CN112667679B

CN112667679B - Data relationship determination method, device and server

Info

Publication number: CN112667679B
Application number: CN202011493496.2A
Authority: CN
Inventors: 卢健
Original assignee: Industrial and Commercial Bank of China Ltd ICBC
Current assignee: Industrial and Commercial Bank of China Ltd ICBC
Priority date: 2020-12-17
Filing date: 2020-12-17
Publication date: 2024-02-13
Anticipated expiration: 2040-12-17
Also published as: CN112667679A

Abstract

The specification provides a method, a device and a server for determining a data relationship. Based on the method, when a large amount of business data is processed in a big data scene, a target adjacency matrix associated with the business data can be constructed by utilizing the acquired business data according to a preset construction rule; carrying out left-multiplication iteration processing and right-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a corresponding target left-multiplication result vector and a corresponding target right-multiplication result vector; and then, according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and a plurality of service data, a relationship ring formed between the data objects based on the service data can be determined. Therefore, the data processing capacity can be effectively reduced, the relationship circle between the data objects can be efficiently found out from a large amount of service data, the processing efficiency is improved, and the technical problems of large data processing capacity and low processing efficiency in the existing method are solved.

Description

Data relationship determination method, device and server

Technical Field

The specification belongs to the technical field of big data, and particularly relates to a data relationship determining method, a data relationship determining device and a data relationship determining server.

Background

In many big data-based processing scenarios (e.g., bank transfer transaction data processing scenarios, etc.), it is often necessary to analyze and process massive amounts of business data accessed by a platform system to find hidden data relationships between data objects (e.g., relationship circles formed based on the business data, etc.).

Based on the existing method, when processing service data and searching data relation between data objects, the technical problems of large data processing amount and low processing efficiency often exist. The above problem is particularly pronounced when it comes to data relationship searching and mining for higher depths.

In view of the above problems, no effective solution has been proposed at present.

Disclosure of Invention

The specification provides a method, a device and a server for determining a data relationship, which can effectively reduce data processing capacity, efficiently determine a relationship circle between data objects from a large amount of service data, and improve processing efficiency.

The specification provides a method for determining a data relationship, which comprises the following steps:

acquiring a plurality of service data; each service data in the plurality of service data respectively carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data;

According to a preset construction rule, constructing a target adjacency matrix associated with the plurality of business data by utilizing the plurality of business data;

carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth;

performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction;

and determining a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data.

In one embodiment, constructing, according to a preset construction rule, a target adjacency matrix associated with the plurality of service data by using the plurality of service data, includes:

Establishing a corresponding relation between a matrix row number and an identity of a data object initiating service data, and establishing a corresponding relation between a matrix column number and an identity of a data object receiving the service data; and determining the data value of the data element at the position indicated by the combination of the corresponding matrix row number and the matrix column number in the matrix according to the service data.

In one embodiment, after constructing the target adjacency matrix associated with the plurality of service data using the plurality of service data according to a preset construction rule, the method further comprises: and storing the target adjacent matrix in a memory in a sparse format.

In one embodiment, according to a preset iteration rule, performing a preset number of left-multiplication iteration processing on the target adjacency matrix to obtain a target left-multiplication result vector, where the method includes:

according to a preset iteration rule, carrying out the current left-multiplication iteration in preset times on the target adjacent matrix to obtain a current left-multiplication result vector, wherein the method comprises the following steps of:

obtaining a last left multiplication result vector;

using the last left multiplication result vector to multiply the target adjacent matrix left to obtain the current left multiplication result vector; the current time of the left multiplication result vector is used for indicating that the identity of the data object pointed by the adjacent data object through the service data exists in the forward direction under the corresponding relation depth of the current time; the current time is less than or equal to the preset time.

In one embodiment, according to a preset iteration rule, performing right multiplication iteration processing on the target adjacency matrix for a preset number of times to obtain a target right multiplication result vector, including:

according to a preset iteration rule, carrying out right multiplication iteration of the current time in preset times on the target adjacent matrix to obtain a right multiplication result vector of the current time, wherein the right multiplication result vector comprises the following steps:

obtaining the right multiplication result vector of the last time;

right multiplying the target adjacent matrix by using the transpose vector of the right multiplication result vector of the last time to obtain the right multiplication result vector of the current time; the right multiplication result vector of the current time is used for indicating that the identity of the data object pointed by the adjacent data object through the service data exists in the reverse direction under the corresponding relation depth of the current time; the current time is less than or equal to the preset time.

In one embodiment, determining a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data comprises:

performing dot multiplication processing on the target left-multiplication result vector and the target right-multiplication result vector to determine the identity of a first data object forming a first relationship circle; the first relationship circle is a relationship circle with a relationship depth smaller than or equal to a preset relationship depth;

According to the identity of the first data object, service data related to the first data object is screened out from a plurality of service data to serve as first service data;

and determining a plurality of first relationship circles according to the identity of the first data object and the first service data.

In one embodiment, determining a plurality of first relationship circles according to the identity of the first data object and the first service data includes:

and calling the GPU through a preset deep learning framework, and generating a plurality of first relation circles according to the identity identification of the first data object and the first service data.

In one embodiment, after the GPU is invoked to generate a plurality of first relationship circles according to the identity of the first data object and the first service data through a preset deep learning framework, the method further includes:

and filtering the identity of the first data object according to the generated multiple first relation circles.

In one embodiment, after determining a plurality of first relationship circles according to the identity of the first data object and the first service data, the method further includes:

And determining the relationship circles corresponding to different relationship depths from the plurality of first relationship circles through aggregation iteration.

In one embodiment, the business data includes transfer transaction data and the data object includes a transaction account.

In one embodiment, after determining a relationship circle formed between data objects based on the business data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, and the plurality of business data, the method further comprises:

and determining whether the transaction account has preset transaction risk according to a relationship circle formed between the transaction accounts based on transfer transaction data.

The specification also provides a method for determining the ring diagram, which comprises the following steps:

acquiring node data and edge data;

according to a preset construction rule, constructing a target adjacency matrix by utilizing the edge data and the node data;

carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating nodes meeting requirements in the forward direction;

performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; wherein the data elements in the target right-hand product result vector are used to indicate nodes that meet the requirements in the reverse direction;

And determining a ring graph formed by the node data and the edge data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, the node data and the edge data.

The specification also provides a data relationship determining device, which comprises:

the acquisition module is used for acquiring a plurality of service data; each service data in the plurality of service data respectively carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data;

the construction module is used for constructing a target adjacency matrix associated with the plurality of business data by utilizing the plurality of business data according to a preset construction rule;

the first processing module is used for carrying out the left-multiplication iteration processing of the preset times on the target adjacent matrix according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth;

The second processing module is used for carrying out right multiplication iteration processing on the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction;

and the determining module is used for determining a relation ring formed between the data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data.

The present specification provides a server comprising a processor and a memory for storing processor-executable instructions, the processor implementing, when executing the instructions, obtaining a plurality of service data; each service data in the plurality of service data respectively carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data; according to a preset construction rule, constructing a target adjacency matrix associated with the plurality of business data by utilizing the plurality of business data; carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth; performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction; and determining a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data.

The present specification provides a computer readable storage medium having stored thereon computer instructions that, when executed, enable obtaining a plurality of business data; each service data in the plurality of service data respectively carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data; according to a preset construction rule, constructing a target adjacency matrix associated with the plurality of business data by utilizing the plurality of business data; carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth; performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction; and determining a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data.

