Risk detection method and device for cross-border e-commerce enterprise, and storage medium

Abstract

The invention relates to a risk detection method and device for a cross-border e-commerce enterprise and a storage medium. The method comprises the steps that entity information and relation information between entities in cross-border e-commerce list data are extracted, a cross-border e-commerce knowledge graph is established, and the entities comprise subscribers, e-commerce enterprises, delivery addresses and commodities; based on the cross-border e-commerce knowledge graph, determining an order risk feature, the order risk feature being used for representing risk behaviors of an order in a transaction amount dimension, a delivery address dimension and a commodity dimension in a cross-border e-commerce transaction process; and based on the cross-border e-commerce knowledge graph and the order risk characteristics, determining a risk assessment result of the target e-commerce enterprise to indicate whether the target e-commerce enterprise has a goods collection and resale risk. According to the embodiment of the invention, the method can combine the characteristics of cross-border e-commerce services, and employs the cross-border e-commerce knowledge graph to more accurately recognize the possible collection and resale risk, thereby providing powerful support for risk management.

Description

Technical Field

[0001] This disclosure relates to the field of risk detection technology, and in particular to a risk detection method, apparatus and storage medium for cross-border e-commerce enterprises. Background Technology

[0002] Cross-border e-commerce, a new type of goods transaction born in the information age, has greatly expanded people's shopping scope and provided convenience for purchasing foreign goods. In recent years, with the deepening of economic globalization and the continuous advancement of modern information technology, cross-border e-commerce has become one of the important forces driving the improvement of the global economy. Cross-border e-commerce retail import refers to a consumption model in which domestic consumers purchase goods from overseas through third-party cross-border e-commerce platforms and have the goods shipped into the country through methods such as "bonded import" or "direct purchase import." Compared with traditional trade methods, this model can reduce time, manpower, and resource costs, simplify trade processes, reduce intermediate costs, and provide consumers with more price-competitive product choices.

[0003] Meanwhile, the sale of goods imported through cross-border e-commerce retail is subject to strict regulation, and the resale of such goods is strictly prohibited. Specifically, these goods are limited to personal use and cannot be resold. Cross-border e-commerce resale refers to the practice where goods imported from overseas are sold through cross-border e-commerce, and then the company or individual resells these goods to other consumers through markups, distribution, and transshipment. To ensure the healthy development of the market, consumers or those who purchase cross-border e-commerce retail imported goods are prohibited from reselling them.

[0004] Currently, machine learning algorithms such as clustering, classification, and association rule mining are commonly used in the field of risk detection to identify potential risks. However, the risk scenarios faced by cross-border e-commerce companies in the consolidation and resale process are complex and ever-changing. Current risk detection methods that rely solely on single data points often struggle to identify abnormal transaction behaviors in consolidation and resale risk identification scenarios in a timely and accurate manner. Summary of the Invention

[0005] In view of this, this disclosure proposes a risk detection method, device and storage medium for cross-border e-commerce enterprises.

[0006] According to one aspect of this disclosure, a risk detection method for cross-border e-commerce enterprises is provided. The method includes:

[0007] Extract entity information and inter-entity relationship information from cross-border e-commerce manifest data, and establish a cross-border e-commerce knowledge graph. Nodes in the cross-border e-commerce knowledge graph are used to indicate the corresponding entities, and edges in the cross-border e-commerce knowledge graph are used to indicate the relationship between the entities corresponding to the two nodes connecting the edges. Entities include orderers, e-commerce companies, delivery addresses, and goods.

[0008] Based on the cross-border e-commerce knowledge graph, the risk characteristics of the orderer are determined. The risk characteristics of the orderer are used to characterize the risk behavior of the orderer in multiple dimensions during the cross-border e-commerce transaction process. These multiple dimensions include transaction amount, delivery address and product.

[0009] Based on cross-border e-commerce knowledge graphs and orderer risk characteristics, the risk assessment results of target e-commerce enterprises are determined. The risk assessment results are used to indicate whether the target e-commerce enterprises have the risk of consolidation and resale.

[0010] In one possible implementation, based on a cross-border e-commerce knowledge graph, the risk characteristics of the purchaser are determined, including:

[0011] Based on the relationship attributes between the orderer and the goods in the cross-border e-commerce knowledge graph, the maximum dutiable value of the goods purchased by the orderer in each first time window is determined.

[0012] The maximum dutiable value of goods purchased in each first time window is compared with the preset purchase limit to determine the orderer's risk characteristics in terms of transaction amount. The orderer's risk characteristics in terms of transaction amount are used to indicate whether the orderer has engaged in abnormal transaction behavior that exceeds the purchase limit.

[0013] In one possible implementation, the orderer risk characteristics at the delivery address level include abnormal distribution characteristics of delivery addresses. Based on a cross-border e-commerce knowledge graph, the orderer risk characteristics are determined, including:

[0014] Based on the relationship attributes between orderers and delivery addresses in the cross-border e-commerce knowledge graph, the number of delivery addresses that the orderer used only once in the second time window and the total number of delivery addresses used by the orderer in the second time window are determined.

[0015] Based on the number of delivery addresses used only once and the total number of delivery addresses used, abnormal distribution characteristics of delivery addresses are determined. These abnormal distribution characteristics are used to indicate whether the orderer has engaged in abnormal transaction behavior by frequently changing delivery addresses.

[0016] In one possible implementation, the orderer risk characteristics at the delivery address dimension include nearest-neighbor address characteristics. Based on a cross-border e-commerce knowledge graph, the orderer risk characteristics are determined, including:

[0017] Based on the relationship attributes between the orderer and the product, and between the orderer and the delivery address in the cross-border e-commerce knowledge graph, determine the delivery address corresponding to each product associated with the orderer;

[0018] For any product associated with the orderer, based on the distance between the delivery addresses corresponding to the product, determine the delivery addresses that have a nearest neighbor relationship among the delivery addresses corresponding to the product;

[0019] Based on the delivery addresses that are adjacent to each product associated with the orderer, the characteristics of the adjacent addresses are determined. The characteristics of the adjacent addresses are used to indicate whether the orderer has abnormal transaction behavior of using scattered transaction addresses for consolidation.

[0020] In one possible implementation, the risk characteristics of the purchaser at the product level include the risk characteristics of the purchaser's unreasonable self-use. Based on the cross-border e-commerce knowledge graph, the purchaser's risk characteristics are determined, including:

[0021] Based on the relationship attributes between the orderer and the goods in the cross-border e-commerce knowledge graph, the quantity, type and frequency of goods purchased by the orderer in the third time window are determined.

[0022] Based on the quantity, type, and frequency of goods purchased by the orderer, the elastic network regression algorithm is used to determine the risk characteristics of the orderer's unreasonable self-use. The risk characteristics of the orderer's unreasonable self-use are used to indicate whether the orderer has any abnormal transaction behavior that does not belong to reasonable self-use.

[0023] In one possible implementation, based on cross-border e-commerce knowledge graphs and orderer risk characteristics, the risk assessment results of the target e-commerce enterprise are determined, including:

[0024] Based on the relationship attributes between target e-commerce enterprises and customers, and between customers and products in the cross-border e-commerce knowledge graph, an association network model is determined. The association network model is used to indicate the relationship between enterprises, customers, and products.

[0025] Based on the association network model and the risk characteristics of the orderer, the association strength of the target e-commerce enterprise in terms of transaction amount, delivery address and product dimensions is determined.

[0026] Based on the correlation strength of the target e-commerce enterprise in various dimensions, the risk assessment results of the target e-commerce enterprise are determined.

[0027] In one possible implementation, the risk assessment results for the target e-commerce enterprise are determined based on the correlation strength across various dimensions, including:

[0028] When the maximum correlation strength of the target e-commerce enterprise in each dimension is greater than or equal to the preset correlation strength threshold, the maximum correlation strength is multiplied by the first weight, and the correlation strength of other dimensions is multiplied by the second weight respectively. The results are summed to determine the risk assessment result of the target e-commerce enterprise, with the first weight being greater than the second weight.

