CN116451128A - False transaction object detection method, false transaction object detection device and server - Google Patents

False transaction object detection method, false transaction object detection device and server Download PDF

Info

Publication number
CN116451128A
CN116451128A CN202210005494.7A CN202210005494A CN116451128A CN 116451128 A CN116451128 A CN 116451128A CN 202210005494 A CN202210005494 A CN 202210005494A CN 116451128 A CN116451128 A CN 116451128A
Authority
CN
China
Prior art keywords
isomorphic
node
user
commodity
target
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210005494.7A
Other languages
Chinese (zh)
Inventor
张浩鑫
王黎
林睿
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Jingdong Century Trading Co Ltd
Beijing Wodong Tianjun Information Technology Co Ltd
Original Assignee
Beijing Jingdong Century Trading Co Ltd
Beijing Wodong Tianjun Information Technology Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Beijing Jingdong Century Trading Co Ltd, Beijing Wodong Tianjun Information Technology Co Ltd filed Critical Beijing Jingdong Century Trading Co Ltd
Priority to CN202210005494.7A priority Critical patent/CN116451128A/en
Publication of CN116451128A publication Critical patent/CN116451128A/en
Pending legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q30/00—Commerce
    • G06Q30/06—Buying, selling or leasing transactions
    • G06Q30/0601—Electronic shopping [e-shopping]
    • G06Q30/0633—Lists, e.g. purchase orders, compilation or processing
    • G06Q30/0635—Processing of requisition or of purchase orders

Landscapes

  • Business, Economics & Management (AREA)
  • Accounting & Taxation (AREA)
  • Finance (AREA)
  • Development Economics (AREA)
  • Economics (AREA)
  • Marketing (AREA)
  • Strategic Management (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The disclosure provides a detection method of a false transaction object, and belongs to the technical field of security. The method comprises the following steps: determining the connection value of the edges of each isomorphic diagram in M first isomorphic diagrams, wherein the M first isomorphic diagrams comprise at least one isomorphic diagrams of users and isomorphic diagrams of commodities, and M is a positive integer; cutting edges with connection value smaller than or equal to preset connection value in the M first isomorphic graphs to obtain M second isomorphic graphs; performing graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs to obtain a target model; and detecting whether a false transaction object exists in the target order according to the target model, wherein the target order is any order. By the technical scheme provided by the embodiment of the disclosure, the problem that the false transaction detection accuracy is low due to inaccurate training model can be solved.

