types of cross-border offshore customers, and then adopt different management methods for offshore customers at different levels, so as to systematically and intelligently intercept and control the abnormal operations of cross-border offshore customers, realizing the transition of risk prevention from manual monitoring to system monitoring and comprehensively improving the risk prevention and control ability of offshore customers. Compared with the existing technology of manually conducting risk investigation on offshore customers by combining data such as the list of abnormal objects provided by a single bank, the embodiments of the present application automatically and intelligently determine the category of offshore customers by combining multi-dimensional information, which is higher than the existing technology in terms of risk investigation efficiency and reliability and can meet the current risk prevention and control requirements for offshore customers. Moreover, the embodiments of the present application analyze by integrating multiple data sources and with the help of a knowledge graph, enriching the existing risk monitoring dimensions and solving the pain point of the current inability to comprehensively and penetratively "understand the customer". In addition, the embodiments of the present application use the LightGBM algorithm to solve the problems of low operation efficiency and poor scalability caused by using conventional models, and the embodiments of the present application use the leaf-wise strategy with depth limit and perform multi-threaded optimization, which can reduce the model error and improve the model accuracy. In summary, the embodiments of the present application comprehensively apply the "AI algorithm model + rule model", enrich the risk monitoring means for cross-border offshore customers, make up for the deficiencies of the existing rule models, and comprehensively improve the accuracy of identifying risk customers.

[0097] It can be seen that Figure 2 In the shown embodiments, the computer device can obtain the historical transaction behavior data of each sample customer, the holding data of each sample customer, and the registered business information of each sample customer among multiple sample customers; the computer device processes the historical transaction behavior data of each sample customer to obtain the transaction behavior feature set of each sample customer, and performs feature screening on the transaction behavior feature set of each sample customer to obtain the sub-transaction behavior feature set of each sample customer; the computer device uses the sub-transaction behavior feature set of each sample customer, the holding data of each sample customer, and the registered business information of each sample customer to train an initial deep learning model to obtain the trained deep learning model as an offshore customer classification model. Training the model based on the screened features can improve the model training efficiency and avoid the adverse effects of useless features on the prediction effect of the model.

[0098] The present application relates to blockchain technology. For example, various types of data or specified types of data involved in the embodiments of the present application can be obtained through a blockchain network. The various types of data involved in the embodiments of the present application have obtained the use authorization of the corresponding data subject or data controller.

[0099] Please refer toFigure 3 , which is a schematic structural diagram of a data analysis device provided by an embodiment of the present application. This device can be applied to the computer device mentioned above. Specifically, the data analysis device may include:

[0100] An acquisition module 301, configured to acquire in-bank data, external data, and public data of an offshore customer in a target bank; the external data includes international service provider data, and the international service provider data includes a first risk list and an equity penetration map provided by an international service provider, and the public data includes a second risk list provided by other institutions and the registered business information of the offshore customer.

[0101] An update module 302, configured to update the equity penetration map by using the in-bank data, the first risk list, and the second risk list to obtain an updated equity penetration map.

[0102] The acquisition module 301 is further configured to, when determining that the offshore customer is a first-type customer according to the updated equity penetration map, acquire a target transaction behavior feature set of the offshore customer and the holding data of the offshore customer.

[0103] A classification module 303, configured to call an offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer.

[0104] In an optional implementation manner, the in-bank data includes historical transaction behavior data, due diligence information, and account opening information, and the update module 302 is specifically configured to:

[0105] Perform entity recognition on the historical transaction behavior data to obtain identification information of a target enterprise having a transaction relationship with the offshore customer;

[0106] Perform relationship extraction on the historical transaction behavior data to obtain a transaction relationship between the offshore customer and the target enterprise;

[0107] Perform attribute extraction on the historical transaction behavior data to obtain transaction attribute information of the offshore customer;

[0108] Perform attribute extraction on the due diligence information and the account opening information to obtain first enterprise attribute information of the offshore customer;

[0109] Generate labels indicating whether the offshore customer is a list customer of an abnormal object and indicating whether the offshore customer is a suspicious customer according to the first risk list and the second risk list as second enterprise attribute information of the offshore customer;

[0110] Update the equity penetration map based on the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer, to obtain an updated equity penetration map.

[0111] In an alternative embodiment, the apparatus may further include a determination module 304. The determination module 304 is configured to:

[0112] If it is determined according to the updated equity penetration map that the offshore customer is a customer on the abnormal object list, the relationship between the offshore customer and the enterprise on the abnormal object list meets the preset conditions, the transaction behavior of the offshore customer does not conform to the regulations, or the offshore customer is suspicious, then determine that the offshore customer is a second type of customer;

[0113] If it is determined according to the updated equity penetration map that the enterprise type of the offshore customer is the target type, then determine that the offshore customer is a third type of customer;

[0114] If the offshore customer is not the second type of customer and not the third type of customer, then determine that the offshore customer is a first type of customer.