[0029] When the maximum correlation strength of the target e-commerce enterprise in each dimension is less than the preset correlation strength threshold, the correlation strength of each dimension is multiplied by the second weight and then summed to determine the risk assessment result of the target e-commerce enterprise.

[0030] According to another aspect of this disclosure, a risk detection device for cross-border e-commerce enterprises is provided. The device includes:

[0031] The first determination module is used to extract entity information and relationship information between entities from the cross-border e-commerce list data, and to establish a cross-border e-commerce knowledge graph. The nodes in the cross-border e-commerce knowledge graph are used to indicate the corresponding entities, and the edges in the cross-border e-commerce knowledge graph are used to indicate the relationship between the entities corresponding to the two nodes of the connecting edge. Entities include orderers, e-commerce companies, delivery addresses, and goods.

[0032] The second determination module is used to determine the risk characteristics of the purchaser based on the cross-border e-commerce knowledge graph. The risk characteristics of the purchaser are used to characterize the risk behavior of the purchaser in multiple dimensions during the cross-border e-commerce transaction process. These multiple dimensions include the transaction amount dimension, the delivery address dimension, and the product dimension.

[0033] The third determination module is used to determine the risk assessment results of the target e-commerce enterprise based on the cross-border e-commerce knowledge graph and the risk characteristics of the orderer. The risk assessment results are used to indicate whether the target e-commerce enterprise has the risk of consolidation and resale.

[0034] In one possible implementation, the second determining module is used for:

[0035] Based on the relationship attributes between the orderer and the goods in the cross-border e-commerce knowledge graph, the maximum dutiable value of the goods purchased by the orderer in each first time window is determined.

[0036] The maximum dutiable value of goods purchased in each first time window is compared with the preset purchase limit to determine the orderer's risk characteristics in terms of transaction amount. The orderer's risk characteristics in terms of transaction amount are used to indicate whether the orderer has engaged in abnormal transaction behavior that exceeds the purchase limit.

[0037] In one possible implementation, the orderer risk characteristics of the delivery address dimension include abnormal distribution characteristics of delivery addresses, and a second determination module is used for:

[0038] Based on the relationship attributes between orderers and delivery addresses in the cross-border e-commerce knowledge graph, the number of delivery addresses that the orderer used only once in the second time window and the total number of delivery addresses used by the orderer in the second time window are determined.

[0039] Based on the number of delivery addresses used only once and the total number of delivery addresses used, abnormal distribution characteristics of delivery addresses are determined. These abnormal distribution characteristics are used to indicate whether the orderer has engaged in abnormal transaction behavior by frequently changing delivery addresses.

[0040] In one possible implementation, the orderer risk characteristics of the delivery address dimension include nearest neighbor address characteristics, and a second determination module is used for:

[0041] Based on the relationship attributes between the orderer and the product, and between the orderer and the delivery address in the cross-border e-commerce knowledge graph, determine the delivery address corresponding to each product associated with the orderer;

[0042] For any product associated with the orderer, based on the distance between the delivery addresses corresponding to the product, determine the delivery addresses that have a nearest neighbor relationship among the delivery addresses corresponding to the product;

[0043] Based on the delivery addresses that are adjacent to each product associated with the orderer, the characteristics of the adjacent addresses are determined. The characteristics of the adjacent addresses are used to indicate whether the orderer has abnormal transaction behavior of using scattered transaction addresses for consolidation.

[0044] In one possible implementation, the orderer risk characteristics at the product dimension include the orderer's risk characteristics of unreasonable self-use, and the second determining module is used for:

[0045] Based on the relationship attributes between the orderer and the goods in the cross-border e-commerce knowledge graph, the quantity, type and frequency of goods purchased by the orderer in the third time window are determined.

[0046] Based on the quantity, type, and frequency of goods purchased by the orderer, the elastic network regression algorithm is used to determine the risk characteristics of the orderer's unreasonable self-use. The risk characteristics of the orderer's unreasonable self-use are used to indicate whether the orderer has any abnormal transaction behavior that does not belong to reasonable self-use.

[0047] In one possible implementation, the third determining module is used for:

[0048] Based on the relationship attributes between target e-commerce enterprises and customers, and between customers and products in the cross-border e-commerce knowledge graph, an association network model is determined. The association network model is used to indicate the relationship between enterprises, customers, and products.

[0049] Based on the association network model and the risk characteristics of the orderer, the association strength of the target e-commerce enterprise in terms of transaction amount, delivery address and product dimensions is determined.

[0050] Based on the correlation strength of the target e-commerce enterprise in various dimensions, the risk assessment results of the target e-commerce enterprise are determined.

[0051] In one possible implementation, the risk assessment results for the target e-commerce enterprise are determined based on the correlation strength across various dimensions, including:

[0052] When the maximum correlation strength of the target e-commerce enterprise in each dimension is greater than or equal to the preset correlation strength threshold, the maximum correlation strength is multiplied by the first weight, and the correlation strength of other dimensions is multiplied by the second weight respectively. The results are summed to determine the risk assessment result of the target e-commerce enterprise, with the first weight being greater than the second weight.

[0053] When the maximum correlation strength of the target e-commerce enterprise in each dimension is less than the preset correlation strength threshold, the correlation strength of each dimension is multiplied by the second weight and then summed to determine the risk assessment result of the target e-commerce enterprise.

[0054] According to another aspect of this disclosure, a risk detection device for cross-border e-commerce enterprises is provided, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-described method.

[0055] According to another aspect of this disclosure, a non-volatile computer-readable storage medium is provided, on which a computer program is stored, which, when executed by a processor, implements the steps of the above-described method.

[0056] According to another aspect of this disclosure, a computer program product is provided, including a computer program or a non-volatile computer-readable storage medium carrying the computer program, wherein the computer program, when executed by a processor, implements the steps of the above-described method.

[0057] According to embodiments of this disclosure, a cross-border e-commerce knowledge graph is established by extracting entity information and inter-entity relationship information from cross-border e-commerce manifest data. Based on this knowledge graph, the risk characteristics of the purchaser are determined. These risk characteristics characterize the purchaser's risk behavior in multiple dimensions during the cross-border e-commerce transaction process, including transaction amount, delivery address, and product. Based on the cross-border e-commerce knowledge graph and the purchaser's risk characteristics, the risk assessment results of the target e-commerce enterprise are determined, enabling timely and accurate identification of whether the target e-commerce enterprise faces the risk of consolidation and resale. This disclosure embodiment, combined with the characteristics of cross-border e-commerce business, deeply explores the risk characteristics and inherent relationships of various trading entities. Utilizing the cross-border e-commerce knowledge graph, it achieves a comprehensive, accurate, and adaptive risk detection method. By deeply analyzing the purchaser's behavioral patterns, potential risks of e-commerce enterprises can be indirectly revealed. Based on the cross-border e-commerce trade flow knowledge graph, it not only focuses on specific transaction behaviors but also, through monitoring and analysis of purchaser activities, more accurately identifies potential consolidation and resale risks, thereby providing strong support for risk management.

[0058] Other features and aspects of this disclosure will become clear from the following detailed description of exemplary embodiments with reference to the accompanying drawings. Attached Figure Description

[0059] The accompanying drawings, which are included in and form part of this specification, illustrate exemplary embodiments, features, and aspects of this disclosure together with the specification and serve to explain the principles of this disclosure.

[0060] Figure 1 A flowchart illustrating a risk detection method for cross-border e-commerce enterprises according to an embodiment of this disclosure is provided.

[0061] Figure 2 A schematic diagram of a cross-border e-commerce knowledge graph according to an embodiment of the present disclosure is shown.

[0062] Figure 3 This diagram illustrates a process for risk detection of cross-border e-commerce enterprises according to an embodiment of the present disclosure.

[0063] Figure 4 A structural diagram of a risk detection device for cross-border e-commerce enterprises according to an embodiment of this disclosure is shown.

[0064] Figure 5 This is a block diagram illustrating a risk detection device 1900 for cross-border e-commerce enterprises according to an exemplary embodiment. Detailed Implementation

[0065] Various exemplary embodiments, features, and aspects of this disclosure will now be described in detail with reference to the accompanying drawings. The same reference numerals in the drawings denote elements that have the same or similar functions. Although various aspects of the embodiments are shown in the drawings, they are not necessarily drawn to scale unless specifically indicated otherwise.