When facing a large amount of accessed service data, the method, the device and the server for determining the data relationship firstly construct and obtain a target adjacent matrix associated with the plurality of service data by utilizing the plurality of acquired service data according to a preset construction rule; respectively carrying out left multiplication iteration processing of preset times and right multiplication iteration processing of preset times on the target adjacent matrix according to a preset iteration rule to obtain a target left multiplication result vector capable of indicating the identity of the data object meeting the requirements in the forward direction and a target right multiplication result vector capable of indicating the identity of the data object meeting the requirements in the reverse direction; and then, according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and a plurality of service data, a relationship circle formed between the data objects based on the service data can be rapidly determined. Therefore, the data processing capacity can be effectively reduced, the relationship circle between the data objects can be efficiently determined from a large amount of service data, the specific data relationship between the data objects can be accurately analyzed, the processing efficiency is improved, and the technical problems of large data processing capacity and low processing efficiency in the existing method are solved.

Drawings

In order to more clearly illustrate the embodiments of the present disclosure, the drawings that are required for the embodiments will be briefly described below, in which the drawings are only some of the embodiments described in the present disclosure, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.

FIG. 1 is a schematic diagram of one embodiment of the structural composition of a system to which the data relationship determination method provided by the embodiments of the present specification is applied;

FIG. 2 is a flow chart of a method for determining data relationships provided by one embodiment of the present disclosure;

FIG. 3 is a schematic diagram of one embodiment of a method for determining data relationships provided by embodiments of the present disclosure, in one example scenario;

FIG. 4 is a schematic diagram of one embodiment of a method for determining data relationships provided by embodiments of the present disclosure, in one example scenario;

FIG. 5 is a schematic diagram of one embodiment of a method for determining data relationships provided by embodiments of the present disclosure, in one example scenario;

FIG. 6 is a flow diagram of a method for determining a torus diagram provided by an embodiment of the present description;

FIG. 7 is a schematic diagram of the structural composition of a server according to one embodiment of the present disclosure;

FIG. 8 is a schematic structural diagram of a data relationship determining apparatus according to an embodiment of the present disclosure;

FIG. 9 is a schematic diagram of one embodiment of a method of determining data relationships provided by embodiments of the present disclosure, in one example scenario;

fig. 10 is a schematic diagram of an embodiment of a method for determining a data relationship provided by the embodiments of the present specification, in one scenario example.

Detailed Description

In order to make the technical solutions in the present specification better understood by those skilled in the art, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only some embodiments of the present specification, not all embodiments. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are intended to be within the scope of the present disclosure.

In view of the existing data relationship-based determination method, when implemented, the method is often implemented by traversing and searching the direct or indirect interrelationships between the data object and other data objects level by level for single data objects. By applying the method, when the related data objects are more, for example, the related massive data objects in a big data scene, the number of the data objects needing to be traversed and searched is huge; or when the depth of the data relationship to be searched is deeper, a large amount of data processing resources and processing time are required to be consumed, and calculation of a larger data amount is performed to search and find the data relationship to be searched. Therefore, when the conventional method is implemented, there are often technical problems of large data processing amount and low processing efficiency.

For the root cause of the above problem, after the creative labor, the present specification considers that a target adjacency matrix which is related to all data objects and interactive data information among the data objects can be constructed based on a large number of data objects related to a large number of business data; furthermore, the target adjacency matrix can be subjected to repeated left-multiplication iterative processing and repeated right-multiplication iterative processing according to a preset iterative rule to obtain a target left-multiplication result vector which can indicate batch data objects meeting the requirements along the positive direction and a target right-multiplication result vector which can indicate batch data objects meeting the requirements along the negative direction; and then, according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and corresponding service data, a relationship ring formed based on the service data among the data objects can be found in batches. Therefore, the data processing capacity can be effectively reduced, the relationship circle between the data objects can be efficiently found out from a large amount of service data, the processing efficiency is improved, and the technical problems of large data processing capacity and low processing efficiency in the existing method are solved.

The embodiment of the specification provides a data relationship determining method. The method for determining the data relationship can be applied to a system comprising a server and front-end equipment. Reference may be made in particular to fig. 1. The front-end equipment and the server in the system can be connected in a wired or wireless mode to carry out specific data interaction.

When in specific implementation, the front-end equipment can access a plurality of service data; each service data in the plurality of service data respectively carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data; and transmitting the accessed plurality of service data to the server.

The server may be configured to obtain a plurality of service data, and construct, according to a preset construction rule, a target adjacency matrix associated with the plurality of service data using the plurality of service data.

Further, the server may perform a pre-set number of left-multiplication iteration processing on the target adjacency matrix according to a pre-set iteration rule, to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth; meanwhile, right multiplication iteration processing of preset times can be carried out on the target adjacent matrix according to a preset iteration rule, so that a target right multiplication result vector is obtained; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction.

Further, the server may determine a relationship circle formed between the data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, and the plurality of service data.

In this embodiment, the server may specifically include a background server applied to a side of a service data processing platform and capable of implementing functions such as data transmission and data processing. Specifically, the server may be, for example, an electronic device having data operation, storage function and network interaction function. Alternatively, the server may be a software program running in the electronic device that provides support for data processing, storage, and network interactions. In the present embodiment, the number of servers is not particularly limited. The server may be one server, several servers, or a server cluster formed by several servers.

In this embodiment, the front-end device may specifically include a front-end electronic device that is applied to a side of a service data interaction platform and is capable of implementing functions such as data acquisition and data transmission. Specifically, the front-end device may be, for example, a desktop computer, a tablet computer, a notebook computer, a smart phone, etc. Alternatively, the front-end device may be a software application capable of running in the electronic device described above. For example, it may be an APP running on a mobile phone, etc.

Referring to fig. 2, the embodiment of the present disclosure provides a method for determining a data relationship. The method is particularly applied to the server side. In particular implementations, the method may include the following.

S21: acquiring a plurality of service data; each service data in the plurality of service data carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data.

S22: and constructing a target adjacency matrix associated with the plurality of business data by utilizing the plurality of business data according to a preset construction rule.

S23: carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth.

S24: performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction.

S25: and determining a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data.

By the embodiment, the data relationship between each data object and other data objects is avoided from being searched level by level in a traversing and searching mode like the prior method, and a target adjacency matrix associated with a plurality of business data is constructed first; carrying out repeated iterative processing on the target adjacent matrix according to a preset iterative rule, and finding out data objects meeting requirements in the forward direction and the reverse direction in batches; and the relationship circle between the data objects can be determined based on the data objects and corresponding service data, so that the data processing capacity is effectively reduced, the processing efficiency is improved, the relationship circle between the data objects can be efficiently found out from a large amount of service data, and the technical problems of large data processing capacity and low processing efficiency in the existing method are solved.

In this embodiment, the above-mentioned service data may be specifically understood as a service data having directivity. Specifically, each service data may carry an identity of a data object (may be abbreviated as an originating object) that originates the service data, and an identity of a data object (may be abbreviated as a receiving object) that receives the service data, and the originating object may point to the receiving object through the service data.

Specifically, the service data may include transfer transaction data, money transfer data, mail data, and the like. Accordingly, the data objects may include account objects participating in transfer transactions, mailbox objects participating in mailing, and the like. Of course, it should be noted that the above-listed service data and data objects are only illustrative. In specific implementation, according to specific application scenarios and processing requirements, the service data may further include other types of service data, and correspondingly, the data object may further include other types of data objects. The present specification is not limited to this.

In this embodiment, the identification is specifically understood as identification information capable of indicating the data object. Specifically, the identification mark may be a name of the data object, a number of the data object, an address parameter of the data object, and the like.

