Methods and devices for identifying fraudulent transactions
By constructing a data node set and performing undirected graph analysis, abnormal clusters are identified, solving the problem of identifying fraudulent order behavior and improving the reliability and accuracy of identification.
Patent Information
- Application Number
- CN202310953608.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-31
- Publication Date
- 2026-11-10
- Estimated Expiration
- 2043-07-31
AI Technical Summary
Existing technologies are insufficient to effectively identify and stop fraudulent order practices, which damage the reputation of shopping websites and infringe upon the rights of other merchants.
By constructing a set of data nodes, merging target data nodes, constructing an undirected graph, analyzing the number of orders in the undirected subgraph, identifying abnormal clusters, and determining the risk of fraudulent orders.
It improves the reliability and accuracy of identifying fraudulent orders, reduces the complexity of analysis, and clearly shows the relationships between orders.
Smart Images

Figure CN116977025B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of information recognition technology, and more specifically, to a method and apparatus for identifying fraudulent transactions. Background Technology
[0002] With the rapid development of e-commerce technology, online shopping has become increasingly convenient, making it the preferred method for consumers. When consumers browse products on shopping websites, the websites typically sort and filter products based on sales volume and positive review rates to help them choose suitable items from a large selection. Consequently, many merchants hire people or use shopping bots to artificially inflate sales figures and secure a favorable ranking for their products.
[0003] Such improper order-brushing behavior can damage the reputation of shopping websites and infringe on the rights of other merchants. Therefore, in order to maintain the safe and stable operation of shopping websites, it is necessary to identify and stop order-brushing behavior. Summary of the Invention
[0004] In view of this, this application provides a method and apparatus for identifying fraudulent order practices, in order to identify and stop such practices.
[0005] To achieve the above objectives, the following solution is proposed:
[0006] A method for identifying fraudulent orders includes:
[0007] A set of data nodes is determined, wherein the set of data nodes includes multiple first data nodes and multiple second data nodes, each first data node corresponds one-to-one with the order number of all orders in the same time period, and each second data node corresponds to the basic order information of all orders in the same time period.
[0008] The target data nodes in the data node set are merged to obtain a target node set corresponding to the data node set. Each target data node is a second data node that is associated with the basic information of the corresponding order. The new data node after merging the target data nodes is associated with the first data node corresponding to each target data node.
[0009] Construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph;
[0010] Based on the number of orders corresponding to each undirected subgraph, determine whether there is abnormal clustering in the undirected subgraph;
[0011] If abnormal clustering exists, it is determined that the users who placed the orders corresponding to the undirected subgraph are at risk of fraudulent orders.
[0012] Optionally, the basic order information includes the recipient's identifier, recipient's mobile phone number, recipient's address, and payment account information;
[0013] The determined set of data nodes includes:
[0014] Create multiple number nodes, multiple identifier nodes, multiple mobile phone number nodes, multiple address nodes, and multiple payment nodes;
[0015] Construct a one-to-one correspondence between each of the numbered nodes and the order numbers of all orders within the same time period, and use each numbered node after constructing the correspondence as a first data node;
[0016] Construct a one-to-one correspondence between each of the aforementioned identifier nodes and the recipient identifiers of all orders within the same time period;
[0017] Construct a one-to-one correspondence between each of the aforementioned mobile phone number nodes and the receiving mobile phone number information of all orders within the same time period;
[0018] Establish a one-to-one correspondence between each of the address nodes and the delivery address information of all orders within the same time period;
[0019] Establish a one-to-one correspondence between each payment node and the payment account information of all orders within the same time period;
[0020] Each identifier node, each mobile phone number node, each address node, and each payment node, after establishing the corresponding relationship, will be used as a second data node.
[0021] Each of the first data nodes and each of the second data nodes are combined to form a data node set.
[0022] Optionally, constructing a one-to-one correspondence between each address node and the delivery address information of all orders within the same time period includes:
[0023] Obtain the original shipping address data for all orders within the same time period;
[0024] Analyze in turn whether there are two or more administrative names corresponding to the same level of administrative region in the original delivery address data of the order;
[0025] If it exists, the administrative region database is retrieved, and the original delivery address data of the order is updated to obtain the delivery address information of the order.
[0026] Establish the correspondence between the address node and the delivery address information.
[0027] Optionally, the step of retrieving the administrative region database and updating the original delivery address data of the order to obtain the delivery address information of the order includes:
[0028] The administrative region database is searched, and the delivery address data is compared with the addresses in each administrative region database to determine multiple candidate addresses that match the delivery address data.
[0029] The candidate addresses are scored based on the degree of overlap between each candidate address and the delivery address data, as well as the character spacing between different levels of administrative names in the delivery address data.
[0030] The delivery address data is updated to the highest-rated alternative address among all alternative addresses to obtain the delivery address information for the order.
[0031] Optionally, merging the target data nodes in the data node set includes:
[0032] Calculate the similarity between every two second data nodes in the data node set;
[0033] Based on the similarity scores, hierarchical clustering is performed on each of the second data nodes to obtain clustering results. The clustering results include one or more cluster subsets, and each cluster subset includes one or more second data nodes.
[0034] Each second data node in each of the clustered subsets is gradually merged as a target data node.
[0035] Optionally, merging the target data nodes in the data node set includes:
[0036] The basic order information corresponding to each second data node in the data node set is segmented into characters, and each word segment in the basic order information is extracted.