[0066] As used herein, the terms “comprising,” “including,” “having,” or variations thereof are open-ended and include one or more of the stated features, integrals, elements, steps, components, or functions, but do not exclude the presence or addition of one or more other features, integrals, elements, steps, components, functions, or groups thereof.

[0067] When an element is referred to as “connected,” “coupled,” “responding,” or a variation thereof relative to another element, it may be directly connected, coupled, or responding to another element, or there may be an intermediate element present.

[0068] Although the terms first, second, third, etc., may be used herein to describe various elements / operations, these elements / operations should not be limited by these terms. These terms are only used to distinguish one element / operation from another. Therefore, without departing from the teachings of the inventive concept, a first element / operation in some embodiments may be referred to as a second element / operation in other embodiments.

[0069] The term “exemplary” as used herein means “serving as an example, embodiment, or illustration.” Any embodiment illustrated herein as “exemplary” is not necessarily to be construed as superior to or better than other embodiments.

[0070] Furthermore, to better illustrate this disclosure, numerous specific details are set forth in the following detailed description. Those skilled in the art will understand that this disclosure can be practiced without certain specific details. In some instances, methods, means, components, and circuits well known to those skilled in the art have not been described in detail in order to highlight the main points of this disclosure.

[0071] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, data stored, data displayed, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant regions.

[0072] Cross-border e-commerce, a new type of goods transaction born in the information age, has greatly expanded people's shopping scope and provided convenience for purchasing foreign goods. In recent years, with the deepening of economic globalization and the continuous advancement of modern information technology, cross-border e-commerce has become one of the important forces driving the improvement of the global economy. Cross-border e-commerce retail import refers to a consumption model in which domestic consumers purchase goods from overseas through third-party cross-border e-commerce platforms and have the goods shipped into the country through methods such as "bonded import" or "direct purchase import." Compared with traditional trade methods, this model can reduce time, manpower, and resource costs, simplify trade processes, reduce intermediate costs, and provide consumers with more price-competitive product choices.

[0073] Meanwhile, the sale of goods imported through cross-border e-commerce retail is subject to strict regulation, and the resale of such goods is strictly prohibited. Specifically, these goods are limited to personal use and cannot be resold. Cross-border e-commerce resale refers to the practice where goods imported from overseas are sold through cross-border e-commerce, and then the company or individual resells these goods to other consumers through markups, distribution, and transshipment. To ensure the healthy development of the market, consumers or those who purchase cross-border e-commerce retail imported goods are prohibited from reselling them.

[0074] Currently, machine learning algorithms such as clustering, classification, and association rule mining are commonly used in the field of risk detection to identify potential risks. However, the risk scenarios faced by cross-border e-commerce companies in the consolidation and resale process are complex and ever-changing. Current risk detection methods that rely solely on single data points often struggle to identify abnormal transaction behaviors in consolidation and resale risk identification scenarios in a timely and accurate manner.

[0075] In view of this, this disclosure provides a risk detection method, apparatus, and storage medium for cross-border e-commerce enterprises. The method of this disclosure establishes a cross-border e-commerce knowledge graph by extracting entity information and inter-entity relationship information from cross-border e-commerce manifest data. Based on the cross-border e-commerce knowledge graph, it determines the risk characteristics of the purchaser. These risk characteristics characterize the purchaser's risk behavior in the cross-border e-commerce transaction process across multiple dimensions, including transaction amount, delivery address, and product. Based on the cross-border e-commerce knowledge graph and the purchaser's risk characteristics, it determines the risk assessment result of the target e-commerce enterprise, enabling timely and accurate identification of whether the target e-commerce enterprise has the risk of reselling goods. This disclosure can combine the characteristics of cross-border e-commerce business to deeply explore the risk characteristics and inherent relationships of various trading entities. Utilizing the cross-border e-commerce knowledge graph, it achieves a comprehensive, accurate, and adaptive risk detection method. Through in-depth analysis of the purchaser's behavioral patterns, it can indirectly reveal the potential risks of e-commerce enterprises. Based on the cross-border e-commerce trade flow knowledge graph, it can not only focus on specific transaction behaviors but also more accurately identify potential reselling risks through monitoring and analysis of purchaser activities, thereby providing strong support for risk management.

[0076] In cross-border e-commerce trade, e-commerce companies often lack obvious behavioral characteristics, making it challenging to directly assess their consolidation and resale risks. Therefore, this disclosure adopts a "people-centric" research approach, focusing the analysis on the group of customers most closely associated with e-commerce companies. By systematically analyzing the transaction behavior patterns, consumption characteristics, and abnormal behaviors of customers, and combining this with a cross-border e-commerce trade flow knowledge graph, the potential consolidation and resale risks of e-commerce companies can be indirectly but effectively identified and assessed. This approach overcomes the limitations of directly observing corporate behavior and allows for the inference of corporate operating characteristics from end-consumer behavioral data, providing a new perspective for risk identification.

[0077] The risk detection method for cross-border e-commerce enterprises disclosed in this embodiment can be deployed on terminal devices or servers. Terminal devices can be any one or more of the following: mobile phones, foldable electronic devices, tablet computers, desktop computers, laptop computers, handheld computers, laptops, ultra-mobile personal computers (UMPCs), netbooks, cellular phones, personal digital assistants (PDAs), and in-vehicle devices, possessing wired or wireless communication capabilities. This disclosure does not impose any special limitations on the specific type of terminal device.

[0078] Servers can be located locally or in the cloud, and can be physical devices or virtual devices such as virtual machines and containers. They have wireless communication capabilities, which can be configured in the server's chip (system) or other components. Wireless communication capabilities can be implemented through mobile communication technologies such as 2G / 3G / 4G / 5G, as well as Wi-Fi, Bluetooth, frequency modulation (FM), data radio, and satellite communication; they can also communicate via wired connections to interact with other devices.

[0079] Figure 1 A flowchart illustrating a risk detection method for cross-border e-commerce enterprises according to an embodiment of this disclosure is provided. Figure 1 As shown, the method includes:

[0080] Step S101: Extract entity information and inter-entity relationship information from the cross-border e-commerce manifest data to establish a cross-border e-commerce knowledge graph.

[0081] The cross-border e-commerce list data can be data from the most recent preset period (such as within the last year), and can be obtained from sources such as historical cross-border e-commerce transaction records, platform order data, and customs declaration data.

[0082] Entity information can be used to indicate entity type and entity attributes. Entities can include orderers, e-commerce companies, delivery addresses, and goods. Among them, e-commerce companies are the core entities and main risk bearers in cross-border e-commerce trade. E-commerce company entities can use enterprise codes as unique identifiers (i.e., unique codes), and their entity attributes can include company name, place of registration, etc.

[0083] The orderer is the initiator of the transaction. To highlight the uniqueness of the orderer's identity and behavioral characteristics, the orderer's ID number can be used as a unique identifier. Its entity attributes may include the orderer's name, mobile phone number, etc.

[0084] The delivery address is a crucial part of the risk detection process for cross-border e-commerce companies, as its abnormal characteristics can increase the company's risk. To effectively identify abnormal situations such as cash on delivery, the entity type of delivery address can use standardized delivery address text as a unique identifier, and its entity attributes can include the latitude and longitude of the delivery address.

[0085] Entity relationship information can be used to indicate the type and attributes of relationships between entities. Relationship types can include those between an e-commerce company and a customer, between a customer and a delivery address, between a customer and a product, and between a delivery address and a product. The relationship attributes for each type of relationship are shown in Table 1 below.

[0086] Table 1

[0087]

[0088] As shown in Table 1, the trade frequency refers to the number of transactions that occur between different entities within a preset statistical period, the commodity price refers to the dutiable value of the commodity in the cross-border e-commerce list, the commodity quantity is used to indicate the quantity purchased by the orderer in a single or multiple transactions, and the number of categories indicates the number of types of the same kind of goods (such as goods under the same tariff code).