In an embodiment, the constructing, according to the preset construction rule, the target adjacency matrix associated with the plurality of service data by using the plurality of service data may include the following when implemented: establishing a corresponding relation between a matrix row number and an identity of a data object initiating service data, and establishing a corresponding relation between a matrix column number and an identity of a data object receiving the service data; and determining the data value of the data element at the position indicated by the combination of the corresponding matrix row number and the matrix column number in the matrix according to the service data.

In this embodiment, the target adjacency matrix may specifically include data information related to the identities of the data objects related to the plurality of service data, and related to the pointing relationships between different data objects based on the service data.

In this embodiment, when the target adjacency matrix is constructed according to a preset construction rule, reference may be made to fig. 3. All data objects (including an initiating object and a receiving object) related to a plurality of business data can be determined first, and the identity of the data objects is obtained.

For example, a total of N data objects are found out according to the plurality of service data, and the identity of each data object in the N data objects is obtained.

Further, the rows and columns of the matrix may be established based on the identity of the data object. For example, referring to fig. 3, a matrix of N rows and N columns may be constructed according to the identities of N data objects. Wherein each matrix row number specifically corresponds to an identity of a data object initiating the service data. Each matrix column number specifically corresponds to an identity of a data object that receives the service data. Also, a combination of a matrix row number and a matrix column number may indicate a data element in the matrix that is located at a corresponding column position of a corresponding row. Any one data element in the matrix may correspond to one service data, and the data value of the data element may be determined according to the corresponding service data.

Therefore, when each data element in the matrix is determined, the combination of the matrix row number and the matrix column number corresponding to the data element can be determined first; and then searching and determining the specific data value of the data element according to the acquired multiple service data and the service data corresponding to the combination of the matrix row number and the matrix column number.

Referring to fig. 3, for example, a combination of matrix row number 1 and matrix column number 3 (which may be denoted as (1, 3)) is used to indicate the data element in the matrix at the 1 row 3 column position. Wherein, the matrix row number 1 corresponds to the identity of the data object a, and the matrix column number 3 corresponds to the identity of the data object C. Thus, in the matrix, the service data corresponding to the data element located at the position indicated by (1, 3) is one that is originated by the data object a and received by the data object C.

When the data value of the data element of (1, 3) is specifically determined, the acquired plurality of service data can be searched, and whether the service data exists or not is determined, namely, the carried identification mark of the initiating object indicates the data object A, and the carried identification mark of the receiving object indicates the service data of the data object B. In the case where the service data is retrieved from the plurality of service data, the data value of the data element is set to 1. In contrast, in the case where the corresponding service data is not retrieved from the plurality of service data, the data value of the data element is set to 0.

Through the embodiment, the target adjacency matrix containing the identity of the data object and the pointing relation generated based on the service data among different data objects is constructed according to a plurality of service data. In the subsequent further processing, the processing of a large amount of service data can be converted into the processing of the target adjacent matrix, so that the subsequent data processing amount can be effectively reduced, and the overall data processing efficiency is improved.

In one embodiment, referring to fig. 3, it is further considered that the data values of the data elements in the target adjacency matrix constructed according to the preset construction rule in the above manner are mostly 1 or 0, which belongs to a sparse matrix. Therefore, the characteristics of the sparse matrix can be utilized to store and fetch the target adjacent matrix.

In one embodiment, after constructing the target adjacency matrix associated with the plurality of service data by using the plurality of service data according to a preset construction rule, the method may further include the following when implemented: and storing the target adjacent matrix in a memory in a sparse format.

Through the embodiment, the characteristics of the sparse matrix can be fully utilized to store the target adjacent matrix by using less storage resources, and meanwhile, subsequent reading and corresponding data processing by using the target adjacent matrix are convenient, so that the data processing efficiency is further improved.

In this embodiment, when the target adjacency matrix is specifically stored, only the matrix row number and the matrix column number (which may also be referred to as coordinates of the data element in the matrix) of the data element with the data value of 1 may be stored. Specifically, a 2 row m column storage vector may be constructed. The first row in the storage vector is used for storing a matrix row number (or an identity of a data object corresponding to the matrix row number) of the data element with the data value of 1, and the second row in the storage vector is used for storing a matrix column number (or an identity of a data object corresponding to the matrix column number) of the data element with the data value of 1.

In one embodiment, when specifically searching for a data object with a data relationship, matrix iterative operation aiming at a target adjacency matrix can be utilized to replace traversing and searching algorithms adopted by the existing method, so that a plurality of data objects with the data relationship can be efficiently found in batches. The plurality of data objects stored in the data relationship can be understood as data objects forming a relationship ring based on the service data.

The relationship ring can be understood as a ring-shaped relationship structure. The relation structure at least comprises two data objects, and different data objects are connected through a service data directed edge to form a ring graph. Any one data object in the relationship structure may be directed back to itself via the business data by other data objects in the relationship structure.

Specifically, reference may be made to fig. 4. The data objects a, B, C, and the service data 1 (the initiating object is the data object a, the receiving object is the data object B), the service data 2 (the initiating object is the data object B, the receiving object is the data object C), and the service data 3 (the initiating object is the data object C, the receiving object is the data object a) related to the data objects form a relationship ring. Wherein data object a points to data object B via traffic data 1, which in turn points to data object C via traffic data 2, which in turn points back to data object a via traffic data 3.

In an embodiment, during implementation, a pre-set number of left-multiplication iteration processing may be performed on the target adjacent matrix according to a pre-set iteration rule, so as to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth; meanwhile, carrying out right multiplication iteration processing on the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction; then, according to the target left-multiplication result vector and the target right-multiplication result vector, determining the identity of the data object forming the corresponding relation circle; and then, according to the identity of the data objects forming the corresponding relationship circle, the relationship circle formed by the data objects based on the service data can be determined according to the service data related to the data objects.

In an embodiment, the foregoing performing, according to a preset iteration rule, a pre-set number of left-multiplication iteration processing on the target adjacency matrix to obtain a target left-multiplication result vector, where in implementation, the method may include the following: according to a preset iteration rule, carrying out the current left-multiplication iteration in preset times on the target adjacent matrix to obtain a current left-multiplication result vector, wherein the method comprises the following steps of: obtaining a last left multiplication result vector; using the last left multiplication result vector to multiply the target adjacent matrix left to obtain the current left multiplication result vector; the current time of the left multiplication result vector is used for indicating that the identity of the data object pointed by the adjacent data object through the service data exists in the forward direction under the corresponding relation depth of the current time; the current time is less than or equal to the preset time.

In this embodiment, the current-time left-hand multiplication result vector may specifically include a plurality of data elements arranged in sequence. The number of the data elements contained in the current multiplied result vector is the same as the total number of the data objects, and each data element in the plurality of data elements in the vector corresponds to the identity of one data object.

If the data value of a certain data element in the vector is determined to be 0 according to the left-hand multiplication result vector of the current time, it is possible to determine that the data object corresponding to the data element does not have a neighboring data object directed to itself in the forward direction (for example, directed clockwise) at the depth of the relationship corresponding to the current time. In contrast, if the data value of a certain data element in the vector is determined to be 1 according to the left-hand multiplication result vector of the current time, the data object corresponding to the data element may be determined such that there is a neighboring data object directed to itself in the forward direction at the depth of the relationship corresponding to the current time.

In this embodiment, the initial left-hand multiplication result vector (i.e., the 0-time left-hand multiplication result vector) may be an N-dimensional vector with all 1 data elements, where N is the total number of data objects involved in the plurality of service data.