[0037] The orders within the same time period are combined to obtain multiple order sets;
[0038] Count the frequency of each word in each order set;
[0039] Calculate the inverse document coefficient of the word segmentation, where the inverse document coefficient is equal to the target natural logarithm, and the target natural logarithm is the natural logarithm of the ratio of the number of order sets to the number of order sets containing the word segmentation;
[0040] Calculate the Tfidf value of each word corresponding to each order set in sequence. The Tfidf value is equal to the product of the frequency of the word in the order set and the inverse document coefficient of the word.
[0041] The word segment with the highest TFIDF value that is greater than the threshold is selected as the target code for the order set;
[0042] The second data node corresponding to the basic information of the orders containing the target code in the order set is merged as the target data node.
[0043] Optionally, character segmentation is performed on the basic order information corresponding to each of the second data nodes in the data node set, and each word segment in the basic order information is extracted, including:
[0044] For each second data node in the data node set, the basic order information is segmented according to the rules corresponding to the matching regular expression, and each word in the basic order information is extracted.
[0045] Optionally, constructing the undirected graph corresponding to each data node in the target node set includes:
[0046] Based on the correspondence between each second data node and each first data node in the target node set, construct an undirected graph corresponding to each data node in the target node set.
[0047] Optionally, determining whether an abnormal cluster exists in an undirected subgraph based on the number of orders corresponding to each order in the undirected subgraph includes:
[0048] The risk threshold is determined based on the total number of order numbers and the total number of user identifiers who placed the order;
[0049] Determine whether the number of orders corresponding to each of the undirected subgraphs exceeds the risk threshold;
[0050] If the risk threshold is exceeded, it is determined that there is an abnormal clustering in the undirected subgraph.
[0051] A device for identifying fraudulent transactions, comprising:
[0052] A determining unit is used to determine a set of data nodes, wherein the set of data nodes includes a plurality of first data nodes and a plurality of second data nodes, each of the first data nodes corresponds one-to-one with the order number of all orders in the same time period, and each of the second data nodes corresponds to the basic order information of all orders in the same time period.
[0053] The merging unit is used to merge each target data node in the data node set to obtain a target node set corresponding to the data node set, wherein each target data node is a second data node that has an association relationship with the corresponding order basic information, and the new data node after merging each target data node has a corresponding relationship with the first data node corresponding to each target data node;
[0054] A construction unit is used to construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph;
[0055] The judgment unit is used to determine whether there is abnormal clustering in the undirected subgraph based on the number of orders corresponding to each order in the undirected subgraph.
[0056] The order-brushing unit is used to determine if there is abnormal clustering in an undirected subgraph, and whether the users who placed the orders for the corresponding orders in that undirected subgraph are at risk of order-brushing.
[0057] As can be seen from the above technical solution, the order-brushing identification method provided in this application can determine a data node set, which includes multiple first data nodes and multiple second data nodes. Each first data node corresponds one-to-one with the order number of all orders within the same time period, and each second data node corresponds to the basic order information of all orders within the same time period. Based on this, the data node set includes data nodes corresponding to the order number and basic order information of each order. The target data nodes in the data node set can be merged to obtain a target node set corresponding to the data node set. Each target data node is a second data node that has an association with the corresponding basic order information. The new data node after merging the target data nodes has a corresponding first data node. Based on this relationship, second data nodes with related relationships can be merged, and the merged second data node still has a corresponding relationship with the first data node. Furthermore, the new second data node obtained after merging multiple second data nodes has a corresponding relationship with multiple first data nodes. Therefore, through the above process, it is possible to understand whether there is a relationship between the first data nodes of different orders. Construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph. Determine whether there is an abnormal cluster in the undirected subgraph based on the number of orders corresponding to each undirected subgraph. If there is an abnormal cluster, it is determined that the users who placed orders for the orders corresponding to the undirected subgraph are at risk of fraudulent orders. Based on this, by constructing an undirected graph, it is possible to clearly understand whether there is a relationship between different orders and the number of related orders, thereby completing the identification of fraudulent orders. It can be seen that this application can use undirected graphs to analyze the relationship between various orders, reducing the complexity of analyzing the relationship between various orders, and accurately and clearly reflecting the relationship between various orders through undirected graphs, thereby improving the reliability and accuracy of this application in identifying fraudulent orders. Attached Figure Description
[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.
[0059] Figure 1 This is a flowchart of a method for identifying fraudulent transactions disclosed in an embodiment of this application;
[0060] Figure 2 This is a structural block diagram of a brushing identification device disclosed in an embodiment of this application;
[0061] Figure 3 This is a hardware structure block diagram of a brushing identification device disclosed in an embodiment of this application. Detailed Implementation
[0062] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0063] The fraudulent order identification method provided in this application can be applied to a wide range of general-purpose or specialized computing device environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor devices, and distributed computing environments including any of the above devices.
[0064] Next, combine Figure 1 This application provides a detailed description of the method for identifying fraudulent transactions, including the following steps:
[0065] Step S1: Determine the set of data nodes.
[0066] Specifically, the data node set includes two types of data nodes, one type being the first data node and the other type being the second data node.
[0067] Each first data node refers to a data node that corresponds one-to-one with the order number of all orders within the same time period, and each second data node refers to a data node that corresponds to the basic order information of all orders within the same time period.
[0068] There is a correspondence between the first data node and the second data node corresponding to the same order.
[0069] The basic order information can be any one or more of the following: delivery address information, delivery mobile phone number information, delivery user identifier, payment account information, and order placement user identifier.
[0070] The length of the time period can be set according to actual needs, and can be one hour or 20 minutes. For example, each first data node can correspond one-to-one with the order number of all orders in the same hour.
[0071] Each second data node in the data node set corresponds to only one first data node.