Description

False transaction object detection method, false transaction object detection device and server
Technical Field
The disclosure belongs to the technical field of security, and particularly relates to a method, a device and a server for detecting false transaction objects.
Background
In the e-commerce industry, there are some shops that employ a large number of users to falsely purchase the shops' merchandise in order to promote sales, and to conduct a bill, thereby generating a false transaction.
Currently, to detect whether a transaction is a spurious transaction, users, orders, and goods are connected to construct an alien composition. The composition mode generally carries out edge connection according to whether two orders come from the same user or whether the two orders contain the same commodity, and finally a different composition of 'user-order-commodity' is built. And training the graphic neural network on the graph to obtain a training model for judging whether an order is an order of a false transaction.
However, the e-commerce platform generates massive transaction behaviors every day, and relates to a large number of commodities and users, and the map formed by directly connecting the users with purchase relations, the commodities and orders can lead to huge scale of the map, the training difficulty is increased rapidly, the training model is possibly inaccurate, and the false transaction detection accuracy is low.
Disclosure of Invention
The embodiment of the disclosure aims to provide a method, a device and a server for detecting a false transaction object, which can solve the problem that a trained model is inaccurate enough, so that false transaction detection accuracy is low.
In order to solve the above technical problems, the present disclosure is implemented as follows:
In a first aspect, an embodiment of the present disclosure provides a method for detecting a false transaction object, including: determining the connection value of the edges of each isomorphic diagram in M first isomorphic diagrams, wherein the M first isomorphic diagrams comprise at least one isomorphic diagrams of users and isomorphic diagrams of commodities, and M is a positive integer; cutting edges with connection value smaller than or equal to preset connection value in the M first isomorphic graphs to obtain M second isomorphic graphs; performing graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs to obtain a target model; and detecting whether a false transaction object exists in the target order according to the target model, wherein the target order is any order.
In a second aspect, an embodiment of the present disclosure provides a detection apparatus for a false transaction object, the detection apparatus including: the system comprises a determining module, a cutting module, a model training module and a detecting module; the determining module is used for determining the connection value of the edges of each isomorphic diagram in M first isomorphic diagrams, wherein the M first isomorphic diagrams comprise at least one isomorphic diagrams of users and isomorphic diagrams of commodities, and M is a positive integer; the clipping module is used for clipping edges with the connection value smaller than or equal to a preset connection value in the M first isomorphic graphs according to the connection value determined by the determining module to obtain M second isomorphic graphs; the model training module is used for carrying out graph training according to the neighbor characteristics of the neighbor nodes in the M second isomorphic graphs obtained by the cutting module, so as to obtain a target model; the detection module is used for detecting whether a false transaction object exists in a target order or not according to the target model trained by the model training module, and the target order is any order.
In a third aspect, embodiments of the present disclosure provide an electronic device comprising a processor, a memory and a program or instruction stored on the memory and executable on the processor, the program or instruction when executed by the processor implementing the steps of the method according to the first aspect.
In a fourth aspect, embodiments of the present disclosure provide a readable storage medium having stored thereon a program or instructions which when executed by a processor perform the steps of the method according to the first aspect.
In a fifth aspect, embodiments of the present disclosure provide a chip comprising a processor and a communication interface, the communication interface being coupled to the processor, the processor being configured to execute programs or instructions to implement the method according to the first aspect.
In a sixth aspect, embodiments of the present disclosure provide a computer program product comprising instructions which, when run on a computer, cause the computer to perform the steps of the method as described in the first aspect.
In the embodiment of the present disclosure, first, a server may determine a connection value of an edge of each isomorphic graph in M first concurrent graphs, and then, the server cuts an edge of the M first concurrent graphs, where the connection value is less than or equal to a preset connection value, to obtain M second isomorphic graphs; then, the server can conduct graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs to obtain a target model; finally, the server can determine whether any order has false transaction objects according to the target model. Because the M first co-graphs include at least one of a user isomorphic graph of the user and a user graph of the commodity, M is a positive integer, the data size of the embodiment of the disclosure when model training is first smaller than the data size of model training using a user-order-commodity isomorphic graph in the related art. Secondly, edges with connection values smaller than or equal to preset connection values in the M first synchronous patterns are cut off, so that on one hand, the processing amount of data in model training can be reduced again, and the rate of model training is improved; on the other hand, deleting some noise data can improve the accuracy of the trained target model, can enable the accuracy of determining the false transaction to be higher, for example, can distinguish whether a user in an order is a bill-refreshing user or a common user more accurately, distinguish whether goods in an order are bill-refreshing goods or common goods more accurately, and distinguish whether an order is a false transaction order more accurately.
Drawings
Fig. 1 is a flow chart of a method for detecting a false transaction object according to an embodiment of the disclosure;
fig. 2 is a schematic structural diagram of an isomorphic diagram according to an embodiment of the disclosure;
fig. 3 is a logic schematic diagram of pruning with isomorphic diagrams according to an embodiment of the present disclosure;
FIG. 4 is a schematic structural diagram of an isomerism diagram provided in an embodiment of the disclosure;
FIG. 5 is a schematic diagram of a multi-task learning framework provided by embodiments of the present disclosure;
fig. 6 is a schematic diagram of a possible structure of a detection device for a false transaction object according to an embodiment of the disclosure;
fig. 7 is a schematic diagram of a possible structure of a server according to an embodiment of the disclosure;
fig. 8 is a hardware schematic of a server according to an embodiment of the disclosure.
Detailed Description
The following description of the technical solutions in the embodiments of the present disclosure will be made clearly and completely with reference to the accompanying drawings in the embodiments of the present disclosure, and it is apparent that the described embodiments are some embodiments of the present disclosure, but not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without inventive effort, based on the embodiments in this disclosure are intended to be within the scope of this disclosure.
The terms first, second and the like in the description and in the claims, are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged, where appropriate, such that embodiments of the disclosure may be practiced in sequences other than those illustrated and described herein, and that the objects identified by "first," "second," etc. are generally of the same type and are not limited to the number of objects, e.g., the first object may be one or more. Furthermore, in the description and claims, "and/or" means at least one of the connected objects, and the character "/", generally means that the associated object is an "or" relationship.
The detection method provided by the embodiment of the disclosure is described in detail below through specific embodiments and application scenarios thereof with reference to the accompanying drawings.
Fig. 1 is a flowchart of a method for detecting a false transaction object according to an embodiment of the present disclosure, as shown in fig. 1, the method includes the following steps S101 to S104:
s101, a server determines the connection value of the edges of each isomorphic diagram in M first isomorphic diagrams.
The M first co-constructs comprise at least one of a user isomorphic diagram and a commodity isomorphic diagram, and M is a positive integer.
Specifically, in the isomorphic diagram of the user, the edge connection value may be associated with the user characteristics of the two nodes of the edge, and in the isomorphic diagram of the commodity, the edge connection value may be associated with the commodity characteristics of the two nodes of the edge.
Typically, the connection nodes of the isomorphic graph are the same type of connection node. The connection nodes of the iso-pattern comprise at least two types of connection nodes.
In the embodiment of the disclosure, the isomorphic diagrams of the users indicate the diagrams connected by the users who purchase the same commodity, and the isomorphic diagrams of the commodities indicate the diagrams connected by the commodities purchased by the same user.
For convenience of description, in the following schematic diagrams, ui represents the user node i, and pj represents the commodity node j.
Fig. 2 is a schematic structural diagram of an isomorphic diagram provided in an embodiment of the disclosure, and as shown in fig. 2 (a), is a user-user isomorphic diagram, where u1, u2, u3, and u4 are connected in pairs, and each connected user has purchased the same commodity. That is, u1 and u2 purchased the same commodity, u1 and u3 purchased the same commodity, u1 and u4 purchased the same commodity, u2 and u3 purchased the same commodity, and u3 and u4 purchased the same commodity; as shown in fig. 2 (b), the two-by-two connection is shown as a commodity-commodity isomorphic diagram, wherein p1, p2, p3, and p4 are connected in pairs, and each connected commodity is a commodity purchased by the same user. That is, p1 and p2 are items purchased by the same user, p1 and p3 are items purchased by the same user, p1 and p4 are items purchased by the same user, p2 and p3 are items purchased by the same user, and p3 and p4 are items purchased by the same user.
It should be noted that the foregoing is merely a simple example of one isomorphic diagram, in practical application, the number of nodes in one isomorphic diagram may be more or less than the foregoing example, the number of edges in one isomorphic diagram may be more or less than the foregoing example, and the nodes in the isomorphic diagram may not be fully connected with other nodes.
S102, the server cuts edges with connection value smaller than or equal to a first preset connection value in the M first isomorphic graphs to obtain M second isomorphic graphs.