Through the embodiment, the identity of the data object pointed by other adjacent data objects through the service data in the forward direction under the preset relation depth corresponding to the preset times can be rapidly found out.

Specifically, for example, referring to fig. 5, the plurality of service data includes: service data 1 and service data 2; wherein, the initiating object of the service data 1 is A, the receiving object is B, the initiating object of the service data 2 is B, and the receiving object is C.

At this time, a target adjacency matrix shown below can be constructed based on the above-described plurality of traffic data:

wherein, matrix row number 1 in the matrix corresponds to initiating object A, matrix row number 2 corresponds to initiating object B, matrix row number 3 corresponds to initiating object C, matrix column number 1 corresponds to receiving object A, matrix column number 2 corresponds to receiving object B, and matrix column number 3 corresponds to receiving object C. The data element indicated by the combination (1, 2) in the matrix corresponds to traffic data 1, with a data value of 1. The data element indicated by the combination (2, 3) corresponds to traffic data 2, with a data value of 1.

The initial left-hand multiplication result vector is generated as (1, 1), and when the first left-hand multiplication iterative process is performed, the last left-hand multiplication result vector (i.e. the initial left-hand multiplication result vector) can be utilized to carry out the left-hand multiplication on the target adjacent matrix according to the preset iterative rule, and the obtained result vector is used as the current left-hand multiplication result vector, i.e. the first left-hand multiplication result vector.

Specifically, the first left-hand iteration process may be performed with reference to the following formula:

based on the above equation, the corresponding first-time left-hand multiplication result vector is noted as (0, 1).

The data value of the first data element in the vector is 0, which may indicate that the data object a has no other data objects (neighboring data objects) pointing in the forward direction under the depth of the relationship (i.e., 1 degree) corresponding to the current time (i.e., the first time). The data value of the second data element being 1 may indicate that at the depth of the relationship (i.e., 1 degree) to which the current time (i.e., the first time) corresponds, data object B has other data object orientations in the forward direction (e.g., data object a). The data value of the third data element being 1 may indicate that at the depth of the relationship (i.e., 1 degree) corresponding to the current time (i.e., the first time), the data object C has other data object orientations (e.g., data object B) in the forward direction.

Therefore, based on the first-time left-hand multiplication result vector (0, 1) can indicate that under the corresponding relation depth of the first time, the identity of the data object pointed by the adjacent data object through the service data exists in the forward direction, so that the identities B and C of the data object pointed by the adjacent data object through the service data exists in the forward direction can be reserved, and the identity A of the data object pointed by other adjacent data objects is filtered out.

Then, in a similar manner, the target left-multiplication result vector is obtained and utilized, and the target adjacent matrix is subjected to secondary left-multiplication to obtain a secondary left-multiplication result vector.

Specifically, the second multiplication iteration process may be performed with reference to the following expression:

based on the above equation, the corresponding second-order left-hand multiplication result vector is noted as (0, 1).

The data value of the first data element in the vector is 0, which may indicate that the data object a has no other data object pointing in the forward direction under the depth of the relationship (i.e., 2 degrees) corresponding to the current time (i.e., the second time). The data value of the second data element being 0 may indicate that at the depth of the relationship (i.e. 2 degrees) corresponding to the current time (i.e. the second time), the data object B is also not pointed to by other data objects in the forward direction. The data value of the third data element being 1 may indicate that at the depth of the relationship (i.e., 2 degrees) corresponding to the current time (i.e., second time), data object C has other data object orientations in the forward direction (e.g., data object B).

Therefore, based on the second-time left-hand multiplication result vector (0, 1) can indicate that under the corresponding relation depth of the second time, the identity of the data object pointed by the adjacent data object through the service data exists in the forward direction, so that the identity C of the data object pointed by the adjacent data object through the service data exists in the forward direction can be reserved, and the identity B of the data object pointed by other data objects is filtered out.

Repeating the processing mode, and carrying out left-multiplication iteration processing on the target adjacent matrix for preset times to obtain a corresponding last left-multiplication result vector as a target left-multiplication result vector.

The data value of each data element in the target multiplied result vector is used for indicating whether the corresponding data object has an adjacent data object pointing to the corresponding data object along the forward direction under the preset relation depth.

In the implementation, the identity of the data object pointing to the adjacent data object along the forward direction under the preset relation depth can be selected from a plurality of data objects according to the data values of the data elements in the target multiplication vector, and the identity of the data object meeting the requirement along the forward direction is used as the identity of the data object.

In this embodiment, the specific value of the preset number of times may be determined according to a preset depth of relationship. The preset relation depth can be flexibly set according to application scenes and processing requirements.

Specifically, for example, the preset depth of relationship may be 3 degrees, and correspondingly, the preset times may be 3 times. For another example, the preset depth of relationship may be 6 degrees, and correspondingly, the preset number of times may be 6 times.

The target left-hand multiplication result vector is similar to other left-hand multiplication result vectors, and can indicate that under the preset relation depth corresponding to the preset times, the identity of the data object pointed by the adjacent data object through the service data exists in the forward direction, namely the identity of the data object meeting the requirement in the forward direction.

In one embodiment, for the data relationship shown in FIG. 5, when proceeding to the third left-hand iteration, the resulting left-hand result vector becomes: (0,) indicating that no other data object is pointed to by the business data in the forward direction for all data objects under the third corresponding relation depth based on the vector, filtering out the used data objects. Further, it can be determined that the data object a, the data object B, and the data object C cannot form a relationship ring by the service data 1 and the service data 2.

For the data relationship shown in fig. 4, according to a preset construction rule, the constructed target adjacency matrix is:performing a first left-multiplication iteration on the target adjacent matrix according to a preset iteration rule in the same manner to obtain a first left-multiplication result vector which is recorded as (1, 1); for the target adjacent momentThe array performs a second left-hand iteration to obtain a second left-hand result vector (1, 1) … …, and performs an nth left-hand iteration on the target adjacent matrix to obtain an nth left-hand result vector (1, 1). At this point, it can be seen that no matter how many left-hand iterations are performed, there are always other data objects pointing to data object a, data object B, and data object C in the forward direction, which cannot be filtered out by matrix iterations on the target adjacency matrix. It can be further determined that the data object a, the data object B, and the data object C form a relationship circle by the service data 1 and the service data 2.

In an embodiment, the performing right multiplication iteration processing on the target adjacency matrix for a preset number of times according to the preset iteration rule to obtain a target right multiplication result vector, and the implementation may include the following: according to a preset iteration rule, carrying out right multiplication iteration of the current time in preset times on the target adjacent matrix to obtain a right multiplication result vector of the current time, wherein the right multiplication result vector comprises the following steps: obtaining the right multiplication result vector of the last time; right multiplying the target adjacent matrix by using the transpose vector of the right multiplication result vector of the last time to obtain the right multiplication result vector of the current time; the right multiplication result vector of the current time is used for indicating that the identity of the data object pointed by the adjacent data object through the service data exists in the reverse direction under the corresponding relation depth of the current time; the current time is less than or equal to the preset time.

In this embodiment, the initial right-multiplication result vector (i.e., the right-multiplication result vector of 0 times) may be an N-dimensional vector with all data elements being 1, where N is the total number of data objects involved in the plurality of service data.

In this embodiment, the current right-multiplication result vector may specifically include a plurality of data elements arranged in sequence. The number of the data elements contained in the right multiplication result vector of the current time is the same as the total number of the data objects, and each data element in the plurality of data elements in the vector corresponds to the identity of one data object.

