CN118115167A - Electronic commerce brush bank behavior detecting system based on big data - Google Patents

Abstract

The invention relates to an electronic commerce bank behavior detection system based on big data, which comprises a transaction detection cloud platform, a related financial platform and a third party payment platform. The transaction detection cloud platform comprises a network construction module, a vertex analysis module, a transaction ring identification module and a behavior identification module. The transaction detection cloud platform performs fusion processing on all commodity transaction subgraphs generated based on the transaction data of the third-party payment platform to generate corresponding commodity transaction graphs; identifying first user vertex sets for commodity transaction with corresponding target user vertices according to the triplet data of the commodity transaction graph, and determining second user vertex sets of each first user vertex; determining a target financial transaction loop taking the target user vertex as a transaction behavior starting point and a transaction behavior ending point based on the first user vertex set and the second user vertex set; a determination is made as to whether the transaction behavior of the corresponding transaction user is a billing behavior based on the historical transaction data sets of the target financial transaction loop and the user vertices.

Description

Electronic commerce brush bank behavior detecting system based on big data

Technical Field

The invention relates to the field of electronic commerce and big data, in particular to an electronic commerce business trip behavior detection system based on big data.

Background

With the continuous development of social economy and internet technology, online shopping becomes the first choice of more and more consumers, and electronic commerce becomes a novel business model which exists by relying on the internet. The e-commerce platform for providing online shopping for consumers is not enumerated, wherein the treasured washing network is the e-commerce platform with the greatest influence and has become the largest Asian. With the continuous development of the Taobao net, new problems are also caused. Sales and benefits are made by "brushing" on the panzer (also known as "credit stir-ups" for short). The business income is improved to a certain extent by the bill swiping action of the e-commerce platform, but the popularization cost of the e-commerce platform is improved by the bill swiping action, so that the serious credit safety problem is caused; false bill-of-use information, on the other hand, renders the consumer susceptible to misleading and thus property damage. Therefore, how to detect the e-commerce transaction is a current urgent problem.

Disclosure of Invention

Aiming at the defects of the prior art, the invention provides an electronic commerce bank behavior detection system based on big data, which comprises a transaction detection cloud platform, a related financial platform and a third party payment platform; the transaction detection cloud platform is respectively in communication connection with the relevant financial platform and the third party payment platform, and the relevant financial platform and the third party payment platform are in communication connection;

The transaction detection cloud platform comprises a network construction module, a vertex analysis module, a transaction ring identification module and a behavior identification module;

The network construction module constructs corresponding commodity transaction subgraphs for users of transaction parties in a relevant financial platform for commodity transaction in a preset transaction period based on transaction data of a third party payment platform, and performs fusion processing on all commodity transaction subgraphs based on user information of a second transaction party for transaction with each first transaction party to generate corresponding commodity transaction graphs, wherein the first transaction party is a commodity consumer in the relevant financial platform, and the second transaction party is a commodity provider in the relevant financial platform;

The vertex analysis module represents the commodity transaction graph in the form of triplet data, and identifies a first user vertex set for commodity transaction with a corresponding target user vertex at a preset time in the commodity transaction graph according to the triplet data, wherein the triplet data comprises a first transaction party, a second transaction party and a transaction time stamp;

The transaction ring identification module acquires all second user vertexes which carry out commodity transaction with each first user vertex in the first user vertex set before the preset moment in the commodity transaction diagram, so as to generate a corresponding second user vertex set for each first user vertex in the first user vertex set, and determines a target financial transaction ring taking a target user vertex as a transaction behavior starting point and a transaction behavior ending point based on the first user vertex set and the second user vertex set;

The behavior recognition module determines whether a transaction behavior related to the target financial transaction loop is a bill-swiping behavior based on structural features of the target financial transaction loop and historical transaction data sets of respective user vertices in the target financial transaction loop.

According to a preferred embodiment, performing fusion processing on all commodity transaction sub-graphs based on user information of a second transaction party transacting with each first transaction party to generate corresponding commodity transaction graphs includes:

Identifying neighbor user vertexes shared between the corresponding first transaction parties based on user information of the second transaction party transacting with each first transaction party, and performing splicing processing on all commodity transaction subgraphs based on attribute information of related edges existing between the corresponding first transaction party and the neighbor user vertexes thereof to generate corresponding commodity transaction graphs, wherein the attribute information of the related edges is used for representing transaction time stamps between the corresponding first transaction party and the neighbor user vertexes thereof.