Illustratively, fig. 3 is a logic schematic diagram of pruning of an isomorphic diagram provided by an embodiment of the present disclosure. In combination with fig. 2 (a), as shown in fig. 3 (a), which is a logic schematic diagram of pruning of a isomorphic graph of a user, the server may determine the connection value of the edge between the user nodes, which is shown in fig. 3 (a), where E34 represents the connection value of the edge between the user node 3 and the user node 4, and if E34 is smaller than the preset connection value, the server may clip the edge between the user node 3 and the user node 4, i.e. as shown in fig. 3 (b). Similarly, referring to fig. 2 (b), as shown in fig. 3 (c), which is a logic diagram of pruning of the isomorphic diagram of the commodity, the server may determine the connection value of the edges between commodity nodes, which is shown in fig. 3 (c), where E24 represents the connection value of the edges between commodity node 2 and commodity node 4, and if E24 is smaller than the preset connection value, the server may cut out the edges between commodity node 2 and commodity node 4.
And S103, the server performs graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs to obtain a target model.
For example, the server may perform average aggregation (Mean aggregation) according to the sampled neighbor features, and perform Graph neural network training by using Graph SAGE, so as to obtain Graph mapping of the M second isomorphic graphs. The Mean aggregation is an aggregation mode of a graph algorithm, and an average value of impression quantities of a node and all neighborhoods of the node is taken:
specifically, in the case that the M second isomorphic graphs are isomorphic graphs of the user, the server may input Graph mapping of the isomorphic graphs of the user into a softmax layerIn the (normalized exponential function), model training is carried out to obtain a user classification model, so that user classification can be realized according to the user classification model, and a bill user and a normal user can be distinguished. Wherein, the liquid crystal display device comprises a liquid crystal display device,
specifically, in the case that the M second isomorphic diagrams are isomorphic diagrams of the commodity, the server may input Graph mapping of the isomorphic diagrams of the commodity into one softmax layer, and perform model training to obtain a commodity classification model, so that commodity classification can be implemented according to the commodity classification model, and the bill-printed commodity and the normal commodity can be distinguished.
Under the condition that the M second isomorphic diagrams comprise the isomorphic diagrams of the user and the isomorphic diagrams of the commodity, the server can also combine Graph subedding of the isomorphic diagrams of the user, graph subedding of the isomorphic diagrams of the commodity and the characteristics of the order, input the characteristics of the order into an MLP (Multiple Layer Perceptron) neural network model for model training to obtain an order classification model, so that order classification can be realized according to the order classification model, and false transaction orders and normal orders can be distinguished.
That is, in the embodiment of the present disclosure, the server may perform multitasking training according to the structural relationship and feature information fusion of the user, the commodity, and the order, to obtain multiple classification models.
S104, the server detects whether a false transaction object exists in the target order according to the target model.
Wherein the target order is any order.
For example, if the target model is the first model, the server may detect whether the user in one order is a billing user according to the first model; if the target model is a second model, the server can detect whether the commodity in an order is a bill-refreshing commodity according to the second model; if the target model is a third model, the server may detect whether an order is a spurious trade order according to the third model.
In general, some false transaction behaviors exist in the electronic commerce platform, and merchants of some stores hire a large number of false purchases of commodities of direct stores, such as false delivery, so that sales of the stores are promoted.
In combination with the method for detecting the false transaction object provided by the embodiment of the disclosure, the false transaction object which can be detected can comprise at least one of a bill user, a bill commodity, a bill order, a bill merchant and a bill comment.
Wherein, the isomorphic diagram of the user and the isomorphic diagram of the comments of the commodity can be combined, and which comments are the bill comments can be determined.
According to the detection method for the false transaction object, firstly, a server can determine the connection value of the edges of each isomorphic graph in M first isomorphic graphs, and then the server cuts edges with the connection value smaller than or equal to the preset connection value in the M first isomorphic graphs to obtain M second isomorphic graphs; then, the server can conduct graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs to obtain a target model; finally, the server can determine whether any order has false transaction objects according to the target model. Because the M first co-graphs include at least one of a user isomorphic graph of the user and a user graph of the commodity, M is a positive integer, the data size of the embodiment of the disclosure when model training is first smaller than the data size of model training using a user-order-commodity isomorphic graph in the related art. Secondly, edges with connection values smaller than or equal to preset connection values in the M first synchronous patterns are cut off, so that on one hand, the processing amount of data in model training can be reduced again, and the rate of model training is improved; on the other hand, deleting some noise data can improve the accuracy of the trained target model, can enable the accuracy of determining the false transaction to be higher, for example, can distinguish whether a user in an order is a bill-refreshing user or a common user more accurately, distinguish whether goods in an order are bill-refreshing goods or common goods more accurately, and distinguish whether an order is a false transaction order more accurately.
Optionally, the embodiment of the present disclosure provides a method for detecting a false transaction object, which may further include the following steps S105 to S107 before S103 described above:
s105, the server acquires the target characteristics of each node of each isomorphic diagram in the M second isomorphic diagrams.
Wherein the target features include attribute features and structural features.
The attribute feature of a node may be an attribute including the node in the isomorphic graph, and the structural feature of a node may be a structural relationship including the node and an adjacent node.
Illustratively, in the isomorphic diagram of the user, the attribute features of the user may include information such as age, registration event, etc. of the user; the structural features of the user may be characterized by an adjacency matrix of isomorphic diagrams of the user.
For example, in the isomorphic diagram of the commodity, the attribute features of the commodity can include information such as the type, the number and the purchase times of the commodity, and the structural features of the commodity can be represented by adopting a connection matrix of the isomorphic diagram of the commodity.
S106, the server determines the target similarity of each node and the neighbor nodes according to the target characteristics.
In the embodiment of the present disclosure, the target similarity is a weighted sum of attribute similarity and structural similarity.
Illustratively, the target similarity of the first node and the second node is a weighted sum of the attribute similarity of the first node and the second node and the structural similarity of the first node and the second node.
The attribute similarity of the first node and the second node is the similarity of the attribute of the node, and the structural similarity of the first node and the second node is the similarity of the structural relationship of the nodes in the isomorphic diagram.
And S107, the server samples the neighbor nodes according to the probability according to the target similarity to obtain neighbor features.
Specifically, the server may determine the collected sampling probability of each node according to the target similarity, and then sample to obtain the neighbor feature of a node according to the collection probability of the node.
For example, assuming that the nodes connected to the node 1 are the node 2 and the node 3, wherein the target similarity s12=5 of the node 1 and the node 2, and the target similarity s13=1 of the node 1 and the node 3, when the neighbor feature of the node 1 is acquired, the probability that the node 2 is acquired is 5/(5+1) =5/6, and the probability that the node 3 is acquired is 1/(5+1) =1/6.
It can be understood that, assuming that there are 100 neighbor nodes in a node, 10 neighbor nodes need to be acquired to determine the neighbor characteristics of the node, sampling can be performed according to the sampling probability of each node.
It should be noted that, in the case where the number of neighbor nodes is less than or equal to the number of neighbor nodes to be acquired, the neighbor nodes may be resampled.
Based on the scheme, when the server acquires the neighbor characteristics of each node, the server can sample according to the target similarity of the nodes in the target isomorphic graph and probability, so that the target similarity of the nodes obtained by sampling is higher, and the obtained neighbor characteristics can more accurately represent the neighbor characteristics of one node and the neighbor nodes.
It can be appreciated that in the embodiment of the present disclosure, the isomorphic diagram in S101 may be an isomorphic diagram generated according to a relationship of a user-order-commodity, or an isomorphic diagram obtained according to a conventional isomorphic diagram of a user-order-commodity, which is not limited in particular in the embodiment of the present disclosure.
Optionally, the method for detecting a false transaction object provided in the embodiment of the present disclosure may further include the following S108 before S101:
s108, splitting the target iso-composition into M first co-compositions by the server.
Wherein the target isograms are user-order-commodity isograms, and the M first co-grams include at least one of a user isomorphic chart and a commodity isomorphic chart.
Fig. 4 is a schematic structural diagram of an iso-composition according to an embodiment of the present disclosure, where, as shown in fig. 4, a transaction a is issued for a commodity p by a user u, and in the heterogeneous diagram formed by the user u (node) and the transaction a (node), an edge exists between the transaction a (node) and the commodity p (node). Specifically, in the isomerism diagram, u1 purchases 4 commodities, namely p1, p2, p3 and p4, and the generated transactions are a1, a2, a3 and a4 in sequence; u2 purchased commodity p4, resulting in transaction a5; u3 purchased commodity p4, resulting in transaction a6; u4 purchased commodity p4, resulting in transaction a7.
With reference to fig. 4, the server may perform a split-graph processing on the iso-graph of fig. 4, and connect users who have purchased the same commodity, so as to obtain an isomorphic graph of the users, and connect commodities purchased by the same user, so as to obtain an isomorphic graph of the commodity.