[0089] Existing graph construction technologies (such as named entity recognition, relation extraction, and pattern construction) can be used to extract entity information and inter-entity relationship information from cross-border e-commerce manifest data. This information can then be organized into a graph structure to obtain a cross-border e-commerce knowledge graph. In this graph, nodes indicate their corresponding entities, and edges indicate the relationship between the entities corresponding to the two nodes connecting those edges.

[0090] This embodiment of the disclosure is based on the core participants and their relationships in the entire cross-border e-commerce trade process. Through a systematic entity relationship modeling method, a multi-dimensional knowledge graph is constructed to meet the risk assessment needs of enterprises. Specifically, it focuses on mining the complex network relationships among trading entities such as e-commerce enterprises, orderers, consignees, and goods. A layered design concept is adopted to achieve refined definitions of entity attributes and relationship attributes, ultimately forming an intelligent cross-border e-commerce knowledge graph that supports risk assessment.

[0091] Figure 2 This diagram illustrates a cross-border e-commerce knowledge graph according to an embodiment of the present disclosure. Figure 2 The cross-border e-commerce knowledge graph shown can be stored in a graph database (such as Neo4j) to support subsequent analysis and computation. Based on the characteristics of cross-border e-commerce business, the knowledge graph can more comprehensively identify e-commerce companies that may face risks of reselling goods during cross-border e-commerce transactions, from dimensions such as the orderer, delivery address, and goods.

[0092] Step S102: Based on the cross-border e-commerce knowledge graph, determine the risk characteristics of the purchaser.

[0093] Among them, the risk characteristics of the orderer can be used to characterize the risk behavior of the orderer in multiple dimensions during the cross-border e-commerce transaction process. These multiple dimensions may include the transaction amount dimension, the delivery address dimension, and the product dimension.

[0094] This disclosed embodiment introduces the concept of "observing enterprises from the perspective of people," which detects the risk of e-commerce enterprises' centralized resale of goods from the perspective of orderers' behavior. It incorporates dimensions such as delivery address, goods, and transaction amount into the risk detection, which can discover anomalies from more dimensions.

[0095] The risk characteristics of orderers based on transaction amount can be used to indicate whether an orderer has engaged in abnormal transactions exceeding purchase limits. For example, from the perspective of transaction amount, it is possible to determine whether abnormal purchasing behavior exists by monitoring whether the amount of goods purchased by an orderer within a short period of time reaches or exceeds the annual purchase limit.

[0096] Order risk characteristics at the delivery address level can include abnormal delivery address distribution characteristics, which can indicate whether an orderer is engaging in abnormal transaction behavior such as frequently changing delivery addresses. Order risk characteristics at the delivery address level can also include neighboring address characteristics, which can indicate whether an orderer is engaging in abnormal transaction behavior such as using multiple transaction addresses for consolidation. For example, from the delivery address level, natural language processing can be performed on delivery addresses periodically to confirm whether an orderer is engaging in risky behavior such as using multiple addresses for one person and using multiple neighboring addresses. "Multiple addresses for one person" mainly refers to a situation where an orderer has multiple addresses, with a significant proportion of addresses used only once. This situation is likely due to the orderer frequently changing delivery addresses to evade platform monitoring, making it difficult for the platform to accurately identify their true consolidation behavior. "Nearby addresses" can be analyzed by examining the distance between delivery addresses of different orderers purchasing the same item, thereby identifying whether different orderers in geographically close locations are purchasing the same item and preventing the phenomenon of using multiple addresses to conceal consolidation behavior.

[0097] The purpose of an individual purchasing cross-border e-commerce goods should be for personal use, not for resale or business purposes. Purchasing goods for resale, commercial operation, or similar purposes does not meet the definition of personal use. The risk characteristics of the purchaser at the product level can include characteristics indicating unreasonable personal use, which can be used to indicate whether the purchaser has engaged in abnormal transaction behavior that does not fall under reasonable personal use. From the product perspective, a reasonableness analysis of personal use can be conducted. For example, reasonable quantities for personal use can be set for different types of goods based on different attributes of the transaction, and the quantity of such goods purchased by the purchaser within a certain period can be detected to see if it exceeds the reasonable personal use data.

[0098] The following will provide a detailed explanation of how the risk characteristics of the orderer in each dimension are determined.

[0099] Regarding the transaction amount dimension, in one possible implementation, step S102 includes:

[0100] Based on the relationship attributes between orderers and goods in the cross-border e-commerce knowledge graph, the maximum dutiable value of goods purchased by orderers in each first time window is determined; the maximum dutiable value of goods purchased in each first time window is compared with the preset purchase amount to determine the orderer's risk characteristics in the dimension of transaction amount.

[0101] The size of the first time window can be preset as needed (e.g., 30 days). Within the time span of the aforementioned cross-border e-commerce manifest data (e.g., within one year), the relationship attributes between the orderer and the goods (e.g., product price, quantity, etc.) are statistically analyzed according to the first time window. For each first time window, the relationship attributes between the orderer and the goods (e.g., product price, quantity, etc.) are statistically analyzed to establish a transaction amount profile for the orderer, including the maximum dutiable value within each first time window.

[0102] Specifically, for any given first time window, the dutiable value of each transaction can be calculated using the list number and product item as the granularity (e.g., calculated based on the unit price of the product and the quantity purchased), and the highest value is selected from all transaction records as the maximum dutiable value for that first time window.

[0103] The preset purchase limit can be set in advance, such as the annual purchase limit stipulated in relevant regulations.

[0104] If the maximum dutiable value of goods purchased by an orderer in any first time window exceeds the preset purchase limit, it can be considered that the orderer has engaged in abnormal transaction behavior exceeding the purchase limit, and the orderer's risk characteristic in the transaction amount dimension can be set to the first preset value (e.g., 1); otherwise, if the maximum dutiable value of goods purchased by all orderers in the first time window does not exceed the preset purchase limit, it can be considered that the orderer has not engaged in abnormal transaction behavior exceeding the purchase limit, and the orderer's risk characteristic in the transaction amount dimension can be set to the second preset value (e.g., 0).

[0105] Regarding the delivery address dimension, in one possible implementation, step S102 includes:

[0106] Based on the relationship attributes between orderers and delivery addresses in the cross-border e-commerce knowledge graph, the number of delivery addresses that the orderer used only once within the second time window and the total number of delivery addresses used by the orderer within the second time window are determined; based on the number of delivery addresses used only once and the total number of delivery addresses used, the abnormal distribution characteristics of delivery addresses are determined.

[0107] The size of the second time window can be preset as needed (e.g., 30 days). Within the time span of the aforementioned cross-border e-commerce list data (e.g., within one year), the relationship attributes between the orderer and the delivery address can be statistically analyzed according to the second time window. This will determine the number of delivery addresses used by the orderer within the second time window, the trade frequency of each delivery address, and establish an orderer address profile. This profile will include the number of delivery addresses used only once by the orderer within each second time window, as well as the total number of delivery addresses used by the orderer within each second time window.

[0108] Specifically, for any second time window, the ratio of the number of delivery addresses used only once by the orderer within that second time window to the total number of delivery addresses used can be calculated. If the ratio for any second time window is greater than a preset ratio threshold (e.g., 0.5), and the total number of delivery addresses used by the orderer within any second time window is greater than a preset quantity threshold (determined based on the distribution characteristics of the total number of delivery addresses for all orderers), it can be considered that the orderer has engaged in abnormal transaction behavior of frequently changing delivery addresses, and the abnormal distribution characteristic of delivery addresses can be set to a first preset value (e.g., 1). Otherwise, it can be considered that the orderer does not have abnormal transaction behavior of frequently changing delivery addresses, and the abnormal distribution characteristic of delivery addresses can be set to a second preset value (e.g., 0).

[0109] This allows us to detect whether the orderer's delivery address distribution is abnormal. If a high percentage of the addresses that the orderer has used only once in the past are used, there may be a risk.