If the right multiplication result vector of the current time is based on the data value of a certain data element in the vector being 0, it is possible to determine that the data object corresponding to the data element does not exist in the reverse direction (e.g., counterclockwise direction) with respect to the depth of the relationship corresponding to the current time. In contrast, if the right multiplication result vector of the current time is based on the determination that the data value of a certain data element in the vector is 1, the data object corresponding to the data element may be determined, and in the depth of the relationship corresponding to the current time, there is a neighboring data object directed to itself in the reverse direction.

The process of performing the right-multiplication iteration process for the preset number of times on the target adjacent matrix is similar to the process of performing the left-multiplication iteration process for the preset number of times on the target adjacent matrix, and reference may be made to the embodiment of performing the right-multiplication iteration process for the preset number of times on the target adjacent matrix, which is not described in detail in this specification.

Repeating the processing mode, and carrying out right multiplication iteration processing on the target adjacent matrix for preset times to obtain a corresponding last right multiplication result vector as a target right multiplication result vector.

The data values of the data elements in the target right-multiplication result vector are used for indicating whether adjacent data objects pointing to the corresponding data objects exist along the reverse direction under the preset relation depth.

In the implementation, the identity of the data object pointing to the adjacent data object along the reverse direction under the preset relation depth can be selected from a plurality of data objects according to the data values of the data elements in the target right-multiplied vector, and the identity of the data object meeting the requirement along the reverse direction is used as the identity of the data object.

Through the embodiment, the identity of the data object pointed by the adjacent data object through the service data on the negative direction under the preset relation depth corresponding to the preset times can be rapidly found out.

In one embodiment, the determining the relationship circle formed by the data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the service data may include the following when implemented:

s1: performing dot multiplication processing on the target left-multiplication result vector and the target right-multiplication result vector to determine the identity of a first data object forming a first relationship circle; the first relationship circle is a relationship circle with a relationship depth smaller than or equal to a preset relationship depth;

s2: according to the identity of the first data object, service data related to the first data object is screened out from a plurality of service data to serve as first service data;

s3: and determining a plurality of first relationship circles according to the identity of the first data object and the first service data.

In this embodiment, the two data objects may be crossed by performing a dot multiplication process on the target left-multiplication result vector and the target right-multiplication result vector, so as to find the identity of the first data object capable of forming a ring structure.

In this embodiment, it should be noted that the first relationship circle determined above may specifically include a relationship circle corresponding to a preset relationship depth, or may include a relationship circle corresponding to a relationship depth smaller than the preset relationship depth.

Specifically, for example, if the preset relationship depth is 3, the determined first relationship circle may include two relationship circles including a relationship circle of 2 degrees and a relationship circle of 3 degrees.

By the above embodiment, a plurality of first relationship circles formed based on the service data can be efficiently determined.

In one embodiment, the determining a plurality of first relationship circles according to the identity of the first data object and the first service data may include the following when implemented: and calling the GPU through a preset deep learning framework, and drawing to generate a plurality of first relation circles according to the identity identification of the first data object and the first service data.

The preset deep learning frame may specifically include a Tensorflow frame, a PyTorch frame, and the like. The Tensorflow is a symbolic mathematical system based on data flow programming (dataflow programming), and is widely applied to the programming implementation of various machine learning algorithms. The above-mentioned PyTorch may be specifically an open source Python machine learning library, and is used for applications such as natural language processing based on Torch. Of course, the above-listed preset deep learning framework is only one illustrative illustration. In particular, other suitable deep learning frameworks may be used as appropriate. The present specification is not limited to this.

By the embodiment, a plurality of first relationship circles can be generated more quickly according to the identity of the first data object and the first service data, so that the overall processing efficiency is further improved.

In one embodiment, after the GPU is invoked to generate a plurality of first relationship circles according to the identity of the first data object and the first service data through the preset deep learning framework, the method may further include the following when implemented: and filtering the identity of the first data object according to the generated multiple first relation circles.

In this embodiment, during implementation, whether the data object that does not form the relationship ring exists in the first data object determined before can be checked according to the generated first relationship ring return, and the data object that does not form the relationship ring is filtered, or the identity of the data object that does not meet the requirement and points to the data object itself is filtered, so that errors introduced by the data object are reduced.

In one embodiment, after determining a plurality of first relationship circles according to the identity of the first data object and the first service data, the method may further include the following when implemented: and determining the relationship circles corresponding to different relationship depths from the plurality of first relationship circles through aggregation iteration.

Through the embodiment, various relationship circles corresponding to different relationship depths can be further determined from the plurality of first relationship circles efficiently and accurately, so that finer data processing is supported, and more diversified use requirements of users are met.

Specifically, for example, the preset relationship depth is 4. By the above embodiment, a relationship circle with a relationship depth of 2 (the relationship circle contains only 2 data objects), a relationship circle with a relationship depth of 3 (the relationship circle contains only 3 data objects), and a relationship circle with a relationship depth of 4 (the relationship circle contains only 4 data objects) can be finely split from the first relationship circle.

In this embodiment, the aggregation iteration may be performed according to the following algorithm to sequentially output the relationship circles corresponding to 2 to n degrees, respectively. Wherein the preset depth of relation is n.

S1: a secondary number (or identity of a secondary) is added to each location point in the matrix row number, represented by the { ID } set.

S2: the sub-numbers of the start point are aggregated according to the end point (stored in key-value form, key being the ID of the end point; value being the set of all sub-numbers pointing to the start point of the end point). The end point may be the last received object calibrated in the first relationship circle, and the start point may be the first initiating object calibrated in the first relationship circle.

S3: updating the secondary number of the current terminal point to be the aggregated secondary number set.

S4: it is determined whether the number of each point belongs to its sub-number set, and if so, the number IDs of the points are output, and the points indicating the numbers belong to a member of an i-degree ring.

S5: steps S2 to S4 are repeatedly performed until an n-degree loop is output.

S6: and outputting all rings according to the connection relation (edge) file between the ring member points under each degree output by the S4 and the points output by the S3, or directly inputting the rings into the visualization tool for displaying.

In one embodiment, the business data may include transfer transaction data and the data object may include a transaction account.

In this embodiment, in the account transfer data processing scenario of a bank or other financial institution, according to the acquired plurality of account transfer transaction data, according to the method provided in the present specification, a relationship circle formed between a plurality of transaction accounts based on the account transfer transaction data may be determined. Therefore, the transfer transaction data can be efficiently and massively processed to mine the relationship circle between transaction accounts so as to carry out subsequent further analysis and processing.

In one embodiment, after determining a relationship circle formed between data objects based on service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, and the plurality of service data, the method may further include, when implemented, the following: and determining whether the transaction account has preset transaction risk according to a relationship circle formed between the transaction accounts based on transfer transaction data.

Through the embodiment, a large amount of transfer transaction data in a transfer data processing scene of a bank or other financial institutions can be efficiently analyzed and processed, so that a transaction account with preset transaction risk can be accurately found out.

Specifically, for example, as shown in fig. 4, the service data is transfer transaction data, and it can be known from the relationship circles: the transfer transaction data initiated by the data object A to the data object B finally flows back to the data object A through the data object C. Furthermore, by combining other related data characteristics, the fact that illegal behaviors such as money laundering and the like are mutually carried out through transfer transactions among the data object A, the data object B and the data object C can be primarily judged, and further the fact that preset transaction risks exist among the data object A, the data object B and the data object C can be judged.

Further, corresponding risk prompt information may be generated for the data object, so as to enable more strict monitoring and analysis for transfer transaction data initiated by the data object subsequently. And (3) carrying out effective punishment in time under the condition of determining that transaction risks do exist.