Based on the scheme, the server can split the traditional heterogeneous graph, so that an isomorphic graph with a simple graph structure is obtained, model training is performed according to the isomorphic graph with a simple graph structure, and compared with model training performed by adopting the heterogeneous graph, the data size during model training can be reduced, the complexity of model training is reduced, and the training efficiency of model training can be improved.
Optionally, the method for detecting a false transaction object provided by the embodiment of the present disclosure, S101 described above may be specifically executed through S11 and S12 described below:
s11, the server determines the connection heat of the nodes of each isomorphic diagram in the M first isomorphic diagrams.
Wherein, the connection heat of the nodes in the isomorphic diagram of the user is related to the purchase quantity of the user, and the connection heat of the nodes in the isomorphic diagram of the commodity is related to the sales quantity of the commodity
For example, the connection hotness may indicate a connection weight of the node with other nodes in the isomorphic graph.
For example, if the M first co-graphs are isomorphic graphs of users, the connection hotness of one user node is related to the purchase amount of the users. Wherein, if the purchase amount of the user is higher, the connection heat of the user node is lower; the lower the purchase amount of the user, the higher the connection heat of the user node.
It should be noted that, if the purchase amount of a user is higher, the probability that the user is a normal user is far higher than the probability that the user is a bill, the edges sent by the user have less help to the server to determine the false transaction later, are noise information in the data processing process, and if the edges sent by the users are all connected, the graph scale is easy to increase.
For example, if the M first co-graphs are isomorphic graphs of the commodity, the connection heat of one commodity node is related to the sales volume of the commodity, wherein if the sales volume of the commodity is higher, the connection heat of the commodity node is lower; the lower the sales of the commodity, the higher the connection heat of the commodity node.
It should be noted that, if the sales volume of a commodity is higher, the probability that the commodity is a normal commodity is far higher than the probability that the commodity is a bill, edges generated by the commodity are less helpful to the server to determine false transactions later, are noise information in the data processing process, and if edges sent by the commodity are all related, the map scale is easy to increase.
And S12, the server determines the connection value of the edges of each isomorphic graph in the M first isomorphic graphs according to the connection heat.
Illustratively, the connection value of an edge indicates the connection weight of the edge between two nodes connected.
For example, for a user's isomorphic diagram, the more items two users purchase together, the higher the value of the connection of the edges between the two users; the less merchandise two users purchase together, the lower the value of the connection of the edge between the two users.
For example, for a isomorphic diagram of two commodities, the more users the two commodities purchase together, the higher the connection value of the edges between the two commodities; the less users the two items purchase together, the lower the value of the connection of the edge between the two items.
Based on the scheme, the server can determine the connection heat of each node in the M first concurrent patterns, and then determine the connection value of each edge in the M first concurrent patterns according to the connection heat of each node. The connection value of the edges of the isomorphic graphs of the users can reflect the situation that the users purchase goods together, the connection value of the edges of the isomorphic graphs of the goods can reflect the situation that whether the goods are purchased by different users or not, and the structure of M second isomorphic graphs can be more simplified by pruning the isomorphic graphs through the connection value of the edges, so that noise information is removed for subsequent machine learning, the processing capacity of graph data is reduced, and therefore the training efficiency of the false transaction model can be improved.
Optionally, the method for detecting a false transaction object provided by the embodiment of the present disclosure, S11 described above may specifically be executed through S11a or S11b described below:
s11a, the server determines the connection heat of the nodes in the isomorphic diagram of the commodity according to the formula (1) and the sales of the commodity.
Wherein vp is i Representing the connection heat of the nodes of the commodity i, S i The sales (sale) of commodity i are indicated.
It should be noted that, in the embodiment of the present disclosure, the sales amount of the commodity may be the total amount of historical sales amounts of the commodity, and sales amounts within a preset period of time may be used, for example, sales amounts within the last month, the last half year, and the last year.
And S11b, the server determines the connection heat of the nodes in the isomorphic diagram of the user according to the formula (2) and the purchase quantity of the user.
Wherein vu is j Representing the connection heat of the nodes of user j, p j Representing the purchase amount (purchase) of user j.
The purchase amount of the user indicates the number of orders for the user to purchase the commodity.
In the embodiment of the disclosure, all the historical orders of the user in the platform can be adopted, and the historical orders in the preset time period can also be adopted, for example, the data of the historical orders in the last month, the last half year and the last year.
Based on the scheme, the server can determine the connection heat of the user node based on the purchase quantity of the user and the connection heat of the commodity node based on the sales quantity of the commodity based on the two formulas.
Optionally, the method for detecting a false transaction object provided by the embodiment of the present disclosure, S12 described above may specifically be executed through S12a or S12b described below:
And S12a, the server determines the connection value of the edges in the isomorphic diagram of the commodity according to the formula (3) and the connection heat of each node.
Wherein EU is ab Representing the connection value of the edge between user a and user b, p ab Representing the collection of items purchased by both user a and user b, p a Representing the collection of items purchased by user a, p b Representing the collection of items purchased by user b.
It can be understood that, in the commodities purchased by two users, if the two users purchase fewer commodities together, i.e. the connection heat of the public commodity is lower, the connection value of the edges of the two users in the isomorphic diagram is lower.
Based on the scheme, the server can accurately determine the connection value of each side in the isomorphic diagram of the user based on the formula (3).
And S12b, the server determines the connection value of the edges in the isomorphic diagram of the user according to the formula (4) and the connection heat of each node.
Wherein EP xy Representing the value of the connection of the edges between commodity x and commodity y, P xy Representing a set of users who have purchased both commodity x and commodity y, U x A user collection representing purchased goods x; u (U) y User set representing purchased goods yCombining; u (U) xy Representing the collection of users who purchased commodity x and commodity y.
It will be appreciated that for two products, the lower the connection heat of the user who has purchased the product, i.e. the lower the connection heat of the public user, the lower the connection value of the edges of the two products between the nodes in the isomorphic diagram.
Based on the scheme, the server can accurately determine the connection value of each side in the commodity isomorphic diagram based on the formula (4).
Optionally, the method for detecting a false transaction object according to the embodiment of the present disclosure, S107 described above may specifically be executed through S71 below:
and S71, the server samples the neighbor nodes according to the probability according to the weighted sum of the attribute similarity and the structure similarity of the nodes to obtain neighbor features.
In an exemplary case that the node is a user node, the server samples neighbor nodes of the user node according to probability according to a weighted sum of attribute similarity of the user node and structure similarity of the user node to obtain neighbor features corresponding to the user; and under the condition that the node is a commodity node, the server samples the neighbor nodes of the commodity node according to the probability according to the weighted sum of the attribute similarity of the commodity node and the structural similarity of the commodity node to obtain the neighbor features corresponding to the user.
Based on the scheme, the server can combine the attribute similarity and the structure similarity of the nodes, samples the neighbor nodes for one node according to the probability, and can enable the sampled neighbor nodes to have higher similarity with the node, so that the accuracy of model training can be improved.
Optionally, the target features include attribute features and structural features. Further, in the method for detecting a false transaction object provided in the embodiment of the present disclosure, S106 described above may be specifically executed by the following S61 to S63:
and S61, the server determines the attribute similarity of each node and the neighbor nodes according to the attribute characteristics of the nodes and the formula (5).
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing the similarity of the attributes of node i and node j, < >>Attribute feature representing node i, < >>Representing the attribute characteristics of node j.
Note that, the attribute similarity between the node i and the node j calculated by the above formula may also be referred to as cosine similarity between the node i and the node j.
S62, the server determines the structural similarity of each node and the neighbor nodes according to the structural characteristics of the nodes and the formula (6).
Wherein, the liquid crystal display device comprises a liquid crystal display device,representing the structural similarity of node i and node j, < >>Representing the structural features of node i->Representing the structural features of node j.
Note that, the structural phase velocity of the node i and the node j calculated by the above formula may also be referred to as the euclidean distance between the node i and the node j.
And S63, the server determines the target similarity of each node and the neighbor nodes according to the attribute similarity, the structure similarity and the formula (7).
Wherein S is ij Representing the target similarity of node i and node j, ω representing the weight factor.
Note that ω is used to balance attribute similarity and structural similarity. The value of ω may be a predetermined value or may be empirically set. For example, ω may be 0.5.
Based on the scheme, the server can calculate the characteristic similarity and the structural similarity of the node and the neighbor node of the node respectively based on the formula.
Optionally, the method for detecting a false transaction object provided by the embodiment of the present disclosure, where the false transaction object includes at least one of a bill user, a bill commodity, and a false transaction order; the above S104 may be specifically executed by S41, S42, or S43 described below:
s41, the server detects whether the target order is a false transaction order according to the target model.
For example, the server may detect whether a trade corresponding to an order is a false trade order according to a trade detection model learned from a user isomorphic diagram and a commodity isomorphic diagram.
S42, the server detects whether the buyer of the target order is a bill-refreshing buyer according to the target model.
For example, the server may detect whether a buyer in an order is a swipe buyer based on a user detection model learned from a user isomorphic diagram.
S43, the server detects whether the commodity of the target order is a bill-refreshing commodity according to the target model.
For example, the server may detect whether the commodity in one order is a bill-of-use commodity according to the commodity obtained by learning the commodity isomorphic diagram.
It should be noted that, the foregoing exemplary description is only performed by using the dummy transaction objects as the bill-refreshing user, the bill-refreshing commodity and the dummy transaction order, and in practical application, more types of dummy transaction objects may be determined, for example, a bill-refreshing comment, a bill-refreshing merchant and the like.
Based on the scheme, the server can execute different classification tasks according to different classification models, for example, the server can classify whether an order is a normal order or a false transaction order, classify whether a buyer is a swipe buyer or a normal buyer, and classify whether a commodity is a normal commodity or a swipe commodity.
Illustratively, fig. 5 is a schematic diagram of a multi-task learning framework provided by an embodiment of the present disclosure, as shown in the figure, including graph-splitting, pruning, similarity-based neighbor feature aggregation, model training, user classification, transaction classification, and commodity classification. Specifically, the server firstly divides the initial heterograms to obtain isomorphic diagrams of commodities and isomorphic diagrams of users; secondly, pruning can be carried out on each isomorphic graph according to the connection value; then, carrying out neighbor feature aggregation based on the similarity according to the isomorphic diagram of the pruned commodity and the isomorphic diagram of the user, and sampling neighbor nodes according to probability according to the aggregated similarity to obtain neighbor features; and then carrying out average aggregation (Mean aggregation) according to the neighbor features and carrying out graph neural network training by adopting graph SAGE to obtain graph subedding of the user and graph subedding of the commodity. Inputting graph enabling of the user into a softmax layer for training to obtain a user classification model; inputting graph enabling of the commodity into a softmax layer for training to obtain a commodity classification model; and inputting the graph enabling of the user and the graph enabling of the commodity into an MLP model for training and transaction classification model. And finally, the server can respectively detect whether the user, the commodity and the transaction in the order are false transaction objects according to the three models obtained through training.
The following is data of model training of isomorphic diagrams and model training of related methods provided by embodiments of the present disclosure, and test results of performance at different angles of the present disclosure.
In the disclosed embodiment, a commonly used graph dataset YelpNYC is selected for testing, wherein 2% of the dataset is a training set, 2% of the dataset is a validation set, and 96% of the dataset is a test set. The number of edges, the memory occupied, and the events consumed by training the graph model for the original iso-graph and the pruned isograph in the disclosed embodiments are compared, respectively. The results are shown in table 1, and it can be seen that, compared with the traditional method of directly adopting the heterogeneous graph to perform model training, the number of edges is reduced by more than half, the memory occupation is also reduced by more than half, and the graph model training time is shortened by nearly half through the graph pruning strategy of the present disclosure.
TABLE 1
Method Number of edges Memory occupancy Graph model training time
Correlation method 49272129 1940MB 40 minutes
Methods of the present disclosure 24142907 836MB 21 minutes
In the embodiment of the disclosure, the performance of the embodiment of the disclosure on user classification, commodity classification and order classification is tested by testing for a total of 4 test indexes of AUC, AP, ACC and F1 of the classification task. The test results are shown in Table 2, the identification results are relatively high, and the accuracy is over 90%. The AUC value is a probability value, a positive sample and a negative sample are randomly selected, and the probability that the positive sample is arranged in front of the negative sample according to the calculated Score value is AUC in the current classification algorithm. The closer the AUC is to 1, the better the classifier performance. ACC represents the proportion of the identified correct number to the total number. AP represents average accuracy (Average Precision), and F1 represents harmonic averaging of precision and recall.
TABLE 2
Node AUC AP ACC F1
User' s 0.9281 0.8049 0.9027 0.7764
Goods commodity 0.8845 0.7206 0.9043 0.7176
Order form 0.9732 0.8803 0.9567 0.8119
It should be noted that, in the method for detecting a false transaction object provided by the embodiment of the present disclosure, the execution body may be a detection device of the false transaction object, or a control module of the method for executing the detection of the false transaction object in the detection device of the false transaction object. In the embodiment of the present disclosure, a method for executing detection of a false transaction object by using a detection device of a false transaction object is taken as an example, and the detection device of the false transaction object provided by the embodiment of the present disclosure is described.
Fig. 6 is a schematic structural diagram of a device for detecting a false transaction object according to the disclosed embodiment, and as shown in fig. 6, the device 600 for detecting a false transaction object includes: a determining module 601, a clipping module 602, a model training module 603 and a detecting module 604; a determining module 601, configured to determine a connection value of an edge of each isomorphic diagram in M first isomorphic diagrams, where the M first isomorphic diagrams include at least one of an isomorphic diagram of a user and an isomorphic diagram of a commodity, and M is a positive integer; a clipping module 602, configured to clip edges with a connection value determined by the determining module 601 in the M first isomorphic graphs being less than or equal to a preset connection value, to obtain M second isomorphic graphs; the model training module 603 is configured to perform graph training according to the neighbor features of the neighbor nodes in the M second isomorphic graphs obtained by clipping by the clipping module 602, so as to obtain a target model; the detection module 604 is configured to detect whether a false transaction object exists in a target order according to the target model trained by the model training module 603, where the target order is any order.
Optionally, the detection device of the false transaction object further includes: an acquisition module and a sampling module; the acquisition module is further used for acquiring target characteristics of each node of each isomorphic graph in the M second isomorphic graphs before the model training module performs graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs; the determining module is also used for determining the target similarity of each node and the neighbor node according to the target characteristics acquired by the acquiring module; and the sampling module is used for sampling the neighbor nodes according to the probability according to the target similarity determined by the determining module to obtain neighbor characteristics.
Optionally, the detection device of the false transaction object further includes: splitting the module; the splitting module is used for splitting the target iso-graph into M first co-graphs before the determining module determines the connection value of the edges of each isograph in the M first co-graphs, wherein the target iso-graph is a user-order-commodity iso-graph, and the M first co-graphs comprise at least one of a user isograph and a commodity isograph.
Optionally, the false transaction object includes at least one of a swiped user, a swiped merchandise, and a false transaction order; the detection module is specifically used for: detecting whether the target order is a false transaction order; or detecting whether the buyer of the target order is a swipe order buyer; or detecting whether the commodity of the target order is a bill-of-use commodity.
Optionally, the determining module is specifically configured to: determining the connection heat of each node in the M first concurrent patterns; determining the connection value of the edges of each isomorphic graph in the M first isomorphic graphs according to the connection heat; the connection heat of the nodes in the isomorphic diagram of the user is related to the purchase quantity of the user, and the connection heat of the nodes in the isomorphic diagram of the commodity is related to the sales quantity of the commodity.
Optionally, the determining module is specifically configured to: determining the connection heat of the nodes of each isomorphic diagram of the commodity according to a first preset formula and the sales of the commodity; the first preset formula is:wherein vp is i Representing the connection heat of the nodes of the commodity i, S i Indicating the sales of commodity i.
Optionally, the determining module is specifically configured to: determining the connection heat of the nodes in the isomorphic diagram of the user according to a second preset formula and the purchase quantity of the user; the second preset formula is:wherein vu is j Connection of nodes representing user jHeat degree, p j Representing the purchase amount of user j.
Optionally, the determining module is specifically configured to: determining the connection value of edges in the isomorphic diagram of the commodity according to a third preset formula and the connection heat of each node; the third preset formula is:wherein EU is ab Representing the connection value of the edge between user a and user b, p ab Representing the collection of items purchased by both user a and user b, p a Representing the collection of items purchased by user a, p b Representing the collection of items purchased by user b.
Optionally, the determining module is specifically configured to: determining the connection value of the edges in the isomorphic graph of the user according to the connection heat of each node of the fourth preset formula; the fourth preset formula is:wherein EP xy Representing the value of the connection of the edges between commodity x and commodity y, P xy Representing a set of users who have purchased both commodity x and commodity y, U x A user collection representing purchased goods x; u (U) y A user set representing purchased goods y; u (U) xy Representing the collection of users who purchased commodity x and commodity y.
Optionally, the target features include attribute features and structural features; the sampling module is specifically used for: and sampling the neighbor nodes according to the probability according to the weighted sum of the attribute similarity and the structure similarity of the nodes to obtain neighbor features.