[0110] Regarding the delivery address dimension, in one possible implementation, step S102 includes:

[0111] Based on the relationship attributes between orderers and products, and between orderers and delivery addresses in the cross-border e-commerce knowledge graph, the delivery addresses corresponding to each product associated with the orderer are determined; for any product associated with the orderer, based on the distance between the delivery addresses corresponding to the product, the delivery addresses with nearest neighbor relationships among the delivery addresses corresponding to the product are determined; based on the delivery addresses with nearest neighbor relationships among the delivery addresses corresponding to each product associated with the orderer, the characteristics of nearest neighbor addresses are determined.

[0112] Specifically, based on the relationship attributes between orderers and products, and between orderers and delivery addresses in the cross-border e-commerce knowledge graph, multiple delivery addresses for the same product by different orderers can be obtained. The latitude and longitude attributes of these delivery addresses can be stored as vectors in a vector database. For example, for a certain product, there are orderers 1 and 2. Orderer 1 has delivery addresses 1 and 2 for this product, and orderer 2 has delivery address 3 for this product. The latitude and longitude attributes of delivery addresses 1-3 can be stored as vectors in the vector database.

[0113] Next, based on the distance between the vectors of the delivery addresses, it can be determined whether the distance between different delivery addresses is less than a preset distance threshold. If the vector distance between two addresses is less than this threshold, they are considered to have a nearest neighbor relationship. For example, based on the distance between the vectors corresponding to delivery addresses 1 and 2 and delivery address 3, it can be determined whether each distance is less than the preset distance threshold. For instance, if the distance between delivery address 2 and delivery address 3 is less than the preset distance threshold, it is considered that there are two nearest neighbor addresses for this type of product (i.e., delivery addresses with a nearest neighbor relationship include delivery address 2 and delivery address 3).

[0114] The above method can be used to calculate the nearest neighbor addresses corresponding to various types of goods. Further, the nearest neighbor addresses corresponding to each good associated with the orderer are obtained. For any orderer, the number of goods associated with that orderer whose corresponding nearest neighbor addresses exceed a preset threshold can be determined as the nearest neighbor address characteristics of that orderer.

[0115] Therefore, from the perspective of the orderer and the product, it is possible to identify whether different orderers with similar geographical locations may be purchasing the same product for the same purpose. That is, to detect whether the neighboring addresses corresponding to different orderers and the same product are neighboring, so as to help identify potential risks of consolidation and resale.

[0116] Regarding the product dimension, in one possible implementation, step S102 includes:

[0117] Based on the relationship attributes between customers and goods in the cross-border e-commerce knowledge graph, the quantity, type, and frequency of goods purchased by the customer within the third time window are determined. Based on the quantity, type, and frequency of goods purchased by the customer, the elastic network regression algorithm is used to determine the risk characteristics of unreasonable self-use by the customer.

[0118] The size of the third time window can be preset as needed (e.g., 30 days). Within the time span of the aforementioned cross-border e-commerce list data (e.g., within one year), the relationship attributes between the orderer and the goods can be statistically analyzed according to the third time window to construct features, including the quantity of goods purchased by the orderer, the type of goods (which can be obtained by statistically analyzing the names of goods under the same tax code), and the frequency of goods purchase (determined based on the trade frequency attribute) within each third time window.

[0119] Regarding the quantity of goods, the quantity of goods purchased for personal use should meet individual needs and normal standards. If the quantity purchased is abnormal, such as purchasing a large quantity of the same product, it may be considered unreasonable. Regarding the type of goods, if the type of goods purchased by the orderer is abnormal, such as purchasing a large quantity of multiple similar products, it may be considered unreasonable. Regarding the frequency of purchase, if the orderer purchases goods too frequently or purchases the same product for a long period of time, it may be considered unreasonable.

[0120] Since multicollinearity may exist among the above statistical features in cross-border e-commerce scenarios, this disclosure adopts the elastic network regression algorithm for modeling. The elastic network autoregression algorithm has great advantages in handling multicollinearity. This algorithm combines the characteristics of ridge regression, making it more robust in handling multicollinearity features. Furthermore, due to the inclusion of the characteristics of Lasso regression, this algorithm can perform feature selection, thereby simplifying the model and improving interpretability. In addition, by adjusting the hyperparameters, this algorithm can flexibly balance the complexity and generalization ability of the model.

[0121] The quantity, type, and frequency of purchases of goods by the customer within each third time window can be used as input features for the elastic network model. These input features can also include the value of the goods (e.g., excessively high value may be considered unreasonable) and the number of purchasing platforms (e.g., purchasing from multiple platforms may be for diversification and could be considered unreasonable) as auxiliary features. When a customer exhibits abnormal transaction behavior that is not for reasonable self-use, the elastic network model can output a first preset value (e.g., 1) as the risk feature of unreasonable self-use; otherwise, the elastic network model can output a second preset value (e.g., 0) as the risk feature of unreasonable self-use.

[0122] The elastic network model can be a pre-trained model using a training set. The training set can be constructed using the aforementioned input features corresponding to each orderer in historical cross-border e-commerce manifest sample data, along with training labels (used to indicate whether the orderer has engaged in abnormal transactions not for reasonable personal use), to train the initial elastic network model, resulting in a pre-trained elastic network model.

[0123] For a specific customer B, examples of customer risk characteristics across various dimensions could be:

[0124] "Purchaser B":

[0125] {"Number of purchase records for goods with proximity address risk": 2,}

[0126] "List of high-risk products purchased from neighboring addresses": ["Vitamin D", "Vitamin A"]

[0127] "Whether the limit for a natural year was temporarily exceeded": 1.

[0128] "Total number of shipping addresses used": 1323

[0129] Number of addresses used only once: 1215

[0130] Percentage of addresses used only once: 0.92

[0131] "Whether the purchased goods exceed reasonable personal use": 1.

[0132] "List of items exceeding reasonable personal use": ["Vitamin D"]

[0133] }

[0134] Step S103: Based on the cross-border e-commerce knowledge graph and the risk characteristics of the orderer, determine the risk assessment results of the target e-commerce enterprise.

[0135] The target e-commerce companies are the primary subjects of the consolidation and resale risk assessment. Based on the risk characteristics of the orderers obtained from the various dimensions described above, it is also necessary to assess the risk situation at the e-commerce company level. The risk assessment results can be used to indicate whether the target e-commerce company faces consolidation and resale risks.

[0136] According to embodiments of this disclosure, a cross-border e-commerce knowledge graph is established by extracting entity information and inter-entity relationship information from cross-border e-commerce manifest data. Based on this knowledge graph, the risk characteristics of the purchaser are determined. These risk characteristics characterize the purchaser's risk behavior in multiple dimensions during the cross-border e-commerce transaction process, including transaction amount, delivery address, and product. Based on the cross-border e-commerce knowledge graph and the purchaser's risk characteristics, the risk assessment results of the target e-commerce enterprise are determined, enabling timely and accurate identification of whether the target e-commerce enterprise faces the risk of consolidation and resale. This disclosure embodiment, combined with the characteristics of cross-border e-commerce business, deeply explores the risk characteristics and inherent relationships of various trading entities. Utilizing the cross-border e-commerce knowledge graph, it achieves a comprehensive, accurate, and adaptive risk detection method. By deeply analyzing the purchaser's behavioral patterns, potential risks of e-commerce enterprises can be indirectly revealed. Based on the cross-border e-commerce trade flow knowledge graph, it not only focuses on specific transaction behaviors but also, through monitoring and analysis of purchaser activities, more accurately identifies potential consolidation and resale risks, thereby providing strong support for risk management.

[0137] In one possible implementation, step S103 includes:

[0138] Based on the relationship attributes between target e-commerce enterprises and customers, and between customers and products in the cross-border e-commerce knowledge graph, a correlation network model is determined; based on the correlation network model and customer risk characteristics, the correlation strength of target e-commerce enterprises in the dimensions of transaction amount, delivery address, and product is determined; based on the correlation strength of target e-commerce enterprises in each dimension, the risk assessment results of target e-commerce enterprises are determined.