According to a preferred embodiment, determining whether a characteristic deviation exists between a current transaction behavior of a first transaction party in the target financial transaction loop and a historical transaction behavior of the first transaction party in the target financial transaction loop according to a historical transaction data set corresponding to a vertex of a user in the target financial transaction loop comprises:

Mapping the multidimensional data feature vector corresponding to each historical transaction data in the historical transaction data set into a plurality of feature subspaces with equal size to obtain feature vectors of the historical transaction data in the feature subspaces with different dimensions, and determining the space coordinate points of the corresponding historical transaction data in the feature subspaces with different dimensions according to the feature vectors;

Counting the space coordinate points in each feature subspace to obtain the feature density of the corresponding feature subspace, acquiring the feature density of each feature subspace in different time windows to determine the density increment of each feature subspace in different time windows, and predicting the density increment of each feature subspace in a preset time period by using poisson distribution;

Fusing the density increment of each characteristic subspace in different time windows and the predicted density increment to obtain the characteristic weight of each characteristic subspace, and constructing a corresponding weight matrix according to the characteristic weights of the historical transaction data sets in different characteristic subspaces;

and determining whether characteristic deviation exists between the current transaction behavior of the first transaction party and the historical transaction behavior of the first transaction party according to the weight matrix and the transaction data generated by the corresponding first transaction party in a preset transaction period.

According to a preferred embodiment, the step of determining whether the transaction activity associated with the financial transaction loop is a billing activity comprises:

Analyzing and obtaining sequence features of vertex sequences of the target financial transaction ring based on structural features of the target financial transaction ring and transaction time stamps between vertices of each user in the target financial transaction ring, and determining whether the sequence features are matched with sequence features of abnormal transaction behaviors;

if so, judging whether the characteristic deviation exists between the current transaction behavior of the first transaction party in the target financial transaction ring and the historical transaction behavior of the first transaction party in the target financial transaction ring according to the historical transaction data set of the corresponding user vertex in the target financial transaction ring; if yes, determining that the transaction behavior related to the financial transaction loop is a bill swiping behavior.

According to a preferred embodiment, the determining whether the characteristic deviation exists between the current transaction behavior of the first transaction party and the historical transaction behavior of the first transaction party according to the weight matrix and the transaction data generated by the corresponding first transaction party in the preset transaction period includes:

Weighting and fusing the feature vectors of each historical transaction data in different feature subspaces according to the weight matrix to obtain a first behavior feature corresponding to the first transaction party, wherein the first behavior feature is used for representing the behavior feature of the historical transaction behavior of the first transaction party;

Mapping transaction data generated by a first transaction party in a preset transaction period into a plurality of feature subspaces with equal size, so as to obtain feature vectors of the transaction data in the feature subspaces with different dimensions, and carrying out weighted fusion on the feature vectors in the different feature subspaces according to the weight matrix, so as to obtain second behavior features corresponding to the first transaction party, wherein the second behavior features are used for representing the behavior features of the current transaction behavior of the first transaction party;

Comparing the first behavioral characteristics with the second behavioral characteristics to determine whether the current transaction behavior of the first transaction party has a characteristic deviation from its historical transaction behavior.

According to a preferred embodiment, determining a target financial transaction loop with a target user vertex as a transaction action starting point and a transaction action ending point based on the first set of user vertices and the second set of user vertices comprises:

Detecting all first financial transaction loops which exist in the first user vertex set and the second user vertex set and take corresponding first user vertices as transaction behavior starting points and transaction behavior ending points, and second financial transaction loops which take corresponding second user vertices as transaction behavior starting points and transaction behavior ending points;

Comparing all target financial transaction loops corresponding to the target user vertexes with all first financial transaction loops corresponding to each first user vertex and all second financial transaction loops corresponding to each second user vertex, and judging whether each target financial transaction loop of the target user vertexes has an overlapping area with loop structures of the corresponding first financial transaction loops and the corresponding second financial transaction loops;

if not, adding the vertex of the target user into the loop structures of the corresponding first financial transaction loop and the second financial transaction loop to generate a new target financial transaction loop for the vertex of the target user;

If so, adding loop structures of a first financial transaction loop and/or a second financial transaction loop with overlapping areas corresponding to the vertexes of the target users so that the length of the first financial transaction loop and/or the second financial transaction loop is increased by 1, and updating the corresponding target financial transaction loop by using the increased first financial transaction loop and/or second financial transaction loop.