[0072] Step S2: Merge each target data node in the data node set to obtain a target node set corresponding to the data node set.
[0073] Specifically, each target data node is a second data node that is related to the corresponding order basic information. The target node set is obtained by merging the second data nodes that are related to the corresponding order basic information and updating the data node set.
[0074] Among them, the order basic information corresponding to the second data node that has a relationship may contain the same or similar special characters.
[0075] The new data node corresponds to the orders of each of the related second data nodes, and the new data node also corresponds to the first data node of each of the related second data nodes.
[0076] The target node set may include each new data node, each first data node, and each second data node that did not participate in the merging.
[0077] Step S3: Construct an undirected graph corresponding to each data node in the target node set.
[0078] Specifically, various methods can be used to retrieve the correspondence between each data node in the target node set and construct an undirected graph corresponding to each data node in the target node set.
[0079] For example, the correspondence between data nodes in a target node set can be retrieved using depth-first search or breadth-first search.
[0080] An undirected graph contains one or more undirected subgraphs, and the undirected subgraphs are independent of each other.
[0081] Step S4: Determine whether there is abnormal clustering in the undirected subgraph based on the number of orders corresponding to each order. If abnormal clustering exists, proceed to step S5.
[0082] Specifically, after constructing the undirected graph, since the data nodes in the undirected graph correspond to orders, the undirected graph containing the data nodes also corresponds to orders.
[0083] Based on the orders corresponding to each data node in the undirected subgraph, the order data corresponding to the undirected subgraph can be determined, and then the presence of abnormal clustering in the undirected subgraph can be judged based on the order data.
[0084] If it is determined that there is an abnormal cluster in the undirected subgraph, proceed to step S5;
[0085] If it is determined that there is no abnormal clustering in the undirected subgraph, it is determined that the users who placed the orders corresponding to the undirected subgraph do not pose a risk of fraudulent orders for the time being.
[0086] Step S5: Determine if the users who placed the orders corresponding to the undirected subgraph are at risk of fraudulent orders.
[0087] Specifically, when there is abnormal clustering in an undirected subgraph, the users who placed orders for the corresponding orders in the undirected subgraph are at risk of fraudulent orders, and technical personnel can be promptly alerted to handle the situation.
[0088] As can be seen from the above technical solution, the order-brushing identification method provided in this application can determine a data node set, which includes multiple first data nodes and multiple second data nodes. Each first data node corresponds one-to-one with the order number of all orders within the same time period, and each second data node corresponds to the basic order information of all orders within the same time period. Based on this, the data node set includes data nodes corresponding to the order number and basic order information of each order. The target data nodes in the data node set can be merged to obtain a target node set corresponding to the data node set. Each target data node is a second data node that has an association with the corresponding basic order information. The new data node after merging the target data nodes has a corresponding first data node. Based on this relationship, second data nodes with related relationships can be merged, and the merged second data node still has a corresponding relationship with the first data node. Furthermore, the new second data node obtained after merging multiple second data nodes has a corresponding relationship with multiple first data nodes. Therefore, through the above process, it is possible to understand whether there is a relationship between the first data nodes of different orders. Construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph. Determine whether there is an abnormal cluster in the undirected subgraph based on the number of orders corresponding to each undirected subgraph. If there is an abnormal cluster, it is determined that the users who placed orders for the orders corresponding to the undirected subgraph are at risk of fraudulent orders. Based on this, by constructing an undirected graph, it is possible to clearly understand whether there is a relationship between different orders and the number of related orders, thereby completing the identification of fraudulent orders. It can be seen that this application can use undirected graphs to analyze the relationship between various orders, reducing the complexity of analyzing the relationship between various orders, and accurately and clearly reflecting the relationship between various orders through undirected graphs, thereby improving the reliability and accuracy of this application in identifying fraudulent orders.
[0089] In some embodiments of this application, the process of step S1, determining the data node set, is described in detail as follows:
[0090] S10. Create multiple number nodes, multiple identifier nodes, multiple mobile phone number nodes, multiple address nodes, and multiple payment nodes.
[0091] Specifically, the basic order information may include the recipient's identifier, recipient's mobile phone number, recipient's address, and payment account information.
[0092] Determine the quantity of all orders within the same time period, and create number nodes, identifier nodes, mobile number nodes, address nodes, and payment nodes with the same quantity.
[0093] S11. Construct a one-to-one correspondence between each of the numbered nodes and the order numbers of all orders within the same time period, and use each numbered node after constructing the correspondence as a first data node.
[0094] Specifically, a one-to-one correspondence can be created between each number node and the order number of each order, and the number node that has a corresponding relationship with the order number is the first data node.
[0095] S12. Construct a one-to-one correspondence between each of the aforementioned identifier nodes and the recipient identifiers of all orders within the same time period.
[0096] Specifically, a one-to-one correspondence can be created between each identifier node and the recipient identifier of each order, so that each identifier node can represent the recipient identifier of each order, and the identifier node that has a corresponding relationship with the recipient identifier belongs to the second data node in the data node set.
[0097] S13. Construct a one-to-one correspondence between each of the mobile phone number nodes and the receiving mobile phone number information of all orders within the same time period.
[0098] Specifically, a one-to-one correspondence can be created between each mobile number node and the receiving mobile number information of each order, so that each mobile number node can represent the receiving mobile number information of each order, and the mobile number node that has a corresponding relationship with the receiving mobile number information belongs to the second data node in the data node set.
[0099] S14. Construct a one-to-one correspondence between each address node and the delivery address information of all orders within the same time period.