As can be seen from the above, in the method for determining a data relationship provided in the embodiments of the present disclosure, when a large amount of service data is processed, a target adjacency matrix associated with the plurality of service data is constructed by using the plurality of acquired service data according to a preset construction rule; respectively carrying out left multiplication iteration processing of preset times and right multiplication iteration processing of preset times on the target adjacent matrix according to a preset iteration rule to obtain corresponding target left multiplication result vectors capable of indicating the identity of the data object meeting the requirements in the forward direction and target right multiplication result vectors capable of indicating the identity of the data object meeting the requirements in the reverse direction; and then the relationship circle formed by the data objects based on the service data can be determined according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data. Therefore, the data processing capacity can be effectively reduced, the relationship circle between the data objects can be efficiently found out from a large amount of service data, the data relationship between the data objects can be accurately analyzed, the processing efficiency is improved, and the technical problems of large data processing capacity and low processing efficiency in the existing method are solved.

Referring to fig. 6, the embodiment of the present disclosure further provides a method for determining a ring map, so as to efficiently find a corresponding ring map. In particular implementations, the method may include the following.

S61: node data and edge data are acquired.

S62: and constructing a target adjacency matrix by utilizing the edge data and the node data according to a preset construction rule.

S63: carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; wherein the data elements in the target left-hand result vector are used to indicate nodes that meet the demand in the forward direction.

S64: performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; wherein the data elements in the target right-hand product result vector are used to indicate nodes that meet the requirements in the reverse direction.

S65: and determining a ring graph formed by the node data and the edge data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, the node data and the edge data.

In this embodiment, the above-mentioned ring graph can be specifically understood as a graph in which a plurality of node data are connected end to end by edge data to form a ring. Specifically, reference may be made to fig. 4.

In this embodiment, the ring graph may specifically include a relationship ring that can characterize the data relationships between the data objects. The relationship circle can comprise interaction circles among friends, funds return circles among transaction accounts, communication circles among departments and the like according to different application scenes.

Through the embodiment, the data processing amount can be effectively reduced in a matrix iteration mode, and the corresponding ring graph is efficiently found out according to a large amount of acquired node data and edge data.

The embodiment of the specification also provides a server, which comprises a processor and a memory for storing instructions executable by the processor, wherein the processor can execute the following steps according to the instructions when being implemented: acquiring a plurality of service data; each service data in the plurality of service data respectively carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data; according to a preset construction rule, constructing a target adjacency matrix associated with the plurality of business data by utilizing the plurality of business data; carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth; performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction; and determining a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data.

In order to more accurately complete the above instructions, referring to fig. 7, another specific server is provided in this embodiment of the present disclosure, where the server includes a network communication port 71, a processor 72, and a memory 73, and the above structures are connected by an internal cable, so that each structure may perform specific data interaction.

The network communication port 71 may be specifically configured to obtain a plurality of service data; each service data in the plurality of service data carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data.

The processor 72 may be specifically configured to construct, according to a preset construction rule, a target adjacency matrix associated with the plurality of service data by using the plurality of service data; carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth; performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction; and determining a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data.

The memory 73 may be used for storing a corresponding program of instructions.

In this embodiment, the network communication port 71 may be a virtual port that binds with different communication protocols, so that different data may be sent or received. For example, the network communication port may be a port responsible for performing web data communication, a port responsible for performing FTP data communication, or a port responsible for performing mail data communication. The network communication port may also be an entity's communication interface or a communication chip. For example, it may be a wireless mobile network communication chip, such as GSM, CDMA, etc.; it may also be a Wifi chip; it may also be a bluetooth chip.

In this embodiment, the processor 72 may be implemented in any suitable manner. For example, the processor may take the form of, for example, a microprocessor or processor, and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a programmable logic controller, and an embedded microcontroller, among others. The description is not intended to be limiting.

In the present embodiment, the memory 73 may include a plurality of layers, and in a digital system, it may be a memory as long as it can hold binary data; in an integrated circuit, a circuit with a memory function without a physical form is also called a memory, such as a RAM, a FIFO, etc.; in the system, the storage device in physical form is also called a memory, such as a memory bank, a TF card, and the like.

The embodiments of the present specification also provide a computer storage medium storing computer program instructions that when executed implement a method of determining a data relationship as described above: acquiring a plurality of service data; each service data in the plurality of service data respectively carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data; according to a preset construction rule, constructing a target adjacency matrix associated with the plurality of business data by utilizing the plurality of business data; carrying out left-multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth; performing right multiplication iteration processing for the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction; and determining a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data.

In the present embodiment, the storage medium includes, but is not limited to, a random access Memory (Random Access Memory, RAM), a Read-Only Memory (ROM), a Cache (Cache), a Hard Disk (HDD), or a Memory Card (Memory Card). The memory may be used to store computer program instructions. The network communication unit may be an interface for performing network connection communication, which is set in accordance with a standard prescribed by a communication protocol.

In this embodiment, the functions and effects of the program instructions stored in the computer storage medium may be explained in comparison with other embodiments, and are not described herein.

Referring to fig. 8, on a software level, the embodiment of the present disclosure further provides a data relationship determining apparatus. The device may comprise in particular the following structural modules.

The acquiring module 81 may be specifically configured to acquire a plurality of service data; each service data in the plurality of service data respectively carries an identity of a data object initiating the service data and an identity of a data object receiving the service data, and the data object initiating the service data points to the data object receiving the service data through the service data;

The construction module 82 may be specifically configured to construct, according to a preset construction rule, a target adjacency matrix associated with the plurality of service data by using the plurality of service data;

the first processing module 83 may be specifically configured to perform a pre-set number of left-multiplication iteration processing on the target adjacency matrix according to a pre-set iteration rule, so as to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating the identity of the data object meeting the requirements along the forward direction, and the preset times are determined according to the preset relation depth;

the second processing module 84 may be specifically configured to perform a right multiplication iteration process on the target adjacency matrix for a preset number of times according to a preset iteration rule, so as to obtain a target right multiplication result vector; the data elements in the target right multiplication result vector are used for indicating the identity of the data object meeting the requirement along the reverse direction;

the determining module 85 may be specifically configured to determine a relationship circle formed between data objects based on the service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, and the plurality of service data.

In one embodiment, the above construction module 82 may be implemented according to the following procedure: establishing a corresponding relation between a matrix row number and an identity of a data object initiating service data, and establishing a corresponding relation between a matrix column number and an identity of a data object receiving the service data; and determining the data value of the data element at the position indicated by the combination of the corresponding matrix row number and the matrix column number in the matrix according to the service data.

In one embodiment, after constructing the target adjacency matrix associated with the plurality of service data by using the plurality of service data according to a preset construction rule, the apparatus is further configured to store the target adjacency matrix in a memory in a sparse format when the apparatus is implemented.

In one embodiment, when the first processing module 83 is implemented, the following procedure may be performed: according to a preset iteration rule, carrying out the current left-multiplication iteration in preset times on the target adjacent matrix to obtain a current left-multiplication result vector, wherein the method comprises the following steps of: obtaining a last left multiplication result vector; using the last left multiplication result vector to multiply the target adjacent matrix left to obtain the current left multiplication result vector; the current time of the left multiplication result vector is used for indicating that the identity of the data object pointed by the adjacent data object through the service data exists in the forward direction under the corresponding relation depth of the current time; the current time is less than or equal to the preset time.