According to a preferred embodiment, the historical transaction data includes transaction time, transaction amount, transaction party information, transaction merchandise, and order number.

According to a preferred embodiment, the sequence of features of the vertex sequence is used to indicate the flow features, duration period and number of users involved of the corresponding financial transaction behaviour.

According to a preferred embodiment, determining a target financial transaction loop with a target user vertex as a transaction action starting point and a transaction action ending point based on the first set of user vertices and the second set of user vertices further comprises:

Searching user vertexes in a corresponding path of a commodity transaction diagram by taking a target user vertex as a transaction behavior starting point, taking the searched loop path as a target financial transaction loop of the target user vertex when the current searching length is smaller than a preset loop length and the transaction time stamp of the user vertex corresponding to the current searching length is smaller than a preset loop closing time, and updating the loop closing time of the next target user vertex;

And stopping searching the user vertexes in the corresponding paths when the current searching length is greater than or equal to the preset loop length or the transaction time stamp of the user vertexes corresponding to the current searching length is greater than or equal to the preset loop closing time.

The invention has the following beneficial effects:

By analyzing the transaction data of the third-party payment platform, the invention identifies the financial transaction loop taking the corresponding transaction user as the transaction action starting point and the transaction action ending point, wherein the financial transaction loop represents that the transaction funds finally flow back to the hand of the transaction user in the transaction loop taking the corresponding transaction user as the transaction action starting point. Therefore, the method and the system avoid the arrangement and the identification of a large amount of redundant transaction data by starting from key features of the bill-refreshing behavior, can more efficiently and accurately identify the bill-refreshing behavior in the network, and provide a good online shopping space for consumers so as to ensure the shopping safety of the consumers.

Drawings

FIG. 1 is a block diagram of a system for detecting a large data-based e-commerce transaction.

Detailed Description

This invention is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the invention and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the invention being indicated by the following claims.

As shown in fig. 1, in one embodiment, the big data based e-commerce facilitation of the detection system comprises a transaction detection cloud platform, a related financial platform, and a third party payment platform. The transaction detection cloud platform is respectively in communication connection with the relevant financial platform and the third party payment platform, and the relevant financial platform and the third party payment platform are in communication connection. Related financial platforms include electronic commerce platforms, take-away platforms, and other platforms that conduct transactions that can be used with merchandise.

The transaction detection cloud platform comprises a network construction module, a vertex analysis module, a transaction ring identification module and a behavior identification module.

The network construction module is used for constructing corresponding commodity transaction subgraphs for users of transaction parties in a relevant financial platform for carrying out commodity transaction in a preset transaction period according to transaction data of a third party payment platform, and carrying out fusion processing on all commodity transaction subgraphs based on user information of a second transaction party for carrying out transaction with each first transaction party to generate corresponding commodity transaction graphs, wherein the first transaction party is a commodity consumer in the relevant financial platform, and the second transaction party is a commodity provider in the relevant financial platform;

the vertex analysis module is used for representing the commodity transaction graph in the form of triplet data, and identifying a first user vertex set for commodity transaction with a corresponding target user vertex at a preset time in the commodity transaction graph according to the triplet data, wherein the triplet data comprises a first transaction party, a second transaction party and a transaction time stamp;

The transaction ring identification module is used for acquiring all second user vertexes which are used for carrying out commodity transaction with each first user vertex in the first user vertex sets before the preset moment in the commodity transaction diagram, so as to generate a corresponding second user vertex set for each first user vertex in the first user vertex sets, and determining a target financial transaction ring taking a target user vertex as a transaction behavior starting point and a transaction behavior ending point based on the first user vertex set and the second user vertex set;

The behavior recognition module is used for determining whether the transaction behavior related to the target financial transaction loop is a bill-refreshing behavior according to the structural characteristics of the target financial transaction loop and the historical transaction data set of each user vertex in the target financial transaction loop.

Specifically, the working method flow of the e-commerce business trip behavior detection system based on big data comprises the following steps:

S1, a network construction module constructs corresponding commodity transaction subgraphs for transaction users in related financial platforms for carrying out commodity transaction in a preset transaction period based on transaction data of a third party payment platform, and fusion processing is carried out on all commodity transaction subgraphs based on user information of second transaction parties for carrying out transaction with each first transaction party to generate corresponding commodity transaction graphs, wherein the first transaction party is a commodity consumer in the related financial platform, and the second transaction party is a commodity provider in the related financial platform.