[0115] In an alternative embodiment, the apparatus may further include a judgment module 305. The judgment module 305 is configured to:

[0116] Invoke transaction class rules to judge whether the transaction behavior of the offshore customer conforms to the regulations according to the updated equity penetration map; the transaction class rules include judgment rules set according to the transaction attribute information of the offshore customer; and / or,

[0117] Invoke static information class rules to judge whether the offshore customer is suspicious according to the updated equity penetration map, the static information class rules include judgment rules set according to the enterprise attribute information of the offshore customer; and / or,

[0118] Invoke suspicious list class rules to judge whether the offshore customer is suspicious according to the updated equity penetration map.

[0119] In an alternative embodiment, the apparatus may further include a training module 306. The training module 306 is configured to:

[0120] Obtain the historical transaction behavior data of each sample customer among multiple sample customers, the holding data of each sample customer, and the registered business information of each sample customer;

[0121] Processing the historical transaction behavior data of each sample customer to obtain a set of transaction behavior characteristics for each sample customer, where the set of transaction behavior characteristics includes multiple transaction behavior characteristics;

[0122] Performing feature screening on the set of transaction behavior characteristics of each sample customer to obtain a set of sub-transaction behavior characteristics for each sample customer, where the set of sub-transaction behavior characteristics includes at least one transaction behavior characteristic for distinguishing the category to which each sample customer belongs;

[0123] Training an initial deep learning model using the set of sub-transaction behavior characteristics of each sample customer, the holding data of each sample customer, and the registered business information of each sample customer to obtain a trained deep learning model as an offshore customer classification model, where the target set of transaction behavior characteristics is the set of sub-transaction behavior characteristics of the offshore customer.

[0124] In an alternative embodiment, the training module 306 is specifically configured to:

[0125] Processing the historical transaction behavior data of each sample customer to obtain a set of transaction behavior characteristics for each sample customer includes:

[0126] Statistical analysis of the transaction behavior data of each sample customer in each time window among multiple time windows and the transaction behavior data of each sample customer in each frequency window among multiple frequency windows based on the historical transaction behavior data of each sample customer;

[0127] Processing the transaction behavior data of the sample customer in each time window through a first discrete function and a first continuous function to obtain the transaction behavior characteristics corresponding to the sample customer in each time window;

[0128] Processing the transaction behavior data of the sample customer in each frequency window through a second discrete function and a second continuous function to obtain the transaction behavior characteristics corresponding to the sample customer in each frequency window;

[0129] Constructing a set of transaction behavior characteristics for each sample customer, where the set of transaction behavior characteristics includes the transaction behavior characteristics corresponding to the sample customer in each time window and the transaction behavior characteristics corresponding to the sample customer in each frequency window.

[0130] In an alternative embodiment, the obtaining module 301 is further specifically configured to determine the number of companies held by the offshore customer and / or the number of companies held by the legal representative of the offshore customer according to the updated equity penetration map; and determine the number of companies held by the offshore customer and / or the number of companies held by the legal representative of the offshore customer as the holding data of the offshore customer.

[0131] It can be seen that Figure 3 In the illustrated embodiment, the data analysis device can obtain in-bank data, external data, and public data of an offshore customer in a target bank; the data analysis device updates the equity penetration map by using the in-bank data, the first risk list included in the external data, and the second risk list included in the public data to obtain an updated equity penetration map; when the data analysis device determines that the offshore customer is a first-type customer according to the updated equity penetration map, it obtains the target transaction behavior feature set of the offshore customer and the holding data of the offshore customer, and calls an offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer. This process can improve the accuracy of identifying the category of the offshore customer.

[0132] Please refer to Figure 4 , which is a schematic structural diagram of a computer device provided by an embodiment of the present application. The computer device described in this embodiment may include: one or more processors 1000 and a memory 2000. The processor 1000 and the memory 2000 may be connected through a bus.

[0133] The processor 1000 may be a central processing module (Central Processing Unit, CPU), and this processor may also be other general-purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), field-programmable gate arrays (Field-Programmable Gate Array, FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc.