In one embodiment, the second processing module 84 may be implemented according to the following procedure: according to a preset iteration rule, carrying out right multiplication iteration of the current time in preset times on the target adjacent matrix to obtain a right multiplication result vector of the current time, wherein the right multiplication result vector comprises the following steps: obtaining the right multiplication result vector of the last time; right multiplying the target adjacent matrix by using the transpose vector of the right multiplication result vector of the last time to obtain the right multiplication result vector of the current time; the right multiplication result vector of the current time is used for indicating that the identity of the data object pointed by the adjacent data object through the service data exists in the reverse direction under the corresponding relation depth of the current time; the current time is less than or equal to the preset time.

In one embodiment, the determining module 85 may be implemented as follows: performing dot multiplication processing on the target left-multiplication result vector and the target right-multiplication result vector to determine the identity of a first data object forming a first relationship circle; the first relationship circle is a relationship circle with a relationship depth smaller than or equal to a preset relationship depth; according to the identity of the first data object, service data related to the first data object is screened out from a plurality of service data to serve as first service data; and determining a plurality of first relationship circles according to the identity of the first data object and the first service data.

In an embodiment, when the determining module 85 is specifically implemented, the GPU may be invoked by a preset deep learning framework to generate a plurality of first relationship circles according to the identity of the first data object and the first service data.

In one embodiment, the apparatus is further configured to filter, when the apparatus is specifically implemented, the identity of the first data object according to the generated plurality of first relationship circles.

In one embodiment, after determining a plurality of first relationship circles according to the identity of the first data object and the first service data, the apparatus may be further configured to determine, when implemented, relationship circles corresponding to different relationship depths from the plurality of first relationship circles through aggregation iteration.

In one embodiment, the business data may include transfer transaction data and the like, and the data object may include a transaction account and the like.

In one embodiment, the apparatus may be further configured to determine whether a preset transaction risk exists in the transaction account according to a relationship circle formed between the transaction accounts based on the transfer transaction data.

It should be noted that, the units, devices, or modules described in the above embodiments may be implemented by a computer chip or entity, or may be implemented by a product having a certain function. For convenience of description, the above devices are described as being functionally divided into various modules, respectively. Of course, when the present description is implemented, the functions of each module may be implemented in the same piece or pieces of software and/or hardware, or a module that implements the same function may be implemented by a plurality of sub-modules or a combination of sub-units, or the like. The above-described apparatus embodiments are merely illustrative, for example, the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.

From the above, the data relationship determining device provided in the embodiments of the present disclosure can effectively reduce the data processing amount, efficiently find the relationship circle between the data objects from a large amount of service data, improve the processing efficiency, and solve the technical problems of large data processing amount and low processing efficiency in the existing method.

The specification also provides a device for determining the ring diagram, which specifically may include the following structural modules: the acquisition module is used for acquiring node data and edge data; the construction module is used for constructing a target adjacency matrix by utilizing the edge data and the node data according to a preset construction rule; the first processing module is used for carrying out the left-multiplication iteration processing of the preset times on the target adjacent matrix according to a preset iteration rule to obtain a target left-multiplication result vector; the data elements in the target left-hand multiplication result vector are used for indicating nodes meeting requirements in the forward direction; the second processing module is used for carrying out right multiplication iteration processing on the target adjacent matrix for preset times according to a preset iteration rule to obtain a target right multiplication result vector; wherein the data elements in the target right-hand product result vector are used to indicate nodes that meet the requirements in the reverse direction; and the determining module is used for determining a ring graph formed by the node data and the edge data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, the node data and the edge data.

In a specific scenario example, the method for determining the relationship data provided in the present specification may be based on a method for efficiently finding a fund relationship loop chart with a risk of fund reflow in a credit scenario by using massive data, so as to perform corresponding risk management.

First, most of the existing methods adopt traditional loop graph searching, and are generally implemented by adopting tools of graph calculation engines (such as GraphX, neo4J, etc.). Such tools still encounter bottlenecks in the performance of such graph computation engines when the amount of data is large or the degree of the graph is deep in a graph lookup scenario (e.g., a graph such as that shown in FIG. 9). The reason is that under such algorithm logic, the complexity of the loop-map query grows exponentially with depth.

In the scene example, a brand-new sparse matrix operation-based method is provided and applied, a relation with fund reflux is found out by searching a loop diagram, specifically, the GPU can be called for parallel calculation based on a deep learning framework such as PyTorch, tensorFlow, the calculation speed is improved, and no matter how many degrees of loops need to be inquired, the loop can be efficiently realized through iteration of corresponding times.

In particular, referring to fig. 10, the following steps may be included.

S101: all points are numbered from 0 as point IDs. Wherein each point corresponds to an identity of an account.

S102: a sparse matrix (adjacent matrix) with data values of 0 and 1 of a data element in the shape of n multiplied by n is constructed according to the ID and the edge data, and is denoted by a symbol A. Where n is the number of accounts.

Specifically, if the edge 0 points to 1, then the value of the element at the corresponding coordinate (0, 1) in the matrix is 1. If there is no pointing relationship between points x and y (i.e., no transaction data for the pointing relationship described above) then 0 is at coordinate (x, y). In practice, this matrix is stored in the memory in a sparse format, and only coordinates with element values of 1 are recorded. Such as:

tensor(indices＝tensor([[53259,32798,39160,...,666,37708,4313],

[13625,32610,33357,...,21206,4190,4190]]),

values＝tensor([1,1,1,...,1,1,1]),

size＝(67632,67632),nnz＝69618,layout＝torch.sparse_coo)。

wherein the indices are a 2 row m column vector representing the coordinates of the element with value 1 in the matrix, the first row is a numbered list of the start point and the second row is a numbered list of the end point. values are a list of all 1's with columns of the number m of edges. Corresponding to the value at each location in the indices.

S103: the iteration is performed according to the following steps. The points that make up the ring are found. This step, when embodied, may include the following.

S103-1: a column vector u of all 1 elements and a row vector v of all 1 elements are initialized. Expressed in the following forms: u= (1,) 1)', v= (1,) 1.

S103-2: the adjacent matrix D is respectively multiplied by u to the right and multiplied by v to the left, so that new u and v are obtained: u=σ (Du).

At this time, if some rows in the adjacency matrix (assuming that the ith row) are all 0 (indicating that the ith point does not point to any other point), then after the first iteration the value of the ith element of the u vector will be 0.

The right multiplication operation is performed: v=σ (vD).

At this time, if some columns in the adjacency matrix (assuming that the j-th column) are all 0 (indicating that no other points point to the j-th point), then after the first iteration, the value of the j-th element of the v vector will be 0. Indicating that this point is also filtered out.

That is, each iteration filters out the points that are currently left over, and the points that are not pointed to by any other points in this state.

The sigma is a 0-1 function, 1 is taken for elements in the vector that are greater than zero, otherwise 0 is taken.

S103-3: the step S302 is repeatedly performed n times, the number of n being the depth of the desired ring.

S103-4: and (3) performing point multiplication on the iterated vector in the dimension 1 multiplied by the iterated vector in the dimension 1: x=u' ·v.

After n iterations, the remaining points in u and v are intersected, and only the points in two cases remain, namely the points forming the replacement (the points are sequentially circularly represented by ring members and are not filtered out), and the points capable of connecting the two rings (the points forming the rings are not disappeared and are not disappeared, and the points are not disappeared, so the points are not disappeared).

The point corresponding to the element that is not 0 in the finally calculated vector x is the point that constitutes the loop or connects the two loops. In this step, the GPU is called to calculate by means of a deep learning framework such as Tensorflow, pyTorch, and the ring query of hundreds of degrees can be completed within 1 second.