[0100] Specifically, a one-to-one correspondence can be created between each address node and the delivery address information of each order, so that each address node can represent the delivery address information of each order, and the address node that has a corresponding relationship with the delivery address information belongs to the second data node in the data node set.
[0101] S15. Construct a one-to-one correspondence between each payment node and the payment account information of all orders within the same time period.
[0102] Specifically, a one-to-one correspondence can be created between each payment node and the payment account information of each order, so that each payment node can represent the payment account information of each order, and the payment node that has a corresponding relationship with the payment account information belongs to the second data node in the data node set.
[0103] S16. After constructing the corresponding relationship, each identifier node, each mobile phone number node, each address node, and each payment node will be used as each second data node.
[0104] Specifically, each second data node can be composed of an identifier node corresponding to the recipient's identifier, a mobile number node corresponding to the recipient's mobile number information, an address node corresponding to the recipient's address information, and a payment node corresponding to the payment account information.
[0105] S17. The first data nodes and the second data nodes are combined into a data node set.
[0106] Specifically, each first data node and each second data node can be combined to form a data node set.
[0107] As can be seen from the above technical solution, the second data node in this application can be composed of multiple types of data nodes such as identification nodes, mobile phone number nodes, address nodes, and payment nodes. Therefore, when merging target data nodes in the future, multiple different types of second data nodes can be considered. Different types of second data nodes indicate different types of order basic information. Therefore, the relationship between each second data node can be considered from multiple dimensions. Thus, by considering whether there is a relationship between each order from multiple dimensions, the reliability and accuracy of this application in identifying fraudulent orders can be further improved.
[0108] In some embodiments of this application, the process of step S14, which involves constructing a one-to-one correspondence between each address node and the delivery address information of all orders within the same time period, is described in detail below:
[0109] S140. Obtain the original delivery address data of all orders within the same time period.
[0110] Specifically, the original shipping address data from each order can be extracted from text.
[0111] S141. Analyze in sequence whether there are two or more administrative names corresponding to the same level of administrative regions in the original delivery address data of the order. If there are two or more administrative names corresponding to the same level of administrative regions, then proceed to step S142.
[0112] Specifically, considering that users who engage in fraudulent order placement may use multiple administrative regions' identifiers to confuse consumers in order to avoid detection, the real delivery address can be determined based on the original delivery address data for fraudulent order placement identification.
[0113] In this process, we can first determine whether there is any possibility of misleading information in the original delivery address data of each order. Therefore, we can sequentially check whether there are two or more administrative names corresponding to the same level of administrative region in the original delivery address data of each order. For example, whether there are two administrative names of municipal-level administrative units at the same time.
[0114] If it is determined that the original delivery address data of the order contains administrative names corresponding to two or more administrative regions at the same level, proceed to step S142.
[0115] If it is determined that there are no two or more administrative names corresponding to the same level of administrative regions in the original delivery address data of the order, the original delivery address data will be directly used as the delivery address information of the order.
[0116] S142. Search the administrative region database and update the original delivery address data to obtain the delivery address information of the order.
[0117] Specifically, each address in the administrative region database can be compared with the delivery address data in turn to retrieve the accurate delivery address information for the order.
[0118] S143. Construct the correspondence between the address node and the delivery address information.
[0119] Specifically, the above method can be used to determine the original delivery address information for each order and to establish a one-to-one correspondence between each address node and each delivery address information.
[0120] As can be seen from the above technical solution, this embodiment provides an optional method for correcting delivery address information. The delivery address information of each order can be determined more accurately through the above method, thereby better judging the relationship between the address nodes corresponding to each delivery address information.
[0121] In some embodiments of this application, the process of step S142, retrieving the administrative region database, updating the original delivery address data, and obtaining the delivery address information of the order, is described in detail below:
[0122] S1420. Retrieve the administrative region database, compare the delivery address data with the addresses in each administrative region database, and determine multiple candidate addresses that match the delivery address data.
[0123] Specifically, when searching the administrative region database, the addresses in each administrative region database can be matched with the delivery address data in turn, and the addresses in the administrative region databases that have the same administrative name as the delivery address data can be used as candidate addresses.
[0124] S1421. The candidate addresses are scored based on the degree of overlap between each candidate address and the delivery address data, as well as the character spacing between administrative names of different levels in the delivery address data.
[0125] Specifically, the overlap score of each candidate address can be determined based on the degree of overlap between the administrative names of each candidate address and the delivery address data.
[0126] The reliability score of the alternative address can be determined based on the character spacing of the administrative names of different levels in the delivery address data.
[0127] The higher the overlap, the higher the overlap score; the shorter the character spacing, the higher the reliability score.
[0128] The score for a candidate address can be the sum of its overlap score and its reliability score.
[0129] S1423. Update the delivery address data to the highest-rated alternative address among all alternative addresses to obtain the delivery address information of the order.
[0130] Specifically, the ratings of each alternative address are sorted, and the alternative address with the highest rating is selected as the order delivery address information.
[0131] As can be seen from the above technical solution, this embodiment provides an optional method for determining the delivery address information of an order using an administrative region database. This method can further improve the practicality and reliability of this application.
[0132] In some embodiments of this application, considering that the basic order information of fraudulent orders has a certain similarity, target data nodes with related relationships can be determined based on similarity. Furthermore, considering that fraudulent orders may contain specific identifiers to help fraudulent users identify them, this embodiment illustrates two optional implementations for step S2: merging the target data nodes in the data node set to obtain a target node set corresponding to the data node set, as follows:
[0133] The first type
[0134] S20. Calculate the similarity between every two second data nodes in the data node set.
[0135] Specifically, we can first calculate the similarity between every two identifier nodes, every two mobile phone number nodes, every two address nodes, and every two payment nodes in the data node set to obtain multiple similarity scores.