[0139] This involves constructing a relational network model based on the relationship attributes between target e-commerce enterprises and customers, and between customers and products, within the cross-border e-commerce knowledge graph, using social network analysis methods. This relational network model can indicate the relationships between e-commerce enterprises, customers, and products. These relationships can include those between e-commerce enterprises and customers, customers and products, and e-commerce enterprises and products, reflecting whether there are direct or indirect transaction connections between the entities. Products in the relational network model can be dual-identified using both tax identification numbers and product names to enhance identification accuracy.

[0140] Building upon this, at the delivery address dimension, the nearest neighbor address feature from the orderer's risk characteristics can be used, combined with the association relationship between e-commerce companies and products in the association network model, to calculate the association strength corresponding to the nearest neighbor address dimension in the delivery address dimension. Alternatively, the abnormal distribution feature of delivery addresses from the orderer's risk characteristics can be used, combined with the association relationship between e-commerce companies and orderers in the association network model, to calculate the association strength corresponding to the abnormal distribution dimension of delivery addresses (which can be termed the "one person, multiple addresses" dimension). At the transaction amount dimension, the orderer's risk characteristics at the transaction amount dimension can be used, combined with the association relationship between e-commerce companies and orderers in the association network model, to calculate the association strength at the transaction amount dimension. At the product dimension, the risk characteristics of the orderer's unreasonable self-use can be used, combined with the association relationship between e-commerce companies and orderers in the association network model, to calculate the association strength at the product dimension.

[0141] When calculating association strength, it can be obtained by quantifying centrality indicators (such as degree centrality and betweenness centrality) in the association network model. The calculation of centrality indicators can be implemented based on existing related technologies. For example, degree centrality can be represented as the proportion of high-risk customers or abnormal goods directly associated with a target e-commerce enterprise in a certain dimension (determined based on relevant customer risk characteristics). Association strength can be used to quantify the degree of association between a target e-commerce enterprise and high-risk customers, abnormal delivery addresses, or abnormal goods in a certain dimension, thereby reflecting the degree of aggregation risk of the target e-commerce enterprise in that dimension. For example, in the delivery address dimension, based on the nearest neighbor address characteristics in the customer risk characteristics, the number of abnormal goods related to the nearest neighbor address characteristics in each product node associated with the e-commerce enterprise in the association network model can be determined. In this case, for e-commerce company A associated with customers B and C, the number of products related to e-commerce company A in the nearest neighbor address features for customer B is 2, and the number of products related to e-commerce company A in the nearest neighbor address features for customer C is 3. Therefore, the number of abnormal products related to the nearest neighbor address features among all product nodes associated with e-commerce company A can be considered to be 2 + 3 = 5. Next, based on the proportion of abnormal products to the total number of products associated with the e-commerce company, the degree centrality index value can be determined as the association strength corresponding to the nearest neighbor address dimension in the delivery address dimension. Similar calculation methods are used to calculate the association strength for other dimensions.

[0142] In determining the risk assessment results of a target e-commerce enterprise based on its correlation strength across various dimensions, this disclosed embodiment addresses the pain point of "low-weighted sub-dimension risk dilution" in the traditional weighted summation model. It proposes a risk index calculation mechanism combining "threshold triggering" and "weighted summation." The threshold triggering mechanism ensures that high-risk sub-dimensions are not diluted; the weighted summation preserves the cumulative effect of multi-dimensional risks, avoiding misjudgments caused by a single threshold. The risk index calculation mechanism combining "threshold triggering" and "weighted summation" will be described below.

[0143] In one possible implementation, during the process of determining the risk assessment results of the target e-commerce enterprise based on the correlation strength of the target e-commerce enterprise in various dimensions, the following can be done:

[0144] When the maximum correlation strength of the target e-commerce enterprise in each dimension is greater than or equal to the preset correlation strength threshold, the maximum correlation strength is multiplied by the first weight, and the correlation strength of other dimensions is multiplied by the second weight respectively. The results are summed to determine the risk assessment result of the target e-commerce enterprise.

[0145] The association strength threshold can be preset as needed (e.g., 0.5), and the association strength thresholds for different dimensions can be the same or different. The first weight and the second weight can be preset as needed, and the first weight can be set to be greater than the second weight (e.g., the first weight can be set to 3 times the second weight).

[0146] When the maximum correlation strength of the target e-commerce enterprise in each dimension is less than the preset correlation strength threshold, the correlation strength of each dimension is multiplied by the second weight and then summed to determine the risk assessment result of the target e-commerce enterprise.

[0147] Risk assessment results can be quantified using a risk index. A higher risk index value indicates a higher probability that the target e-commerce company poses a risk of reselling goods through consolidation. For example, when the risk index exceeds a risk threshold (e.g., 0.5), the target e-commerce company can be marked as having a risk of reselling goods through consolidation. The risk index can be calculated as follows:

[0148] ;

[0149] in, Let be the association strength in the i-th dimension. Let w be the correlation strength of the j-th dimension, n be the total number of dimensions, w be the second weight mentioned above (e.g., 0.25), 3w be the first weight mentioned above, and T be the correlation strength threshold (e.g., 0.5).

[0150] The following is an example of how the risk index of a cross-border e-commerce company C, which is marked as having the risk of consolidation and resale, is calculated:

[0151] E-commerce company C:

[0152] {

[0153] "Number of high-risk goods at neighboring addresses": 12.

[0154] "List of high-risk products for neighboring addresses": ["Fish oil", "Vitamin D"...]

[0155] "Nearest neighbor address dimension association strength": 0.65

[0156] "Number of subscribers who temporarily exceeded the annual limit": 3

[0157] "Correlation strength in transaction amount dimension": 0.3

[0158] Number of subscribers with multiple addresses: 2.

[0159] "Association strength across multiple addresses per person": 0.25

[0160] "The quantity of goods exceeds the number of orders placed for reasonable personal use": 5.

[0161] "Product-level correlation strength": 0.4

[0162] "List of items exceeding reasonable personal use": ["Vitamin D",...]

[0163] Risk Index: 0.725

[0164] }

[0165] As shown in the example above, the correlation strength of the nearest address dimension is 0.65, which reflects the high frequency of receiving health products in the same logistics park; there are 3 customers who temporarily exceed the limit of the natural year, and the calculated correlation strength is 0.3; there are 2 customers with the risk of one person having multiple addresses, and the correlation strength is 0.25; there are 5 customers who purchase goods that exceed reasonable personal use, and the correlation strength is 0.4, which reflects the bulk purchase of health products.

[0166] After introducing the "threshold trigger" and "weighted summation" risk index calculation mechanisms, the risk index calculated according to the above formula is: Risk Index = min(1, 0.25×0.3 + 0.25×0.25 + 0.25×0.4 + 3×0.25×0.65)) = 0.725. In this case, e-commerce company C is marked as having the risk of reselling consolidated goods. However, if the traditional weighted method is used, the risk index = 0.25×0.3 + 0.25×0.25 + 0.25×0.4 + 0.25×0.65 = 0.4, and e-commerce company C is marked as not having the risk of reselling consolidated goods. It can be seen that using the traditional weighted method to calculate the risk index may indeed dilute the impact of high-risk sub-dimensions, thus underestimating the overall risk.

[0167] Figure 3 This diagram illustrates a process for risk detection of cross-border e-commerce enterprises according to embodiments of the present disclosure. Figure 3 As shown, the risk detection process for cross-border e-commerce enterprises in this embodiment mainly includes:

[0168] 1. Construct a knowledge graph based on cross-border e-commerce manifest data, as described in step S101 above.

[0169] 2. Risk analysis of the purchaser, please refer to step S102 above.

[0170] 3. Risk analysis for e-commerce enterprises, please refer to step S103 above.