Optionally, the preset transaction period is a preset time period of the system, which may be set to one day, three days or one week, that is, the abnormality detection false detection rate of the commodity transaction exceeding a certain time period is higher.

Specifically, performing fusion processing on all commodity transaction subgraphs based on user information of a second transaction party transacting with each first transaction party to generate corresponding commodity transaction graphs includes:

Identifying neighbor user vertexes shared between the corresponding first transaction parties based on user information of the second transaction party transacting with each first transaction party, and performing splicing processing on all commodity transaction subgraphs based on attribute information of related edges existing between the corresponding first transaction party and the neighbor user vertexes thereof to generate corresponding commodity transaction graphs, wherein the attribute information of the related edges is used for representing transaction time stamps between the corresponding first transaction party and the neighbor user vertexes thereof.

Optionally, the transaction data includes transaction time, transaction finance, transaction party information, transaction merchandise, and order numbers, and the transaction party information includes user information of the first transaction party and user information of the second transaction party. The transaction time stamp is a corresponding transfer time point when the users of the two transaction sides finish transfer.

And S2, the vertex analysis module represents the commodity transaction diagram in the form of triple data, and identifies a first user vertex set for commodity transaction with a corresponding target user vertex at a preset time in the commodity transaction diagram according to the triple data, wherein the triple data comprises a first transaction party, a second transaction party and a transaction time stamp.

Optionally, the preset time is a certain transaction time point preset by the system. The target user vertex can be a first transaction party (the first user vertex for commodity transaction with the target user vertex is a second transaction party) or a second transaction party (the first user vertex for commodity transaction with the target user vertex is a first transaction party); the first user vertex in the first user vertex set may be a first transaction party or a second transaction party, and vice versa.

S3, a transaction ring identification module acquires all second user vertexes which carry out commodity transaction with each first user vertex in the first user vertex sets before the preset moment in the commodity transaction diagram, so as to generate a corresponding second user vertex set for each first user vertex in the first user vertex sets, and a target financial transaction ring taking a target user vertex as a transaction behavior starting point and a transaction behavior ending point is determined based on the first user vertex set and the second user vertex set.

Optionally, in the present invention, the meaning of the financial transaction loop is characterized in the transaction loop with the corresponding user as the transfer party, and the transaction funds ultimately flow back to the user.

Specifically, determining a target financial transaction loop with a target user vertex as a transaction behavior start point and a transaction behavior end point based on the first set of user vertices and the second set of user vertices includes:

Detecting all first financial transaction loops which exist in the first user vertex set and the second user vertex set and take corresponding first user vertices as transaction behavior starting points and transaction behavior ending points, and second financial transaction loops which take corresponding second user vertices as transaction behavior starting points and transaction behavior ending points;

Comparing all target financial transaction loops corresponding to the target user vertexes with all first financial transaction loops corresponding to each first user vertex and all second financial transaction loops corresponding to each second user vertex, and judging whether each target financial transaction loop of the target user vertexes has an overlapping area with loop structures of the corresponding first financial transaction loops and the corresponding second financial transaction loops;

if not, adding the vertex of the target user into the loop structures of the corresponding first financial transaction loop and the second financial transaction loop to generate a new target financial transaction loop for the vertex of the target user;

If so, adding loop structures of a first financial transaction loop and/or a second financial transaction loop with overlapping areas corresponding to the vertexes of the target users so that the length of the first financial transaction loop and/or the second financial transaction loop is increased by 1, and updating the corresponding target financial transaction loop by using the increased first financial transaction loop and/or second financial transaction loop.

Specifically, determining, based on the first set of user vertices and the second set of user vertices, a target financial transaction loop with target user vertices as a transaction behavior start point and a transaction behavior end point further includes:

Searching user vertexes in a corresponding path of a commodity transaction diagram by taking a target user vertex as a transaction behavior starting point, taking the searched loop path as a target financial transaction loop of the target user vertex when the current searching length is smaller than a preset loop length and the transaction time stamp of the user vertex corresponding to the current searching length is smaller than a preset loop closing time, and updating the loop closing time of the next target user vertex;

And stopping searching the user vertexes in the corresponding paths when the current searching length is greater than or equal to the preset loop length or the transaction time stamp of the user vertexes corresponding to the current searching length is greater than or equal to the preset loop closing time.