S104, according to the steps of S103, filtering out non-looped edges by iteratively screening out points X (points represented by elements which are not 0 in X) through a matrix.

The specific filtering method is to screen out only the edges of which the starting point and the ending point belong to X (the visualization can be directly performed up to this point, all the rings with n degrees and less are shown). 7 ten thousand intra-edge ring queries were tested herein and iterated using the GPU version of pytorch. All turns at 3 degrees are output with only 9 ms.

S105: the loops of 2 to n degrees are sequentially output by polymerization iterations. Specific algorithmic processes include the following.

S105-1: each point in X is added with a sub-number, represented by the ID set.

S105-2: the sub-numbers of the start point are aggregated according to the end point (stored in key-value form, key being the ID of the end point; value being the set of all sub-numbers pointing to the start point of the end point).

S105-3: updating the secondary number of the current terminal point to be the aggregated secondary number set.

S105-4: it is determined whether the number of each point belongs to its sub-number set, and if so, the number IDs of the points are output, and the points indicating the numbers belong to a member of an i-degree ring.

S105-5: steps S502 to S504 are repeatedly performed until an n-degree loop is output.

S106: and outputting all rings according to the connection relation (edge) file between the ring member points at each degree output by the S104 and the points output by the S103, or directly inputting the rings into the visualization tool for displaying.

And further analyzing whether the risk of funds backflow exists according to the searched ring, and performing corresponding risk management and control.

Through the above scene example, it is verified that the method provided in the present specification has the following features:

1. the calculation speed is faster. The points which only form the ring can be quickly screened out based on matrix iteration, and GPU resources can be called for acceleration.

2. Rings of any degree can be output. All ring graphs between two degrees and n degrees can be searched out once only by setting the number of iterations n. And the calculation speed only increases linearly, not exponentially. The time taken is on the order of seconds.

The method provided by the specification can be directly applied to the loop diagram searching scene with any degree on one hand; the method can also be used for loop preliminary screening for searching complex loop diagrams (such as directed time sequence loop diagrams, multi-path time sequence flow loop diagrams and the like), can greatly reduce the edge searching space, and only leaves a loop forming edge for downstream logic processing. The calculation efficiency of various ring graphs is greatly improved.

Although the present description provides method operational steps as described in the examples or flowcharts, more or fewer operational steps may be included based on conventional or non-inventive means. The order of steps recited in the embodiments is merely one way of performing the order of steps and does not represent a unique order of execution. When implemented by an apparatus or client product in practice, the methods illustrated in the embodiments or figures may be performed sequentially or in parallel (e.g., in a parallel processor or multi-threaded processing environment, or even in a distributed data processing environment). 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, it is not excluded that additional identical or equivalent elements may be present in a process, method, article, or apparatus that comprises a described element. The terms first, second, etc. are used to denote a name, but not any particular order.

Those skilled in the art will also appreciate that, in addition to implementing the controller in a pure computer readable program code, it is well possible to implement the same functionality by logically programming the method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Such a controller can be regarded as a hardware component, and means for implementing various functions included therein can also be regarded as a structure within the hardware component. Or even means for achieving the various functions may be regarded as either software modules implementing the methods or structures within hardware components.

The description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, classes, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

From the above description of embodiments, it will be apparent to those skilled in the art that the present description may be implemented in software plus a necessary general hardware platform. Based on such understanding, the technical solutions of the present specification may be embodied essentially in the form of a software product, which may be stored in a storage medium, such as a ROM/RAM, a magnetic disk, an optical disk, etc., and include several instructions to cause a computer device (which may be a personal computer, a mobile terminal, a server, or a network device, etc.) to perform the methods described in the various embodiments or portions of the embodiments of the present specification.

Various embodiments in this specification are described in a progressive manner, and identical or similar parts are all provided for each embodiment, each embodiment focusing on differences from other embodiments. The specification is operational with numerous general purpose or special purpose computer system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.

Although the present specification has been described by way of example, it will be appreciated by those skilled in the art that there are many variations and modifications to the specification without departing from the spirit of the specification, and it is intended that the appended claims encompass such variations and modifications as do not depart from the spirit of the specification.

Claims

1. A method for determining a data relationship, comprising:

determining a relationship circle formed between data objects based on service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data;

wherein determining a relationship circle formed between data objects based on service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data comprises: performing dot multiplication processing on the target left-multiplication result vector and the target right-multiplication result vector to determine the identity of a first data object forming a first relationship circle; the first relationship circle is a relationship circle with a relationship depth smaller than or equal to a preset relationship depth; according to the identity of the first data object, service data related to the first data object is screened out from a plurality of service data to serve as first service data; and determining a plurality of first relationship circles according to the identity of the first data object and the first service data.

2. The method of claim 1, wherein constructing a target adjacency matrix associated with the plurality of business data using the plurality of business data according to a preset construction rule comprises:

3. The method of claim 1, wherein after constructing the target adjacency matrix associated with the plurality of business data using the plurality of business data according to a preset construction rule, the method further comprises: and storing the target adjacent matrix in a memory in a sparse format.

4. The method of claim 1, wherein performing a pre-set number of left-hand iterations on the target adjacency matrix according to a pre-set iteration rule to obtain a target left-hand result vector, comprises:

Obtaining a last left multiplication result vector;

5. The method of claim 1, wherein performing a right-multiplication iteration process on the target adjacency matrix for a preset number of times according to a preset iteration rule to obtain a target right-multiplication result vector, includes:

obtaining the right multiplication result vector of the last time;

6. The method of claim 1, wherein determining a plurality of first relationship circles based on the identity of the first data object and the first business data comprises:

7. The method of claim 6, wherein after invoking the GPU to generate a plurality of first relationship circles from the identity of the first data object and the first service data through a preset deep learning framework, the method further comprises:

8. The method of claim 1, wherein after determining a plurality of first relationship circles based on the identity of the first data object and the first traffic data, the method further comprises:

9. The method of claim 1, wherein the business data comprises transfer transaction data and the data object comprises a transaction account.

10. The method of claim 9, wherein after determining a relationship circle formed between data objects based on traffic data based on the target left-hand multiplication result vector, the target right-hand multiplication result vector, and the plurality of traffic data, the method further comprises:

11. A method of determining a torus map, comprising:

acquiring node data and edge data;

Determining a ring graph formed by node data and edge data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, the node data and the edge data;

the method for determining the ring graph formed by the node data and the edge data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector, the node data and the edge data comprises the following steps: performing dot multiplication processing on the target left-multiplication result vector and the target right-multiplication result vector to determine first node data forming a ring graph; selecting edge data related to the first node data from the edge data according to the first node data, and taking the edge data as first edge data; and determining a plurality of ring graphs according to the first node data and the first edge data.

12. A data relationship determination apparatus, comprising:

the determining module is used for determining a relationship ring formed between data objects based on service data according to the target left-hand multiplication result vector, the target right-hand multiplication result vector and the plurality of service data;

the determining module is specifically configured to perform dot multiplication processing on the target left-multiplication result vector and the target right-multiplication result vector, so as to determine an identity of a first data object forming a first relationship circle; the first relationship circle is a relationship circle with a relationship depth smaller than or equal to a preset relationship depth; according to the identity of the first data object, service data related to the first data object is screened out from a plurality of service data to serve as first service data; and determining a plurality of first relationship circles according to the identity of the first data object and the first service data.

13. A server comprising a processor and a memory for storing processor-executable instructions, which when executed by the processor implement the steps of the method of any one of claims 1 to 10.

14. A computer readable storage medium having stored thereon computer instructions which when executed implement the steps of the method of any of claims 1 to 10.