Optionally, the determining module is specifically configured to: determining attribute similarity of each node and the neighbor node according to the attribute characteristics of the nodes and a fifth preset formula; determining the structural similarity of each node and the neighbor node according to the structural characteristics of the nodes and a sixth preset formula; determining the target similarity of each node and the neighbor node according to the attribute similarity, the structure similarity and a seventh preset formula; the fifth preset formula is: The sixth preset formula is:the seventh preset formula is: />Wherein (1)>Representing the similarity of the attributes of node i and node j, < >>Attribute feature representing node i, < >>Representing the attribute characteristics of node j; />Representing the structural similarity of node i and node j, < >>Representing the structural features of node i->Representing the structural characteristics of the node j; s is S ij Representing the target similarity of node i and node j, ω representing the weight factor.
The embodiment of the disclosure provides a detection device for a false transaction object, firstly, the detection device for the false transaction object can firstly determine the connection value of the edge of each isomorphic graph in M first isomorphic graphs, and then the detection device for the false transaction object cuts the edge of which the connection value is smaller than or equal to a preset connection value in the M first isomorphic graphs to obtain M second isomorphic graphs; then, the detection device of the false transaction object can conduct graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs to obtain a target model; finally, the detection device of the false transaction object can determine whether any order has the false transaction object according to the target model. Because the M first co-graphs include at least one of a user isomorphic graph of the user and a user graph of the commodity, M is a positive integer, the data size of the embodiment of the disclosure when model training is first smaller than the data size of model training using a user-order-commodity isomorphic graph in the related art. Secondly, edges with connection values smaller than or equal to preset connection values in the M first synchronous patterns are cut off, so that on one hand, the processing amount of data in model training can be reduced again, and the rate of model training is improved; on the other hand, deleting some noise data can improve the accuracy of the trained target model, can enable the accuracy of determining the false transaction to be higher, for example, can distinguish whether a user in an order is a bill-refreshing user or a common user more accurately, distinguish whether goods in an order are bill-refreshing goods or common goods more accurately, and distinguish whether an order is a false transaction order more accurately.
The detection device of the false transaction object in the embodiment of the disclosure may be a device, or may be a component, an integrated circuit, or a chip in a terminal. The device may be a mobile electronic device or a non-mobile electronic device. By way of example, the mobile electronic device may be a cell phone, tablet computer, notebook computer, palm computer, vehicle-mounted electronic device, wearable device, UMPC (ultra-mobile personal computer, ultra mobile personal computer), netbook or PDA (personal digital assistant ), etc., and the non-mobile electronic device may be a server, NAS (NetworkAttached Storage ), PC (personal computer, personal computer), TV (television), teller machine or self-service machine, etc., and the embodiments of the present disclosure are not limited in particular.
The detection device of the false transaction object in the embodiment of the disclosure may be a device having an operating system. The operating system may be an Android operating system, an iOS operating system, or other possible operating systems, and the embodiments of the present disclosure are not limited specifically.
The detection device for a false transaction object provided in the embodiments of the present disclosure can implement each process implemented by the embodiments of the method in fig. 1 to 5, and in order to avoid repetition, a description is omitted here.
Optionally, as shown in fig. 7, the embodiment of the present disclosure further provides a server 700, including a processor 701, a memory 702, and a program or an instruction stored in the memory 702 and capable of running on the processor 701, where the program or the instruction implements each process of the above embodiment of the method for detecting a false transaction object when executed by the processor 701, and the same technical effects can be achieved, and for avoiding repetition, a detailed description is omitted herein.
It should be noted that, the electronic device in the embodiment of the disclosure includes the mobile electronic device and the non-mobile electronic device described above.
It should be noted that the server 800 shown in fig. 8 is only an example, and should not impose any limitation on the functions and the scope of use of the embodiments of the present disclosure.
As shown in fig. 8, the server 800 includes a central processing unit (Central Processing Unit, CPU) 801 that can perform various appropriate actions and processes according to a program stored in a ROM (Read only memory) 802 or a program loaded from a storage section 808 into a RAM (RandomAccess Memory ) 803. In the RAM 803, various programs and data required for system operation are also stored. The CPU 801, ROM 802, and RAM 803 are connected to each other by a bus 804. An I/O (Input/Output) interface 805 is also connected to bus 804.
The following components are connected to the I/O interface 805: an input portion 806 including a keyboard, mouse, etc.; an output portion 807 including a CRT (Cathode ray tube), an LCD (Liquid Crystal Display ), and the like, and a speaker, and the like; a storage section 808 including a hard disk or the like; and a communication section 809 including a network interface card such as a LAN (local area network) card, a modem, or the like. The communication section 809 performs communication processing via a network such as the internet. The drive 810 is also connected to the I/O interface 805 as needed. A removable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 810 as needed so that a computer program read out therefrom is mounted into the storage section 808 as needed.
In particular, according to embodiments of the present disclosure, the processes described below with reference to flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method shown in the flowcharts. In such an embodiment, the computer program may be downloaded and installed from a network via the communication section 809, and/or installed from the removable media 811. When executed by a central processing unit (CPU 801), performs the various functions defined in the system of the present application.
The embodiment of the disclosure provides a server, firstly, the server can determine the connection value of the edge of each isomorphic graph in M first isomorphic graphs, and then the server cuts the edge with the connection value smaller than or equal to the preset connection value in the M first isomorphic graphs to obtain M second isomorphic graphs; then, the server can conduct graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs to obtain a target model; finally, the server can determine whether any order has false transaction objects according to the target model. Because the M first co-graphs include at least one of a user isomorphic graph of the user and a user graph of the commodity, M is a positive integer, the data size of the embodiment of the disclosure when model training is first smaller than the data size of model training using a user-order-commodity isomorphic graph in the related art. Secondly, edges with connection values smaller than or equal to preset connection values in the M first synchronous patterns are cut off, so that on one hand, the processing amount of data in model training can be reduced again, and the rate of model training is improved; on the other hand, deleting some noise data can improve the accuracy of the trained target model, can enable the accuracy of determining the false transaction to be higher, can more accurately distinguish whether a user in an order is a bill-refreshing user or a common user, more accurately distinguish whether goods in the order are bill-refreshing goods or common goods, and more accurately distinguish whether the order is a false transaction order.
The embodiment of the disclosure further provides a readable storage medium, where a program or an instruction is stored on the readable storage medium, and when the program or the instruction is executed by a processor, the processes of the above-mentioned method embodiment for detecting a false transaction object are implemented, and the same technical effects can be achieved, so that repetition is avoided, and no description is repeated here.
Wherein the processor is a processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium such as a Read-Only Memory (ROM), a random access Memory (RandomAccess Memory, RAM), a magnetic disk or an optical disk, and the like.
The embodiment of the disclosure further provides a chip, where the chip includes a processor and a communication interface, where the communication interface is coupled to the processor, and the processor is configured to run a program or an instruction, implement each process of the above-mentioned method embodiment for detecting a false transaction object, and achieve the same technical effect, so that repetition is avoided, and no further description is given here.
It should be understood that the chips referred to in the embodiments of the present disclosure may also be referred to as system-on-chip chips, chip systems, or system-on-chip chips, etc.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element. Furthermore, it should be noted that the scope of the methods and apparatus in the embodiments of the present disclosure is not limited to performing the functions in the order shown or discussed, but may also include performing the functions in a substantially simultaneous manner or in an opposite order depending on the functions involved, e.g., the described methods may be performed in an order different from that described, and various steps may also be added, omitted, or combined. Additionally, features described with reference to certain examples may be combined in other examples.
From the above description of the embodiments, it will be clear to those skilled in the art that the above-described embodiment method may be implemented by means of software plus a necessary general hardware platform, but of course may also be implemented by means of hardware, but in many cases the former is a preferred embodiment. Based on such understanding, the technical solution of the present disclosure may be embodied essentially or in a part contributing to the prior art in the form of a software product stored in a storage medium (e.g., ROM/RAM, magnetic disk, optical disk), including several instructions for causing a terminal (which may be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to perform the method described in the embodiments of the present disclosure.
The embodiments of the present disclosure have been described above with reference to the accompanying drawings, but the present disclosure is not limited to the above-described embodiments, which are merely illustrative and not restrictive, and many forms may be made by those of ordinary skill in the art without departing from the spirit of the disclosure and the scope of the claims, which are all within the protection of the present disclosure.