[0134] The memory 2000 may be a high-speed RAM memory or a non-volatile memory, such as a disk memory. Among them, the memory 2000 is used to store a computer program, and the computer program includes program instructions. The processor 1000 is configured to call the program instructions to execute the following steps:

[0135] Obtain in-bank data, external data, and public data of an offshore customer in a target bank; the external data includes international service provider data, and the international service provider data includes a first risk list and an equity penetration map provided by the international service provider, and the public data includes a second risk list provided by other institutions and the registered business information of the offshore customer;

[0136] Update the equity penetration map using the in-line data, the first risk list, and the second risk list to obtain an updated equity penetration map;

[0137] When determining that the offshore customer is a first-class customer based on the updated equity penetration map, obtain the target transaction behavior feature set of the offshore customer and the holding data of the offshore customer;

[0138] Call the offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer.

[0139] In one embodiment, the in-line data includes historical transaction behavior data, due diligence information, and account opening information. When updating the equity penetration map using the in-line data, the first risk list, and the second risk list to obtain an updated equity penetration map, the processor 1000 is configured to call the program instructions and specifically perform the following steps:

[0140] Perform entity recognition on the historical transaction behavior data to obtain the identification information of the target enterprise having a transaction relationship with the offshore customer;

[0141] Perform relationship extraction on the historical transaction behavior data to obtain the transaction relationship between the offshore customer and the target enterprise;

[0142] Perform attribute extraction on the historical transaction behavior data to obtain the transaction attribute information of the offshore customer;

[0143] Perform attribute extraction on the due diligence information and the account opening information to obtain the first enterprise attribute information of the offshore customer;

[0144] Generate labels indicating whether the offshore customer is an abnormal object list customer and indicating whether the offshore customer is a suspicious customer according to the first risk list and the second risk list as the second enterprise attribute information of the offshore customer;

[0145] Update the equity penetration map according to the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer to obtain an updated equity penetration map.

[0146] In one embodiment, the processor 1000 is configured to call the program instructions and further perform the following steps:

[0147] If it is determined according to the updated equity penetration chart that the offshore customer is a customer on the abnormal object list, the relationship between the offshore customer and the enterprise on the abnormal object list meets the preset conditions, the transaction behavior of the offshore customer does not conform to the regulations, or the offshore customer is suspicious, then it is determined that the offshore customer is a type-II customer;

[0148] If it is determined according to the updated equity penetration chart that the enterprise type of the offshore customer is the target type, then it is determined that the offshore customer is a type-III customer;

[0149] If the offshore customer is not the type-II customer and not the type-III customer, then it is determined that the offshore customer is a type-I customer.

[0150] In one embodiment, the processor 1000 is configured to call the program instructions and further perform the following steps:

[0151] Call the transaction class rules to determine whether the transaction behavior of the offshore customer conforms to the regulations according to the updated equity penetration chart; the transaction class rules include the judgment rules set according to the transaction attribute information of the offshore customer; and / or,

[0152] Call the static information class rules to determine whether the offshore customer is suspicious according to the updated equity penetration chart, the static information class rules include the judgment rules set according to the enterprise attribute information of the offshore customer; and / or,

[0153] Call the suspicious list class rules to determine whether the offshore customer is suspicious according to the updated equity penetration chart.

[0154] In one embodiment, the processor 1000 is configured to call the program instructions and further perform the following steps:

[0155] Obtain the historical transaction behavior data of each sample customer among multiple sample customers, the shareholding data of each sample customer, and the registered business information of each sample customer;

[0156] Process the historical transaction behavior data of each sample customer to obtain the transaction behavior feature set of each sample customer, and the transaction behavior feature set includes multiple transaction behavior features;

[0157] Perform feature screening on the transaction behavior feature set of each sample customer to obtain the sub-transaction behavior feature set of each sample customer, and the sub-transaction behavior feature set includes at least one transaction behavior feature for distinguishing the category to which each sample customer belongs;

[0158] Train an initial deep learning model using the set of sub - transaction behavior characteristics of each sample customer, the holding data of each sample customer, and the registered business information of each sample customer to obtain a trained deep learning model as an offshore customer classification model, where the target transaction behavior characteristic set is the set of sub - transaction behavior characteristics of the offshore customer.

[0159] In one embodiment, when processing the historical transaction behavior data of each sample customer to obtain the set of transaction behavior characteristics of each sample customer, the processor 1000 is configured to call the program instructions and specifically execute the following steps:

[0160] Statistically analyze the transaction behavior data of each sample customer in each time window among multiple time windows and the transaction behavior data of each sample customer in each frequency window among multiple frequency windows according to the historical transaction behavior data of each sample customer;

[0161] Process the transaction behavior data of the sample customer in each time window through a first discrete function and a first continuous function to obtain the transaction behavior characteristics corresponding to each time window of the sample customer;

[0162] Process the transaction behavior data of the sample customer in each frequency window through a second discrete function and a second continuous function to obtain the transaction behavior characteristics corresponding to each frequency window of the sample customer;

[0163] Construct the set of transaction behavior characteristics of each sample customer, where the set of transaction behavior characteristics includes the transaction behavior characteristics corresponding to each time window of the sample customer and the transaction behavior characteristics corresponding to each frequency window of the sample customer.