Optionally, the preset loop length is a length threshold preset by the system, that is, the target financial transaction loop exceeding the length threshold cannot be used for detecting abnormal transaction behaviors, which increases the false detection rate. The preset loop closing time is a time threshold preset by the system and has the same effect as the preset loop length, namely, the target financial transaction loop which can not be used as a brushing list for detection is screened out.

S4, the behavior recognition module determines whether the transaction behavior related to the target financial transaction loop is a bill-refreshing behavior based on the structural characteristics of the target financial transaction loop and the historical transaction data set of each user vertex in the target financial transaction loop.

Optionally, the structural feature is used to characterize a number of user vertices corresponding to the target financial transaction ring and a vertex type of each user vertex, the vertex type including a first transaction party and a second transaction party.

Specifically, the step of determining whether the transaction activity associated with the financial transaction loop is a billing activity includes:

Analyzing and obtaining sequence features of vertex sequences of the target financial transaction ring based on structural features of the target financial transaction ring and transaction time stamps between vertices of each user in the target financial transaction ring, and determining whether the sequence features are matched with sequence features of abnormal transaction behaviors;

if so, judging whether the characteristic deviation exists between the current transaction behavior of the first transaction party in the target financial transaction ring and the historical transaction behavior of the first transaction party in the target financial transaction ring according to the historical transaction data set of the corresponding user vertex in the target financial transaction ring; if yes, determining that the transaction behavior related to the financial transaction loop is a bill swiping behavior.

Optionally, the abnormal transaction behavior is an improper transaction behavior with money transaction involving a third party, including a store commodity transaction, a take-away transaction.

Optionally, the sequence features of the vertex sequence are used for indicating flow features, duration periods and the number of users involved of the corresponding financial transaction actions; the sequence features of the abnormal transaction behavior are used to indicate the flow features, duration periods, and number of users involved corresponding to the abnormal transaction behavior.

Specifically, determining whether the characteristic deviation exists between the current transaction behavior of the first transaction party in the target financial transaction loop and the historical transaction behavior of the first transaction party in the target financial transaction loop according to the historical transaction data set of the corresponding user vertex in the target financial transaction loop comprises:

Mapping the multidimensional data feature vector corresponding to each historical transaction data in the historical transaction data set into a plurality of feature subspaces with equal size to obtain feature vectors of the historical transaction data in the feature subspaces with different dimensions, and determining the space coordinate points of the corresponding historical transaction data in the feature subspaces with different dimensions according to the feature vectors;

Counting the space coordinate points in each feature subspace to obtain the feature density of the corresponding feature subspace, acquiring the feature density of each feature subspace in different time windows to determine the density increment of each feature subspace in different time windows, and predicting the density increment of each feature subspace in a preset time period by using poisson distribution;

Fusing the density increment of each characteristic subspace in different time windows and the predicted density increment to obtain the characteristic weight of each characteristic subspace, and constructing a corresponding weight matrix according to the characteristic weights of the historical transaction data sets in different characteristic subspaces;

and determining whether characteristic deviation exists between the current transaction behavior of the first transaction party and the historical transaction behavior of the first transaction party according to the weight matrix and the transaction data generated by the corresponding first transaction party in a preset transaction period.

Optionally, the historical transaction data includes transaction time, transaction amount, transaction party information, transaction merchandise, and order number.

Specifically, the determining, according to the weight matrix and the transaction data generated by the corresponding first transaction party in the preset transaction period, whether the current transaction behavior of the first transaction party has a characteristic deviation from the historical transaction behavior of the first transaction party includes:

Weighting and fusing the feature vectors of each historical transaction data in different feature subspaces according to the weight matrix to obtain a first behavior feature corresponding to the first transaction party, wherein the first behavior feature is used for representing the behavior feature of the historical transaction behavior of the first transaction party;

Mapping transaction data generated by a first transaction party in a preset transaction period into a plurality of feature subspaces with equal size, so as to obtain feature vectors of the transaction data in the feature subspaces with different dimensions, and carrying out weighted fusion on the feature vectors in the different feature subspaces according to the weight matrix, so as to obtain second behavior features corresponding to the first transaction party, wherein the second behavior features are used for representing the behavior features of the current transaction behavior of the first transaction party;

Comparing the first behavioral characteristics with the second behavioral characteristics to determine whether the current transaction behavior of the first transaction party has a characteristic deviation from its historical transaction behavior.