[0171] This disclosure constructs an analytical framework of "orderer behavior - enterprise operation chain and consolidation and resale risk" to achieve the mapping and transformation of individual behavioral data into enterprise risk identification. The implementation path includes: behavioral data anchoring: focusing on the multi-dimensional behavior of orderers in cross-border transactions (including but not limited to transaction amount, address risk, and product dimensions), and extracting a set of abnormal behavior features; relationship reconstruction and risk quantification: using knowledge graph relationship mining technology, by analyzing the strength of the association between orderers and enterprises, the positive correlation between individual behavior and enterprise resale risk is mined, thereby quantifying the enterprise risk index. Based on the above research ideas, the method of this disclosure analyzes the orderer's order data to analyze the orderer's risk, thereby linking it to e-commerce enterprises and detecting the consolidation and resale risk behavior of e-commerce enterprises. It can indirectly but effectively identify and assess the potential consolidation and resale risks of e-commerce enterprises. This research approach overcomes the limitations of directly observing enterprise behavior and can infer the enterprise's operating characteristics through end-consumer behavioral data, providing a new perspective for risk identification.

[0172] Figure 4 A structural diagram of a risk detection device for cross-border e-commerce enterprises according to an embodiment of this disclosure is shown. Figure 4 As shown, the device may include:

[0173] The first determining module 401 is used to extract entity information and relationship information between entities from the cross-border e-commerce list data, and to establish a cross-border e-commerce knowledge graph. The nodes in the cross-border e-commerce knowledge graph are used to indicate the corresponding entities, and the edges in the cross-border e-commerce knowledge graph are used to indicate the relationship between the entities corresponding to the two nodes of the connecting edge. The entities include the orderer, e-commerce enterprise, delivery address and goods.

[0174] The second determination module 402 is used to determine the risk characteristics of the orderer based on the cross-border e-commerce knowledge graph. The risk characteristics of the orderer are used to characterize the risk behavior of the orderer in multiple dimensions during the cross-border e-commerce transaction process. The multiple dimensions include the transaction amount dimension, the delivery address dimension, and the product dimension.

[0175] The third determination module 403 is used to determine the risk assessment results of the target e-commerce enterprise based on the cross-border e-commerce knowledge graph and the risk characteristics of the orderer. The risk assessment results are used to indicate whether the target e-commerce enterprise has the risk of consolidation and resale.

[0176] In one possible implementation, the second determining module 402 is used for:

[0177] Based on the relationship attributes between the orderer and the goods in the cross-border e-commerce knowledge graph, the maximum dutiable value of the goods purchased by the orderer in each first time window is determined.

[0178] The maximum dutiable value of goods purchased in each first time window is compared with the preset purchase limit to determine the orderer's risk characteristics in terms of transaction amount. The orderer's risk characteristics in terms of transaction amount are used to indicate whether the orderer has engaged in abnormal transaction behavior that exceeds the purchase limit.

[0179] In one possible implementation, the orderer risk characteristics of the delivery address dimension include abnormal distribution characteristics of delivery addresses, and the second determining module 402 is used for:

[0180] Based on the relationship attributes between orderers and delivery addresses in the cross-border e-commerce knowledge graph, the number of delivery addresses that the orderer used only once in the second time window and the total number of delivery addresses used by the orderer in the second time window are determined.

[0181] Based on the number of delivery addresses used only once and the total number of delivery addresses used, abnormal distribution characteristics of delivery addresses are determined. These abnormal distribution characteristics are used to indicate whether the orderer has engaged in abnormal transaction behavior by frequently changing delivery addresses.

[0182] In one possible implementation, the orderer risk characteristics of the delivery address dimension include nearest neighbor address characteristics, and the second determination module 402 is used for:

[0183] Based on the relationship attributes between the orderer and the product, and between the orderer and the delivery address in the cross-border e-commerce knowledge graph, determine the delivery address corresponding to each product associated with the orderer;

[0184] For any product associated with the orderer, based on the distance between the delivery addresses corresponding to the product, determine the delivery addresses that have a nearest neighbor relationship among the delivery addresses corresponding to the product;

[0185] Based on the delivery addresses that are adjacent to each product associated with the orderer, the characteristics of the adjacent addresses are determined. The characteristics of the adjacent addresses are used to indicate whether the orderer has abnormal transaction behavior of using scattered transaction addresses for consolidation.

[0186] In one possible implementation, the orderer risk characteristics at the product dimension include the orderer's risk characteristics of unreasonable self-use, and the second determining module 402 is used for:

[0187] Based on the relationship attributes between the orderer and the goods in the cross-border e-commerce knowledge graph, the quantity, type and frequency of goods purchased by the orderer in the third time window are determined.

[0188] Based on the quantity, type, and frequency of goods purchased by the orderer, the elastic network regression algorithm is used to determine the risk characteristics of the orderer's unreasonable self-use. The risk characteristics of the orderer's unreasonable self-use are used to indicate whether the orderer has any abnormal transaction behavior that does not belong to reasonable self-use.

[0189] In one possible implementation, the third determining module 403 is used for:

[0190] Based on the relationship attributes between target e-commerce enterprises and customers, and between customers and products in the cross-border e-commerce knowledge graph, an association network model is determined. The association network model is used to indicate the relationship between enterprises, customers, and products.

[0191] Based on the association network model and the risk characteristics of the orderer, the association strength of the target e-commerce enterprise in terms of transaction amount, delivery address and product dimensions is determined.

[0192] Based on the correlation strength of the target e-commerce enterprise in various dimensions, the risk assessment results of the target e-commerce enterprise are determined.

[0193] In one possible implementation, the risk assessment results for the target e-commerce enterprise are determined based on the correlation strength across various dimensions, including:

[0194] When the maximum correlation strength of the target e-commerce enterprise in each dimension is greater than or equal to the preset correlation strength threshold, the maximum correlation strength is multiplied by the first weight, and the correlation strength of other dimensions is multiplied by the second weight respectively. The results are summed to determine the risk assessment result of the target e-commerce enterprise, with the first weight being greater than the second weight.

[0195] When the maximum correlation strength of the target e-commerce enterprise in each dimension is less than the preset correlation strength threshold, the correlation strength of each dimension is multiplied by the second weight and then summed to determine the risk assessment result of the target e-commerce enterprise.

[0196] According to embodiments of this disclosure, a cross-border e-commerce knowledge graph is established by extracting entity information and inter-entity relationship information from cross-border e-commerce manifest data. Based on this knowledge graph, the risk characteristics of the purchaser are determined. These risk characteristics characterize the purchaser's risk behavior in multiple dimensions during the cross-border e-commerce transaction process, including transaction amount, delivery address, and product. Based on the cross-border e-commerce knowledge graph and the purchaser's risk characteristics, the risk assessment results of the target e-commerce enterprise are determined, enabling timely and accurate identification of whether the target e-commerce enterprise faces the risk of consolidation and resale. This disclosure embodiment, combined with the characteristics of cross-border e-commerce business, deeply explores the risk characteristics and inherent relationships of various trading entities. Utilizing the cross-border e-commerce knowledge graph, it achieves a comprehensive, accurate, and adaptive risk detection method. By deeply analyzing the purchaser's behavioral patterns, potential risks of e-commerce enterprises can be indirectly revealed. Based on the cross-border e-commerce trade flow knowledge graph, it not only focuses on specific transaction behaviors but also, through monitoring and analysis of purchaser activities, more accurately identifies potential consolidation and resale risks, thereby providing strong support for risk management.

[0197] In some embodiments, the functions or modules of the apparatus provided in this disclosure can be used to perform the methods described in the above method embodiments. The specific implementation can be referred to the description of the above method embodiments, and for the sake of brevity, it will not be repeated here.

[0198] This disclosure also provides a risk detection device for cross-border e-commerce enterprises, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0199] This disclosure also provides a non-volatile computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0200] This disclosure also provides a computer program product, including a computer program or a non-volatile computer-readable storage medium carrying the computer program, wherein the computer program, when executed by a processor, implements the steps of the above method.

[0201] Figure 5 This is a block diagram illustrating a risk detection device 1900 for cross-border e-commerce enterprises according to an exemplary embodiment. For example, device 1900 can be provided as a server or terminal device. (Refer to...) Figure 5 The apparatus 1900 includes a processing component 1922, which further includes one or more processors, and memory resources represented by memory 1932 for storing instructions, such as application programs, that can be executed by the processing component 1922. The application programs stored in memory 1932 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 1922 is configured to execute instructions to perform the methods described above.