[0164] In one embodiment, when obtaining the holding data of the offshore customer, the processor 1000 is configured to call the program instructions and specifically execute the following steps:

[0165] Determine the number of companies held by the offshore customer and / or the number of companies held by the legal representative of the offshore customer according to the updated equity penetration map;

[0166] Determine the number of companies held by the offshore customer and / or the number of companies held by the legal representative of the offshore customer as the holding data of the offshore customer.

[0167] In specific implementation, the processor 1000 described in the embodiments of the present application can execute Figure 1 the implementation described in the Figure 2 embodiment or

[0168] In each embodiment of the present application, each functional module can be integrated into a processing module, or each module can exist physically alone, or two or more modules can be integrated into one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module.

[0169] Those of ordinary skill in the art can understand that all or part of the processes in the above-described embodiment methods can be completed by instructing relevant hardware through a computer program. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-described method embodiments. Among them, the computer-readable storage medium can be volatile or non-volatile. For example, the computer storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc. The computer-readable storage medium mainly includes a program storage area and a data storage area. Among them, the program storage area can store an operating system, application programs required for at least one function, etc.; the data storage area can store data created according to the use of blockchain nodes, etc.

[0170] Among them, the blockchain referred to in the present application is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm. Blockchain, essentially a decentralized database, is a string of data blocks associated using cryptographic methods. Each data block contains information about a batch of network transactions, which is used to verify the validity (anti-counterfeiting) of the information and generate the next block. The blockchain can include a blockchain underlying platform, a platform product service layer, and an application service layer, etc.

[0171] The above-disclosed is only a preferred embodiment of the present application. Of course, the scope of rights of the present application cannot be limited thereby. Those of ordinary skill in the art can understand all or part of the processes of the above-described embodiments, and the equivalent changes made according to the claims of the present application still fall within the scope covered by the present application.

Claims

1. A data analysis method, characterized in that, it includes: Obtaining in-bank data, external data, and public data of offshore customers in the target bank; the external data includes international service provider data, and the international service provider data includes a first risk list and an equity penetration map provided by the international service provider, and the public data includes a second risk list provided by other institutions and the registered business information of the offshore customer; Updating the equity penetration map by using the in-bank data, the first risk list, and the second risk list to obtain an updated equity penetration map; When determining that the offshore customer is a first-type customer according to the updated equity penetration map, obtaining the target transaction behavior feature set of the offshore customer and the holding data of the offshore customer; Invoking an offshore customer classification model to classify the offshore customer according to the target transaction behavior feature set, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer; Wherein, the in-bank data includes historical transaction behavior data, due diligence information, and account opening information, and updating the equity penetration map by using the in-bank data, the first risk list, and the second risk list to obtain an updated equity penetration map includes: Performing entity recognition on the historical transaction behavior data to obtain the identification information of the target enterprise having a transaction relationship with the offshore customer; Performing relationship extraction on the historical transaction behavior data to obtain the transaction relationship between the offshore customer and the target enterprise; Performing attribute extraction on the historical transaction behavior data to obtain the transaction attribute information of the offshore customer; Performing attribute extraction on the due diligence information and the account opening information to obtain the first enterprise attribute information of the offshore customer; Generating labels indicating whether the offshore customer is an abnormal object list customer and indicating whether the offshore customer is a suspicious customer according to the first risk list and the second risk list as the second enterprise attribute information of the offshore customer; Updating the equity penetration map according to the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer to obtain an updated equity penetration map.

2. The method according to claim 1, characterized in that, the method further includes: If it is determined according to the updated equity penetration map that the offshore customer is an abnormal object list customer, the relationship between the offshore customer and the abnormal object list enterprise meets a preset condition, the transaction behavior of the offshore customer does not conform to the specification, or the offshore customer is suspicious, then determining that the offshore customer is a second-type customer; If it is determined according to the updated equity penetration map that the enterprise type of the offshore customer is the target type, then determining that the offshore customer is a third-type customer; If the offshore customer is not the second-type customer and not the third-type customer, then determining that the offshore customer is a first-type customer.