By analyzing the transaction data of the third-party payment platform, the invention identifies the financial transaction loop taking the corresponding transaction user as the transaction action starting point and the transaction action ending point, wherein the financial transaction loop represents that the transaction funds finally flow back to the hand of the transaction user in the transaction loop taking the corresponding transaction user as the transaction action starting point. Therefore, the method and the device avoid the arrangement and the identification of a large amount of redundant transaction data by starting from the key characteristics of the bill-refreshing behavior, and can more efficiently and accurately identify the bill-refreshing behavior in the network.

The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of the present invention. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.

The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather to enable any modification, equivalent replacement, improvement or the like to be made within the spirit and principles of the invention.

Claims (9)

1. The electronic commerce bank behavior detection system based on big data is characterized by comprising a transaction detection cloud platform, a related financial platform and a third party payment platform; the transaction detection cloud platform is respectively in communication connection with the relevant financial platform and the third party payment platform, and the relevant financial platform and the third party payment platform are in communication connection;

The transaction detection cloud platform comprises a network construction module, a vertex analysis module, a transaction ring identification module and a behavior identification module;

The network construction module constructs corresponding commodity transaction subgraphs for users of transaction parties in a relevant financial platform for commodity transaction in a preset transaction period based on transaction data of a third party payment platform, and performs fusion processing on all commodity transaction subgraphs based on user information of a second transaction party for transaction with each first transaction party to generate corresponding commodity transaction graphs, wherein the first transaction party is a commodity consumer in the relevant financial platform, and the second transaction party is a commodity provider in the relevant financial platform;

The vertex analysis module represents the commodity transaction graph in the form of triplet data, and identifies a first user vertex set for commodity transaction with a corresponding target user vertex at a preset time in the commodity transaction graph according to the triplet data, wherein the triplet data comprises a first transaction party, a second transaction party and a transaction time stamp;

The transaction ring identification module acquires all second user vertexes which carry out commodity transaction with each first user vertex in the first user vertex set before the preset moment in the commodity transaction diagram, so as to generate a corresponding second user vertex set for each first user vertex in the first user vertex set, and determines a target financial transaction ring taking a target user vertex as a transaction behavior starting point and a transaction behavior ending point based on the first user vertex set and the second user vertex set;

The behavior recognition module determines whether a transaction behavior related to the target financial transaction loop is a bill-swiping behavior based on structural features of the target financial transaction loop and historical transaction data sets of respective user vertices in the target financial transaction loop.

2. The system of claim 1, wherein the fusing all commodity transaction subgraphs based on user information of the second transaction party transacting with each first transaction party to generate a corresponding commodity transaction map comprises:

Identifying neighbor user vertexes shared between the corresponding first transaction parties based on user information of the second transaction party transacting with each first transaction party, and performing splicing processing on all commodity transaction subgraphs based on attribute information of related edges existing between the corresponding first transaction party and the neighbor user vertexes thereof to generate corresponding commodity transaction graphs, wherein the attribute information of the related edges is used for representing transaction time stamps between the corresponding first transaction party and the neighbor user vertexes thereof.

3. The system of claim 2, wherein determining whether a characteristic deviation exists between a current transaction activity of a first transaction party in the target financial transaction loop and its historical transaction activity based on the historical transaction data set of a corresponding user vertex in the target financial transaction loop comprises:

Mapping the multidimensional data feature vector corresponding to each historical transaction data in the historical transaction data set into a plurality of feature subspaces with equal size to obtain feature vectors of the historical transaction data in the feature subspaces with different dimensions, and determining the space coordinate points of the corresponding historical transaction data in the feature subspaces with different dimensions according to the feature vectors;

Counting the space coordinate points in each feature subspace to obtain the feature density of the corresponding feature subspace, acquiring the feature density of each feature subspace in different time windows to determine the density increment of each feature subspace in different time windows, and predicting the density increment of each feature subspace in a preset time period by using poisson distribution;

Fusing the density increment of each characteristic subspace in different time windows and the predicted density increment to obtain the characteristic weight of each characteristic subspace, and constructing a corresponding weight matrix according to the characteristic weights of the historical transaction data sets in different characteristic subspaces;

and determining whether characteristic deviation exists between the current transaction behavior of the first transaction party and the historical transaction behavior of the first transaction party according to the weight matrix and the transaction data generated by the corresponding first transaction party in a preset transaction period.