[0202] Device 1900 may also include a power supply component 1926 configured to perform power management of device 1900, a wired or wireless network interface 1950 configured to connect device 1900 to a network, and an input / output interface 1958 (I / O interface). Device 1900 can operate on an operating system, such as Windows Server, stored in memory 1932. TM macOS X TM Unix TM Linux TM FreeBSD TM Or similar.

[0203] In an exemplary embodiment, a non-volatile computer-readable storage medium is also provided, such as a memory 1932 including computer program instructions that can be executed by a processing component 1922 of the device 1900 to perform the above-described method.

[0204] Computer-readable storage media can be tangible devices capable of holding and storing programs / instructions used by instruction execution devices. Computer-readable storage media can be, for example—but not limited to—electrical storage devices, magnetic storage devices, optical storage devices, electromagnetic storage devices, semiconductor storage devices, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), static random access memory (SRAM), portable compact disc read-only memory (CD-ROM), digital multifunction disc (DVD), memory sticks, floppy disks, mechanical encoding devices, such as punch cards or recessed protrusions storing instructions thereon, and any suitable combination of the foregoing. The computer-readable storage media used herein are not to be construed as transient signals themselves, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., light pulses through fiber optic cables), or electrical signals transmitted through wires.

[0205] The computer program (or computer-readable program instructions) described herein can be downloaded from a computer-readable storage medium to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage medium in the respective computing / processing device.

[0206] The computer program (or computer program instructions) used to perform the operations of this disclosure may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, state setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may execute entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuitry, such as programmable logic circuitry, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), is personalized by utilizing state information from the computer-readable program instructions to implement various aspects of this disclosure.

[0207] Various aspects of this disclosure are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this disclosure. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0208] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0209] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0210] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0211] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or technical improvements to the embodiments in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A risk detection method for cross-border e-commerce enterprises, characterized in that, The method includes: Extract entity information and inter-entity relationship information from cross-border e-commerce manifest data, and establish a cross-border e-commerce knowledge graph. Nodes in the cross-border e-commerce knowledge graph are used to indicate the corresponding entities, and edges in the cross-border e-commerce knowledge graph are used to indicate the relationship between the entities corresponding to the two nodes connecting the edges. The entities include the orderer, e-commerce enterprise, delivery address, and goods. Based on the cross-border e-commerce knowledge graph, the risk characteristics of the orderer are determined. The risk characteristics of the orderer are used to characterize the risk behavior of the orderer in multiple dimensions during the cross-border e-commerce transaction process. The multiple dimensions include the transaction amount dimension, the delivery address dimension, and the product dimension. Based on the cross-border e-commerce knowledge graph and the risk characteristics of the orderer, the risk assessment results of the target e-commerce enterprise are determined. The risk assessment results are used to indicate whether the target e-commerce enterprise has the risk of consolidation and resale.

2. The method of claim 1, wherein, The process of determining the risk characteristics of the purchaser based on the cross-border e-commerce knowledge graph includes: Based on the relationship attributes between the orderer and the goods in the cross-border e-commerce knowledge graph, the maximum dutiable value of the goods purchased by the orderer in each first time window is determined. The maximum tax-paid price of the goods purchased in each first time window is compared with the preset purchase limit to determine the orderer's risk characteristics in the transaction amount dimension. The orderer's risk characteristics in the transaction amount dimension are used to indicate whether the orderer has engaged in abnormal transaction behavior that exceeds the purchase limit.

3. The method of claim 1, wherein, The orderer risk characteristics at the delivery address level include abnormal delivery address distribution characteristics. The determination of orderer risk characteristics based on the cross-border e-commerce knowledge graph includes: Based on the relationship attributes between the orderer and the delivery address in the cross-border e-commerce knowledge graph, the number of delivery addresses that the orderer used only once in the second time window and the total number of delivery addresses used by the orderer in the second time window are determined. Based on the number of delivery addresses that have only been used once and the total number of delivery addresses that have been used, the abnormal distribution characteristics of delivery addresses are determined. These abnormal distribution characteristics are used to indicate whether the orderer has engaged in abnormal transaction behavior by frequently changing delivery addresses.

4. The method of claim 1, wherein, The orderer risk characteristics at the delivery address level include nearest-neighbor address characteristics. The determination of orderer risk characteristics based on the cross-border e-commerce knowledge graph includes: Based on the relationship attributes between the orderer and the product, and between the orderer and the delivery address in the cross-border e-commerce knowledge graph, determine the delivery addresses corresponding to each product associated with the orderer. For any product associated with the orderer, based on the distance between the delivery addresses corresponding to the product, determine the delivery addresses that have a nearest neighbor relationship among the delivery addresses corresponding to the product; Based on the delivery addresses that are adjacent to each product associated with the orderer, the adjacent address features are determined. The adjacent address features are used to indicate whether the orderer has abnormal transaction behavior of using scattered transaction addresses for consolidation.

5. The method of claim 1, wherein, The risk characteristics of the purchaser at the product level include the risk characteristics of the purchaser's unreasonable personal use. The determination of the purchaser's risk characteristics based on the cross-border e-commerce knowledge graph includes: Based on the relationship attributes between the orderer and the product in the cross-border e-commerce knowledge graph, the quantity, type and frequency of the product purchased by the orderer within the third time window are determined. Based on the quantity, type, and frequency of goods purchased by the orderer, the elastic network regression algorithm is used to determine the risk characteristics of the orderer's unreasonable self-use. These risk characteristics are used to indicate whether the orderer has any abnormal transaction behavior that does not belong to reasonable self-use.

6. The method of claim 1, wherein, The risk assessment results for determining the target e-commerce enterprise based on the cross-border e-commerce knowledge graph and the risk characteristics of the orderer include: Based on the relationship attributes between target e-commerce enterprises and customers, and between customers and products in the cross-border e-commerce knowledge graph, an association network model is determined. The association network model is used to indicate the association relationship between enterprises, customers, and products. Based on the aforementioned network model and the risk characteristics of the orderer, the correlation strength of the target e-commerce enterprise in terms of transaction amount, delivery address, and product is determined. Based on the correlation strength of the target e-commerce enterprise in various dimensions, the risk assessment result of the target e-commerce enterprise is determined.

7. The method of claim 6, wherein, The determination of the risk assessment result of the target e-commerce enterprise based on the correlation strength of the target e-commerce enterprise in various dimensions includes: When the maximum correlation strength of the target e-commerce enterprise in each dimension is greater than or equal to the preset correlation strength threshold, the maximum correlation strength is multiplied by the first weight, and the correlation strength of other dimensions is multiplied by the second weight respectively. The results are summed to determine the risk assessment result of the target e-commerce enterprise, where the first weight is greater than the second weight. When the maximum correlation strength of the target e-commerce enterprise in each dimension is less than the preset correlation strength threshold, the correlation strength of each dimension is multiplied by the second weight and then summed to determine the risk assessment result of the target e-commerce enterprise.

8. A risk detection device for cross-border e-commerce enterprises, characterized in that, The device includes: The first determination module is used to extract entity information and relationship information between entities from the cross-border e-commerce list data, and to establish a cross-border e-commerce knowledge graph. The nodes in the cross-border e-commerce knowledge graph are used to indicate the corresponding entities, and the edges in the cross-border e-commerce knowledge graph are used to indicate the relationship between the entities corresponding to the two nodes connecting the edges. The entities include the orderer, e-commerce enterprise, delivery address and goods. The second determining module is used to determine the risk characteristics of the orderer based on the cross-border e-commerce knowledge graph. The risk characteristics of the orderer are used to characterize the risk behavior of the orderer in multiple dimensions during the cross-border e-commerce transaction process. The multiple dimensions include transaction amount dimension, delivery address dimension and product dimension. The third determining module is used to determine the risk assessment result of the target e-commerce enterprise based on the cross-border e-commerce knowledge graph and the risk characteristics of the orderer. The risk assessment result is used to indicate whether the target e-commerce enterprise has the risk of reselling goods. 9.A risk detection device for cross-border e-commerce enterprises, comprising a memory, a processor and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A non-transitory computer readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

11. A computer program product comprising a computer program, or a non-volatile computer-readable storage medium carrying a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images