3. The method according to claim 2, characterized in that, the method further includes: Invoke transaction - type rules to determine whether the trading behavior of the offshore customer complies with the regulations according to the updated equity penetration map; the transaction - type rules include judgment rules set according to the trading attribute information of the offshore customer; and / or, Invoke static - information - type rules to determine whether the offshore customer is suspicious according to the updated equity penetration map; the static - information - type rules include judgment rules set according to the enterprise attribute information of the offshore customer; and / or, Invoke suspicious - list - type rules to determine whether the offshore customer is suspicious according to the updated equity penetration map.

4. The method according to claim 1, wherein, the method further includes: Obtain the historical trading behavior data of each sample customer among multiple sample customers, the holding data of each sample customer, and the registered business information of each sample customer; Process the historical trading behavior data of each sample customer to obtain a trading behavior feature set of each sample customer, the trading behavior feature set including multiple trading behavior features; Perform feature screening on the trading behavior feature set of each sample customer to obtain a sub - trading behavior feature set of each sample customer, the sub - trading behavior feature set including at least one trading behavior feature for distinguishing the category to which each sample customer belongs; Train an initial deep - learning model using the sub - trading behavior feature set of each sample customer, the holding data of each sample customer, and the registered business information of each sample customer to obtain a trained deep - learning model as an offshore customer classification model, and the target trading behavior feature set is the sub - trading behavior feature set of the offshore customer.

5. The method according to claim 4, wherein, the processing the historical trading behavior data of each sample customer to obtain a trading behavior feature set of each sample customer includes: Statistically analyze the trading behavior data of the sample customer in each time window among multiple time windows and the trading behavior data of the sample customer in each frequency window among multiple frequency windows according to the historical trading behavior data of each sample customer; Process the trading behavior data of the sample customer in each time window through a first discrete - type function and a first continuous - type function to obtain the trading behavior features corresponding to the sample customer in each time window; Process the trading behavior data of the sample customer in each frequency window through a second discrete - type function and a second continuous - type function to obtain the trading behavior features corresponding to the sample customer in each frequency window; Construct the trading behavior feature set of each sample customer, the trading behavior feature set including the trading behavior features corresponding to the sample customer in each time window and the trading behavior features corresponding to the sample customer in each frequency window.

6. The method according to claim 1, wherein, the obtaining the holding data of the offshore customer includes: Determine the number of companies held by the offshore customer and / or the number of companies held by the legal representative of the offshore customer according to the updated equity penetration map; Determine the number of companies held by the offshore customer's holding company and / or the number of companies held by the legal representative of the offshore customer as the holding data of the offshore customer.

7. A data analysis device, characterized in that, it includes: An acquisition module, configured to acquire in-bank data, external data, and public data of an offshore customer in a target bank; the external data includes international service provider data, and the international service provider data includes a first risk list and an equity penetration map provided by an international service provider, and the public data includes a second risk list provided by other institutions and the registered business information of the offshore customer; An update module, configured to update the equity penetration map by using the in-bank data, the first risk list, and the second risk list to obtain an updated equity penetration map; The acquisition module is further configured to, when determining that the offshore customer is a first-class customer according to the updated equity penetration map, acquire a set of target transaction behavior characteristics of the offshore customer and the holding data of the offshore customer; A classification module, configured to call an offshore customer classification model to classify the offshore customer according to the set of target transaction behavior characteristics, the holding data of the offshore customer, and the registered business information of the offshore customer to obtain the category of the offshore customer; wherein, the in-bank data includes historical transaction behavior data, due diligence information, and account opening information, and the update module is specifically configured to: Perform entity recognition on the historical transaction behavior data to obtain identification information of target enterprises having a transaction relationship with the offshore customer; Perform relationship extraction on the historical transaction behavior data to obtain the transaction relationship between the offshore customer and the target enterprise; Perform attribute extraction on the historical transaction behavior data to obtain transaction attribute information of the offshore customer; Perform attribute extraction on the due diligence information and the account opening information to obtain first enterprise attribute information of the offshore customer; Generate labels indicating whether the offshore customer is an abnormal object list customer and indicating whether the offshore customer is a suspicious customer according to the first risk list and the second risk list as the second enterprise attribute information of the offshore customer; Update the equity penetration map according to the identification information of the target enterprise, the transaction relationship between the offshore customer and the target enterprise, the transaction attribute information of the offshore customer, the first enterprise attribute information of the offshore customer, and the second enterprise attribute information of the offshore customer to obtain an updated equity penetration map.

8. A computer device, characterized in that, it includes a processor and a memory, the processor and the memory are connected to each other, wherein, the memory is used to store computer program instructions, and the processor is configured to execute the program instructions to implement the method according to any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, computer program instructions are stored in the computer-readable storage medium, and when the computer program instructions are executed by a processor, they are used to execute the method according to any one of claims 1-6.

Images