4. The system of claim 3, wherein the step of determining whether the transaction activity associated with the financial transaction loop is a billing activity comprises:

Analyzing and obtaining sequence features of vertex sequences of the target financial transaction ring based on structural features of the target financial transaction ring and transaction time stamps between vertices of each user in the target financial transaction ring, and determining whether the sequence features are matched with sequence features of abnormal transaction behaviors;

if so, judging whether the characteristic deviation exists between the current transaction behavior of the first transaction party in the target financial transaction ring and the historical transaction behavior of the first transaction party in the target financial transaction ring according to the historical transaction data set of the corresponding user vertex in the target financial transaction ring; if yes, determining that the transaction behavior related to the financial transaction loop is a bill swiping behavior.

5. The system of claim 4, wherein determining whether the current transaction behavior of the first transaction party has a characteristic deviation from its historical transaction behavior based on the weight matrix and the transaction data generated by the corresponding first transaction party over a predetermined transaction period comprises:

Weighting and fusing the feature vectors of each historical transaction data in different feature subspaces according to the weight matrix to obtain a first behavior feature corresponding to the first transaction party, wherein the first behavior feature is used for representing the behavior feature of the historical transaction behavior of the first transaction party;

Mapping transaction data generated by a first transaction party in a preset transaction period into a plurality of feature subspaces with equal size, so as to obtain feature vectors of the transaction data in the feature subspaces with different dimensions, and carrying out weighted fusion on the feature vectors in the different feature subspaces according to the weight matrix, so as to obtain second behavior features corresponding to the first transaction party, wherein the second behavior features are used for representing the behavior features of the current transaction behavior of the first transaction party;

Comparing the first behavioral characteristics with the second behavioral characteristics to determine whether the current transaction behavior of the first transaction party has a characteristic deviation from its historical transaction behavior.

6. The system of claim 5, wherein determining a target financial transaction loop having target user vertices as transaction initiation points and transaction termination points based on the first set of user vertices and the second set of user vertices comprises:

Detecting all first financial transaction loops which exist in the first user vertex set and the second user vertex set and take corresponding first user vertices as transaction behavior starting points and transaction behavior ending points, and second financial transaction loops which take corresponding second user vertices as transaction behavior starting points and transaction behavior ending points;

Comparing all target financial transaction loops corresponding to the target user vertexes with all first financial transaction loops corresponding to each first user vertex and all second financial transaction loops corresponding to each second user vertex, and judging whether each target financial transaction loop of the target user vertexes has an overlapping area with loop structures of the corresponding first financial transaction loops and the corresponding second financial transaction loops;

if not, adding the vertex of the target user into the loop structures of the corresponding first financial transaction loop and the second financial transaction loop to generate a new target financial transaction loop for the vertex of the target user;

If so, adding loop structures of a first financial transaction loop and/or a second financial transaction loop with overlapping areas corresponding to the vertexes of the target users so that the length of the first financial transaction loop and/or the second financial transaction loop is increased by 1, and updating the corresponding target financial transaction loop by using the increased first financial transaction loop and/or second financial transaction loop.

7. The system of claim 6, wherein the historical transaction data includes transaction time, transaction amount, transaction party information, transaction merchandise, and order number.

8. The system of claim 7, wherein the sequence of characteristics of the vertex sequence is used to indicate flow characteristics, duration periods, and number of users involved for the corresponding financial transaction activity.

9. The system of claim 8, wherein determining a target financial transaction loop having target user vertices as transaction initiation points and transaction termination points based on the first set of user vertices and the second set of user vertices further comprises:

Searching user vertexes in a corresponding path of a commodity transaction diagram by taking a target user vertex as a transaction behavior starting point, taking the searched loop path as a target financial transaction loop of the target user vertex when the current searching length is smaller than a preset loop length and the transaction time stamp of the user vertex corresponding to the current searching length is smaller than a preset loop closing time, and updating the loop closing time of the next target user vertex;

And stopping searching the user vertexes in the corresponding paths when the current searching length is greater than or equal to the preset loop length or the transaction time stamp of the user vertexes corresponding to the current searching length is greater than or equal to the preset loop closing time.