Claims (14)

1. A method for detecting a false transaction object, the method comprising:
determining the connection value of the edges of each isomorphic diagram in M first isomorphic diagrams, wherein the M first isomorphic diagrams comprise at least one isomorphic diagrams of users and isomorphic diagrams of commodities, and M is a positive integer;
cutting edges with connection value smaller than or equal to preset connection value in the M first isomorphic graphs to obtain M second isomorphic graphs;
performing graph training according to neighbor characteristics of neighbor nodes in the M second isomorphic graphs to obtain a target model;
and detecting whether a false transaction object exists in the target order according to the target model, wherein the target order is any order.
2. The method of claim 1, wherein prior to the graph training based on the neighbor characteristics of the neighbor nodes in the M second isomorphic graphs, the method further comprises:
Acquiring target characteristics of each node of each isomorphic diagram in the M second isomorphic diagrams;
determining target similarity of each node and the neighbor node according to the target characteristics;
and sampling the neighbor nodes according to the probability according to the target similarity to obtain neighbor characteristics.
3. The method of claim 1, wherein prior to determining the connection value of the edges of each isomorphic graph in the M first isomorphic graphs, the method further comprises:
splitting the target iso-composition into the M first co-compositions, wherein the target iso-composition is a user-order-commodity iso-composition, and the M first co-compositions comprise at least one of a user isomorphic chart and a commodity isomorphic chart.
4. The method of claim 1, wherein the spurious transaction object comprises at least one of a swipe user, a swipe merchandise, and a spurious transaction order; the detecting whether the target order has false transaction objects comprises the following steps:
detecting whether the target order is a false trade order; or alternatively, the process may be performed,
detecting whether the buyer of the target order is a swipe order buyer; or alternatively, the process may be performed,
and detecting whether the commodity of the target order is a bill-by-bill commodity.
5. The method of claim 1, wherein determining the connection value of the edges of each isomorphic graph in the M first isomorphic graphs comprises:
Determining the connection heat of the nodes of each isomorphic diagram in the M first isomorphic diagrams;
determining the connection value of the edges of each isomorphic graph in the M first isomorphic graphs according to the connection heat;
the connection heat of the nodes in the isomorphic diagram of the user is related to the purchase quantity of the user, and the connection heat of the nodes in the isomorphic diagram of the commodity is related to the sales quantity of the commodity.
6. The method of claim 5, wherein determining the connection hotness of the nodes of each isomorphic graph in the M first isomorphic graphs comprises:
determining the connection heat of nodes in the isomorphic diagram of the commodity according to a first preset formula and the sales of the commodity;
the first preset formula is:
wherein vp is i Representing the connection heat of the nodes of the commodity i, S i Indicating the sales of commodity i.
7. The method of claim 5, wherein determining the connection hotness of the nodes of each isomorphic graph in the M first isomorphic graphs comprises:
determining the connection heat of nodes in the isomorphic diagram of the user according to a second preset formula and the purchase quantity of the user;
the second preset formula is:
wherein vu is j Representing the connection heat of the nodes of user j, p j Representing the purchase amount of user j.
8. The method of claim 6, wherein determining the connection value of the edges of each isomorphic graph in the M first isomorphic graphs based on the connection heat comprises:
determining the connection value of edges in the isomorphic diagram of the commodity according to a third preset formula and the connection heat of each node; or alternatively, the process may be performed,
the third preset formula is:
wherein EU is ab Representing the connection value of the edge between user a and user b, p ab Representing the collection of items purchased by both user a and user b, p a Representing the collection of items purchased by user a, p b Representing the collection of items purchased by user b.
9. The method of claim 6, wherein determining the connection value of the edges of each isomorphic graph in the M first isomorphic graphs based on the connection heat comprises:
determining the connection value of the edges in the isomorphic graph of the user according to the connection heat of each node of the fourth preset formula;
the fourth preset formula is:
wherein EP xy Representing the value of the connection of the edges between commodity x and commodity y, P xy Representing a set of users who have purchased both commodity x and commodity y, U x A user collection representing purchased goods x; u (U) y A user set representing purchased goods y; u (U) xy Representing the collection of users who purchased commodity x and commodity y.
10. The method of claim 2, wherein the target features include attribute features and structural features; and sampling the neighbor nodes according to the target similarity according to probability to obtain neighbor characteristics, wherein the method comprises the following steps:
and sampling the neighbor nodes according to the probability according to the weighted sum of the attribute similarity and the structure similarity of the nodes to obtain neighbor features.
11. The method of claim 10, wherein said determining the target similarity for each node and neighbor nodes based on the target features comprises:
determining attribute similarity of each node and the neighbor node according to the attribute characteristics of the nodes and a fifth preset formula;
determining the structural similarity of each node and the neighbor node according to the structural characteristics of the nodes and a sixth preset formula;
determining the target similarity of each node and the neighbor node according to the attribute similarity, the structure similarity and a seventh preset formula;
the fifth preset formula is:
the sixth preset formula is:
the seventh preset formula is:
wherein, the liquid crystal display device comprises a liquid crystal display device,representing the similarity of the attributes of node i and node j, < >>Attribute feature representing node i, < >>Representing the attribute characteristics of node j; />Representing the structural similarity of node i and node j, < > >Representing the structural features of node i->Representing the structural characteristics of the node j; s is S ij Representing the target similarity of node i and node j, ω representing the weight factor.
12. A device for detecting a false transaction object, the device comprising: the system comprises a determining module, a cutting module, a model training module and a detecting module;
the determining module is used for determining the connection value of the edges of each isomorphic diagram in M first isomorphic diagrams, wherein the M first isomorphic diagrams comprise at least one isomorphic diagrams of users and isomorphic diagrams of commodities, and M is a positive integer;
the clipping module is configured to clip, according to the connection value determined by the determining module, an edge of each isomorphic graph in the M first isomorphic graphs, where the connection value of each isomorphic graph is less than or equal to a preset connection value, to obtain M second isomorphic graphs;
the model training module is used for carrying out graph training according to the neighbor characteristics of the neighbor nodes in the M second isomorphic graphs obtained by the cutting module, so as to obtain a target model;
and the detection module is used for detecting whether a false transaction object exists in the target order according to the target model obtained through training by the model training module.
13. A server comprising a processor, a memory and a program or instruction stored on the memory and executable on the processor, which when executed by the processor, implements the steps of the method of detecting a false transaction object according to any one of claims 1 to 11.
14. A readable storage medium, wherein a program or instructions is stored on the readable storage medium, which when executed by a processor, implements the steps of the method for detecting a false transaction object according to any of claims 1 to 11.
CN202210005494.7A 2022-01-05 2022-01-05 False transaction object detection method, false transaction object detection device and server Pending CN116451128A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210005494.7A CN116451128A (en) 2022-01-05 2022-01-05 False transaction object detection method, false transaction object detection device and server