[0136] S21. Based on the similarity scores, hierarchical clustering is performed on each of the second data nodes to obtain clustering results. The clustering results include one or more cluster subsets, and each cluster subset includes one or more second data nodes.
[0137] Specifically, the second data node with the highest similarity can be selected iteratively from each similarity level. When the highest similarity exceeds the similarity threshold, the second data node corresponding to that similarity level is clustered to obtain more than one cluster subset. The data nodes contained in each cluster subset are all second data nodes with related relationships.
[0138] The similarity threshold can be set according to actual needs. Generally, the similarity threshold can be 0.7.
[0139] The similarity between the basic information of each two orders can be calculated using any of the following methods: common characters, edit distance, or Hamming distance.
[0140] S22. Gradually merge each second data node in each of the clustered subsets as a target data node.
[0141] Specifically, by merging the second data nodes in each cluster subset step by step, multiple second data nodes can be obtained. The second data nodes in the data node set are then replaced with the merged second data nodes to obtain the target node set.
[0142] Each second data node in the target node set has more than one corresponding first data node, and similarly, each first data node also has more than one corresponding second data node.
[0143] The second type
[0144] S20. Perform character segmentation on the basic order information corresponding to each of the second data nodes in the data node set, and extract each word from the basic order information.
[0145] Specifically, the basic order information corresponding to each second data node can be segmented into characters to obtain each word corresponding to the basic order information of the order.
[0146] S21. Combine the orders within the same time period to obtain multiple order sets.
[0147] Specifically, orders can be combined according to their order of occurrence to obtain multiple order sets. Each order set contains multiple orders, and the combination of all the orders in the order set can yield all the orders within the same time period. Different order sets contain different orders.
[0148] S22. Count the frequency of each word in each order set.
[0149] Specifically, the frequency of each word in each order set can be counted sequentially to obtain multiple frequency counts corresponding to the word, and each frequency count corresponds to an order set.
[0150] S23. Calculate the inverse document coefficient of the word segmentation, wherein the inverse document coefficient is equal to the target natural logarithm, and the target natural logarithm is the natural logarithm of the ratio of the number of order sets to the number of order sets containing the word segmentation.
[0151] Specifically, the inverse document coefficient reflects the uniqueness of word segmentation. The higher the uniqueness of the word segmentation, the higher the inverse document coefficient. Common word segmentations, such as Guangzhou City or Guangdong Province, have relatively low inverse document coefficients.
[0152] The inverse document coefficient can be the natural logarithm of the ratio of the number of order sets to the number of order sets containing the given word.
[0153] S24. Calculate the Tfidf value of each word corresponding to each order set in sequence. The Tfidf value is equal to the product of the frequency of the word in the order set and the inverse document coefficient of the word.
[0154] Specifically, the Tfidf value of each word corresponding to each order set is calculated sequentially. The Tfidf value of the word is equal to the product of the frequency of the word in the order set and the inverse document coefficient of the word.
[0155] S25. Select the word segment with the highest Tfidf value that is greater than the threshold as the target code for the order set.
[0156] Specifically, the Tfidf values corresponding to the order set can be sorted, the highest Tfidf value can be selected, and when the highest Tfidf value is greater than the threshold, the word segment corresponding to the highest Tfidf value can be determined as the target code for the order set.
[0157] S26. Merge the second data node corresponding to the basic information of the orders containing the target code in the order set as the target data node.
[0158] Specifically, the target code for each order set can be determined sequentially, and the second data nodes corresponding to the basic information of the orders containing the target code in the order set can be merged.
[0159] As can be seen from the above technical solutions, this application provides two optional methods for determining the existence of related second data nodes and merging related second data nodes. The above methods can determine whether there is a relationship between second data nodes from two dimensions: similarity and special word segmentation, which further improves the reliability and accuracy of this application.
[0160] In some embodiments of this application, the process of step S20, which involves character segmentation of the basic order information corresponding to each of the second data nodes in the data node set and extraction of each segmented word from the basic order information, is described in detail below:
[0161] S21. For each second data node in the data node set, the basic order information is segmented according to the rules corresponding to the matching regular expression, and each word in the basic order information is extracted.
[0162] Specifically, multiple regular expressions can be pre-defined, and different regular expressions can be set for different basic order information. For example, the regular expressions corresponding to the shipping address information and the payment account information can be different.
[0163] The order basic information corresponding to each second data node is segmented using a pre-set regular expression to obtain each word in the order basic information.
[0164] Word segmentation can be administrative identifiers that frequently appear in the basic order information, or special characters that appear less frequently under normal circumstances.
[0165] As can be seen from the above technical solution process, this embodiment provides an optional method for character segmentation of basic order information. Through the above method, character segmentation of basic order information can be performed in a targeted manner, further improving the possibility of recognizing special word segments.
[0166] In some embodiments of this application, the process of step S3, constructing the undirected graph corresponding to each data node in the target node set, is described in detail as follows:
[0167] S30. Based on the correspondence between each second data node and each first data node in the target node set, construct an undirected graph corresponding to each data node in the target node set.
[0168] Specifically, each second data node can be connected to its corresponding first data node based on the correspondence between each second data node and each first data node in the target node set, thus forming an undirected graph.
[0169] As can be seen from the above technical solution, this embodiment provides an optional way to construct an undirected graph, and the construction of an undirected graph can be further improved through the above method.
[0170] In some embodiments of this application, the process of determining whether an undirected subgraph has abnormal clustering based on the number of orders corresponding to each order in the undirected subgraph is described in detail, and the steps are as follows:
[0171] S40. Determine the risk threshold based on the total number of order numbers and the total number of user identifiers who placed the order.
[0172] Specifically, risk thresholds can be set based on the number of order numbers or the number of individual orders, as well as the number of user identifiers for placing the orders.
[0173] S41. Determine whether the number of orders corresponding to each undirected subgraph exceeds the risk threshold. If it exceeds the risk threshold, proceed to step S42.
[0174] Specifically, determine the number of orders corresponding to each undirected subgraph and compare whether the number of orders exceeds the risk threshold. If it exceeds the risk threshold, execute step S42. If it does not exceed the risk threshold, determine that there is no abnormal clustering in the undirected subgraph.
[0175] S42. Determine that there is an anomalous clustering in the undirected subgraph.
[0176] Specifically, an undirected subgraph with an order quantity exceeding the risk threshold can be an undirected subgraph with abnormal clustering.
[0177] As can be seen from the above technical solution, this embodiment provides an optional method for determining whether there is abnormal clustering in an undirected subgraph. Through the above method, risk thresholds can be set in a targeted manner, thereby achieving better judgment of abnormal clustering.
[0178] The brushing identification device provided in the embodiments of this application is described below. The brushing identification device described below can be referred to in correspondence with the brushing identification method described above.
[0179] See Figure 2 , Figure 2 This is a schematic diagram of a brushing identification device disclosed in an embodiment of this application.
[0180] like Figure 2 As shown, the brushing identification device may include:
[0181] Determining unit 1 is used to determine a set of data nodes, wherein the set of data nodes includes a plurality of first data nodes and a plurality of second data nodes, each of the first data nodes corresponds one-to-one with the order number of all orders in the same time period, and each of the second data nodes corresponds to the basic order information of all orders in the same time period.
[0182] Merging unit 2 is used to merge each target data node in the data node set to obtain a target node set corresponding to the data node set, wherein each target data node is a second data node that has an association relationship with the corresponding order basic information, and the new data node after merging each target data node has a corresponding relationship with the first data node corresponding to each target data node;
[0183] Construction unit 3 is used to construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph;
[0184] Judgment unit 4 is used to determine whether there is abnormal clustering in the undirected subgraph based on the number of orders corresponding to each order in the undirected subgraph;
[0185] Unit 5 is used to determine if there is abnormal clustering, and if so, that the user who placed the order for the order corresponding to the undirected subgraph is at risk of fraudulent activity.
[0186] Optionally, the determining unit may include:
[0187] The node creation unit is used to create multiple numbered nodes, multiple identifier nodes, multiple mobile phone number nodes, multiple address nodes, and multiple payment nodes.
[0188] The order number relationship construction unit is used to construct a one-to-one correspondence between each number node and the order number of all orders in the same time period, and to use each number node after constructing the correspondence as each first data node.
[0189] The receiving user identifier relationship construction unit is used to construct a one-to-one correspondence between each of the identifier nodes and the receiving user identifiers of all orders within the same time period;
[0190] The receiving mobile phone number information construction unit is used to construct a one-to-one correspondence between each of the mobile phone number nodes and the receiving mobile phone number information of all orders within the same time period;
[0191] The delivery address information construction unit is used to construct a one-to-one correspondence between each of the address nodes and the delivery address information of all orders within the same time period;
[0192] The payment account information construction unit is used to construct a one-to-one correspondence between each payment node and the payment account information of all orders within the same time period;
[0193] The second data node determination unit is used to identify each identifier node, each mobile phone number node, each address node and each payment node after the corresponding relationship is constructed as each second data node.
[0194] The node set combination unit is used to combine each of the first data nodes and each of the second data nodes into a data node set.
[0195] Optionally, the delivery address information construction unit may include:
[0196] The address data acquisition unit is used to acquire the original delivery address data of all orders within the same time period;
[0197] The address data analysis unit is used to sequentially analyze whether there are two or more administrative names corresponding to the same level of administrative regions in the original delivery address data of the order;
[0198] The administrative region database retrieval unit is used to retrieve the administrative region database when there are two or more administrative regions of the same level with corresponding administrative names, update the original delivery address data of the order, and obtain the delivery address information of the order.
[0199] The correspondence construction unit is used to construct the correspondence between the address node and the delivery address information.
[0200] Optionally, the administrative region database retrieval unit may include:
[0201] The first administrative region database retrieval unit is used to retrieve the administrative region database, compare the delivery address data with the address of each administrative region database, and determine multiple candidate addresses that match the delivery address data.
[0202] The second administrative region database retrieval unit is used to score the candidate addresses based on the degree of overlap between each candidate address and the delivery address data, as well as the character spacing between different levels of administrative names in the delivery address data.
[0203] The third administrative region database retrieval unit is used to update the delivery address data with the highest-rated alternative address among all alternative addresses, thereby obtaining the delivery address information of the order.
[0204] Optionally, the merging unit may include:
[0205] A similarity calculation unit is used to calculate the similarity between every two second data nodes in the data node set;
[0206] A node clustering unit is used to perform hierarchical clustering on each of the second data nodes according to each similarity to obtain a clustering result. The clustering result includes one or more clustering subsets, and each clustering subset includes one or more second data nodes.
[0207] The node merging unit is used to gradually merge each second data node in each of the clustered subsets as a target data node.
[0208] Optionally, the merging unit may also include:
[0209] The word segmentation and extraction unit is used to perform character segmentation on the basic order information corresponding to each of the second data nodes in the data node set, and extract each word from the basic order information.
[0210] The order combination unit is used to combine various orders within the same time period to obtain multiple order sets;
[0211] The frequency counting unit is used to count the frequency of each word in each order set;
[0212] The inverse document coefficient calculation unit is used to calculate the inverse document coefficient of the word segmentation, wherein the inverse document coefficient is equal to the target natural logarithm, and the target natural logarithm is the natural logarithm of the ratio of the number of order sets to the number of order sets containing the word segmentation;
[0213] The Tfidf value calculation unit is used to calculate the Tfidf value of each word corresponding to each order set in turn. The Tfidf value is equal to the product of the frequency of the word in the order set and the inverse document coefficient of the word.
[0214] The target code determination unit is used to select the word segment with the highest Tfidf value that is greater than the threshold as the target code of the order set.
[0215] The target data node merging unit is used to merge the second data node corresponding to the basic information of the order containing the target code in the order set as the target data node.
[0216] Optionally, the word segmentation extraction unit may include:
[0217] The first word segmentation and extraction unit is used to segment the basic order information corresponding to each second data node in the data node set according to the rules corresponding to the matching regular expression, and extract each word from the basic order information.
[0218] Optionally, the building block may include:
[0219] An undirected graph construction unit is used to construct an undirected graph corresponding to each data node in the target node set based on the correspondence between each second data node and each first data node in the target node set.
[0220] Optionally, the decision unit may include:
[0221] The risk threshold determination unit is used to determine the risk threshold based on the total number of order numbers and the total number of order user identifiers.
[0222] The quantity comparison unit is used to determine whether the number of orders corresponding to each of the undirected subgraphs exceeds the risk threshold.
[0223] An abnormal clustering determination unit is used to determine that an abnormal clustering exists in an undirected subgraph when the number of orders corresponding to the undirected subgraph exceeds the risk threshold.
[0224] The order-brushing identification device provided in this application embodiment can be applied to order-brushing identification equipment, such as PC terminals, cloud platforms, servers, and server clusters. Optionally, Figure 3 The hardware structure block diagram of the order-brushing identification device is shown below. Figure 3 The hardware structure of the order-brushing identification device may include: at least one processor 1, at least one communication interface 2, at least one memory 3, and at least one communication bus 4;
[0225] In this embodiment of the application, the number of processor 1, communication interface 2, memory 3, and communication bus 4 is at least one, and processor 1, communication interface 2, and memory 3 communicate with each other through communication bus 4;
[0226] Processor 1 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.
[0227] Memory 3 may include high-speed RAM, and may also include non-volatile memory, such as at least one disk storage device;
[0228] The memory stores a program, which the processor can call. The program is used for:
[0229] A set of data nodes is determined, wherein the set of data nodes includes multiple first data nodes and multiple second data nodes, each first data node corresponds one-to-one with the order number of all orders in the same time period, and each second data node corresponds to the basic order information of all orders in the same time period.
[0230] The target data nodes in the data node set are merged to obtain a target node set corresponding to the data node set. Each target data node is a second data node that is associated with the basic information of the corresponding order. The new data node after merging the target data nodes is associated with the first data node corresponding to each target data node.
[0231] Construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph;
[0232] Based on the number of orders corresponding to each undirected subgraph, determine whether there is abnormal clustering in the undirected subgraph;
[0233] If abnormal clustering exists, it is determined that the users who placed the orders corresponding to the undirected subgraph are at risk of fraudulent orders.
[0234] Optionally, the refined and extended functions of the program can be referred to the above description.
[0235] This application embodiment also provides a readable storage medium that can store a program suitable for execution by a processor, the program being used for:
[0236] A set of data nodes is determined, wherein the set of data nodes includes multiple first data nodes and multiple second data nodes, each first data node corresponds one-to-one with the order number of all orders in the same time period, and each second data node corresponds to the basic order information of all orders in the same time period.
[0237] The target data nodes in the data node set are merged to obtain a target node set corresponding to the data node set. Each target data node is a second data node that is associated with the basic information of the corresponding order. The new data node after merging the target data nodes is associated with the first data node corresponding to each target data node.
[0238] Construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph;
[0239] Based on the number of orders corresponding to each undirected subgraph, determine whether there is abnormal clustering in the undirected subgraph;
[0240] If abnormal clustering exists, it is determined that the users who placed the orders corresponding to the undirected subgraph are at risk of fraudulent orders.
[0241] Optionally, the refined and extended functions of the program can be referred to the above description.
[0242] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0243] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.
[0244] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. The various embodiments of this application can be combined with each other. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A method for identifying fraudulent order placement, characterized in that, include: A set of data nodes is determined, wherein the set of data nodes includes multiple first data nodes and multiple second data nodes, each first data node corresponds one-to-one with the order number of all orders in the same time period, and each second data node corresponds to the basic order information of all orders in the same time period. The target data nodes in the data node set are merged to obtain a target node set corresponding to the data node set. Each target data node is a second data node that is associated with the basic information of the corresponding order. The new data node after merging the target data nodes is associated with the first data node corresponding to each target data node. Construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph; Based on the number of orders corresponding to each undirected subgraph, determine whether there is abnormal clustering in the undirected subgraph; If abnormal clustering exists, it is determined that the users who placed the orders corresponding to the undirected subgraph are at risk of fraudulent orders. The process of merging the target data nodes in the data node set includes: The basic order information corresponding to each second data node in the data node set is segmented into characters, and each word segment in the basic order information is extracted. The orders within the same time period are combined to obtain multiple order sets; Count the frequency of each word in each order set; Calculate the inverse document coefficient of the word segmentation, where the inverse document coefficient is equal to the target natural logarithm, and the target natural logarithm is the natural logarithm of the ratio of the number of order sets to the number of order sets containing the word segmentation; Calculate the Tfidf value of each word corresponding to each order set in sequence. The Tfidf value is equal to the product of the frequency of the word in the order set and the inverse document coefficient of the word. The word segment with the highest TFIDF value that is greater than the threshold is selected as the target code for the order set; The second data node corresponding to the basic information of the orders containing the target code in the order set is merged as the target data node.
2. The method for identifying fraudulent transactions according to claim 1, characterized in that, The basic order information includes the recipient's identifier, recipient's mobile phone number, delivery address, and payment account information; The determined set of data nodes includes: Create multiple number nodes, multiple identifier nodes, multiple mobile phone number nodes, multiple address nodes, and multiple payment nodes; Construct a one-to-one correspondence between each of the numbered nodes and the order numbers of all orders within the same time period, and use each numbered node after constructing the correspondence as a first data node; Construct a one-to-one correspondence between each of the aforementioned identifier nodes and the recipient identifiers of all orders within the same time period; Construct a one-to-one correspondence between each of the aforementioned mobile phone number nodes and the receiving mobile phone number information of all orders within the same time period; Establish a one-to-one correspondence between each of the address nodes and the delivery address information of all orders within the same time period; Establish a one-to-one correspondence between each payment node and the payment account information of all orders within the same time period; Each identifier node, each mobile phone number node, each address node, and each payment node, after establishing the corresponding relationship, will be used as a second data node. Each of the first data nodes and each of the second data nodes are combined to form a data node set.
3. The method for identifying fraudulent transactions according to claim 2, characterized in that, The process of establishing a one-to-one correspondence between each address node and the delivery address information of all orders within the same time period includes: Obtain the original shipping address data for all orders within the same time period; Analyze in turn whether there are two or more administrative names corresponding to the same level of administrative region in the original delivery address data of the order; If it exists, the administrative region database is retrieved, and the original delivery address data of the order is updated to obtain the delivery address information of the order. Establish the correspondence between the address node and the delivery address information.
4. The method for identifying fraudulent transactions according to claim 3, characterized in that, The process of retrieving the administrative region database updates the original delivery address data of the order, resulting in the delivery address information of the order, including: The administrative region database is searched, and the delivery address data is compared with the addresses in each administrative region database to determine multiple candidate addresses that match the delivery address data. The candidate addresses are scored based on the degree of overlap between each candidate address and the delivery address data, as well as the character spacing between different levels of administrative names in the delivery address data. The delivery address data is updated to the highest-rated alternative address among all alternative addresses to obtain the delivery address information for the order.
5. The method for identifying fraudulent transactions according to claim 1, characterized in that, The basic order information corresponding to each of the second data nodes in the data node set is segmented into characters, and each word segment in the basic order information is extracted, including: For each second data node in the data node set, the basic order information is segmented according to the rules corresponding to the matching regular expression, and each word in the basic order information is extracted.
6. The method for identifying fraudulent transactions according to claim 1, characterized in that, The construction of the undirected graph corresponding to each data node in the target node set includes: Based on the correspondence between each second data node and each first data node in the target node set, construct an undirected graph corresponding to each data node in the target node set.
7. The method for identifying fraudulent transactions according to claim 1, characterized in that, The step of determining whether an undirected subgraph has abnormal clustering based on the number of orders corresponding to each order includes: The risk threshold is determined based on the total number of order numbers and the total number of user identifiers who placed the order; Determine whether the number of orders corresponding to each of the undirected subgraphs exceeds the risk threshold; If the risk threshold is exceeded, it is determined that there is an abnormal clustering in the undirected subgraph.
8. A device for identifying fraudulent transactions, characterized in that, include: A determining unit is used to determine a set of data nodes, wherein the set of data nodes includes a plurality of first data nodes and a plurality of second data nodes, each of the first data nodes corresponds one-to-one with the order number of all orders in the same time period, and each of the second data nodes corresponds to the basic order information of all orders in the same time period. A merging unit is used to merge the target data nodes in the data node set to obtain a target node set corresponding to the data node set. Each target data node is a second data node that is associated with the corresponding order basic information. The new data nodes after merging the target data nodes are respectively associated with the first data node corresponding to each target data node. The merging process of the target data nodes in the data node set includes character segmentation of the order basic information corresponding to each second data node in the data node set, extracting each word from the order basic information; and grouping the orders within the same time period. The process involves merging multiple order sets; counting the frequency of each word in each order set; calculating the inverse document coefficient of each word, where the inverse document coefficient is equal to the target natural logarithm, which is the natural logarithm of the ratio of the number of order sets to the number of order sets containing the word; sequentially calculating the Tfidf value of each word corresponding to each order set, where the Tfidf value is equal to the product of the frequency of the word in the order set and the inverse document coefficient of the word; selecting the word with the highest Tfidf value that is greater than a threshold as the target code for that order set; and merging the second data nodes corresponding to the basic information of the orders containing the target code in that order set as target data nodes. A construction unit is used to construct an undirected graph corresponding to each data node in the target node set, wherein the undirected graph includes at least one independent undirected subgraph; The judgment unit is used to determine whether there is abnormal clustering in the undirected subgraph based on the number of orders corresponding to each order in the undirected subgraph. The order-brushing unit is used to determine if there is abnormal clustering in an undirected subgraph, and whether the users who placed the orders for the corresponding orders in that undirected subgraph are at risk of order-brushing.
Citation Information
Patent Citations
User recognition method and device and computer equipment
CN110648195A
Address information detection method and device, equipment, medium and product
CN114528364A