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210005494.7A CN116451128A (en) 2022-01-05 2022-01-05 False transaction object detection method, false transaction object detection device and server

Publications (1)

Publication Number Publication Date
CN116451128A true CN116451128A (en) 2023-07-18

Family

ID=87134245

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210005494.7A Pending CN116451128A (en) 2022-01-05 2022-01-05 False transaction object detection method, false transaction object detection device and server

Country Status (1)

Country Link
CN (1) CN116451128A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116629984A (en) * 2023-07-24 2023-08-22 中信证券股份有限公司 Product information recommendation method, device, equipment and medium based on embedded model

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116629984A (en) * 2023-07-24 2023-08-22 中信证券股份有限公司 Product information recommendation method, device, equipment and medium based on embedded model
CN116629984B (en) * 2023-07-24 2024-02-06 中信证券股份有限公司 Product information recommendation method, device, equipment and medium based on embedded model

Similar Documents

Publication Publication Date Title
CN105184574B (en) A kind of detection method for applying mechanically trade company&#39;s classification code fraud
US10902484B1 (en) System and method for carbon footprint determination
US8025220B2 (en) Cardholder localization based on transaction data
US10204349B2 (en) Analyzing customer segments
US10521856B1 (en) Categorizing financial transactions based on spending patterns
US9542710B1 (en) Categorizing financial transactions based on business preferences
CN111401777A (en) Enterprise risk assessment method and device, terminal equipment and storage medium
WO2010141270A2 (en) Systems and methods to summarize transaction data
US20100010870A1 (en) System and Method for Tuning Demand Coefficients
CN107578270A (en) A kind of construction method, device and the computing device of financial label
WO2009014753A2 (en) Method for scoring products, services, institutions, and other items
US20140229233A1 (en) Consumer spending forecast system and method
CN110659961A (en) Method and device for identifying off-line commercial tenant
CN110942392A (en) Service data processing method, device, equipment and medium
WO2021202151A1 (en) Systems and methods for advanced velocity profile preparation and analysis
CN107133862A (en) Dynamic produces the method and system of the detailed transaction payment experience of enhancing credit evaluation
CN112529575A (en) Risk early warning method, equipment, storage medium and device
CN115375177A (en) User value evaluation method and device, electronic equipment and storage medium
CN116451128A (en) False transaction object detection method, false transaction object detection device and server
CN105303447A (en) Method and device for carrying out credit rating through network information
WO2019221844A1 (en) Item-specific value optimization tool
JP2021197089A (en) Output device, output method, and output program
CN112464051A (en) Abnormal data detection method and device and computer readable storage medium
US20230100685A1 (en) Financial product information collecting platform system, financial product information collecting method, and computer program for the same
CN112581291B (en) Risk assessment change detection method, apparatus, device and storage medium

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination