CN112245937A - Resource feature extraction method, device, equipment and medium based on complex network - Google Patents

Resource feature extraction method, device, equipment and medium based on complex network Download PDF

Info

Publication number
CN112245937A
CN112245937A CN202011287407.9A CN202011287407A CN112245937A CN 112245937 A CN112245937 A CN 112245937A CN 202011287407 A CN202011287407 A CN 202011287407A CN 112245937 A CN112245937 A CN 112245937A
Authority
CN
China
Prior art keywords
node
sub
network
resource transfer
resource
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202011287407.9A
Other languages
Chinese (zh)
Inventor
胡雨松
李晓森
陈守志
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Tencent Technology Shenzhen Co Ltd
Original Assignee
Tencent Technology Shenzhen Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Tencent Technology Shenzhen Co Ltd filed Critical Tencent Technology Shenzhen Co Ltd
Priority to CN202011287407.9A priority Critical patent/CN112245937A/en
Publication of CN112245937A publication Critical patent/CN112245937A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
    • A63F13/00—Video games, i.e. games using an electronically generated display having two or more dimensions
    • A63F13/85—Providing additional services to players
    • A—HUMAN NECESSITIES
    • A63—SPORTS; GAMES; AMUSEMENTS
    • A63F—CARD, BOARD, OR ROULETTE GAMES; INDOOR GAMES USING SMALL MOVING PLAYING BODIES; VIDEO GAMES; GAMES NOT OTHERWISE PROVIDED FOR
    • A63F13/00—Video games, i.e. games using an electronically generated display having two or more dimensions
    • A63F13/70—Game security or game management aspects
    • A63F13/79—Game security or game management aspects involving player-related data, e.g. identities, accounts, preferences or play histories
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Software Systems (AREA)
  • Evolutionary Computation (AREA)
  • Mathematical Physics (AREA)
  • Data Mining & Analysis (AREA)
  • Physics & Mathematics (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Business, Economics & Management (AREA)
  • Computer Security & Cryptography (AREA)
  • General Business, Economics & Management (AREA)
  • Artificial Intelligence (AREA)
  • Data Exchanges In Wide-Area Networks (AREA)

Abstract

The application discloses a resource feature extraction method, device, equipment and medium based on a complex network, and relates to the field of complex networks. The method comprises the following steps: acquiring a resource transfer relationship; constructing a resource transfer network according to the resource transfer relationship, wherein the resource transfer network comprises at least two nodes, the nodes are used for representing user accounts for carrying out resource transfer, directed edges are correspondingly arranged between the nodes, and the directed edges are used for representing the resource flow direction in the resource transfer; segmenting the resource transfer network according to a preset sub-network structure to obtain at least two segmented sub-network structures, wherein different preset sub-network structures are used for representing resource transfer relations of different topology types; and obtaining the resource characteristics corresponding to the resource transfer network according to the topology type to which the segmented sub-network structure belongs. Comprehensive resource characteristics can be extracted from the resource transfer network, so that the efficiency and the accuracy of identifying abnormal resource transfer are improved.

Description

Resource feature extraction method, device, equipment and medium based on complex network
Technical Field
The present application relates to the field of complex networks, and in particular, to a resource feature extraction method, apparatus, device, and medium based on a complex network.
Background
The complex network is a network with high complexity and has the characteristics of huge number of nodes and various different characteristics of a network structure.
Taking the complex network as a resource transfer network as an example, in the resource transfer network, each node represents a transfer user for transferring resources, and a connection line between the nodes represents a resource transfer relationship. Part of the nodes and part of the connecting lines form a sub-network structure (Motif), and the resource transfer network comprises a plurality of sub-network structures which repeatedly appear. In the related art, the sub-network structure is used as the resource feature, and the frequency of the occurrence of a plurality of sub-network structures in the resource transfer network is calculated, so that whether the resource transfer represented by the sub-network structure is abnormal or not can be judged.
In the above technical solution, since the nodes at different positions may have different topological characteristics, the computer device may not accurately determine whether the resource transfer is abnormal.
Disclosure of Invention
The embodiment of the application provides a resource feature extraction method, a resource feature extraction device, a resource feature extraction equipment and a resource feature extraction medium based on a complex network. By dividing the resource transfer network by using the preset sub-networks representing the resource transfer relations of different transaction types, comprehensive resource characteristics can be extracted from the resource transfer network, so that the efficiency and accuracy of identifying abnormal resource transfer are improved. The technical scheme comprises the following steps:
according to an aspect of the present application, there is provided a resource feature extraction method based on a complex network, the method including:
acquiring a resource transfer relationship;
constructing a resource transfer network according to the resource transfer relationship, wherein the resource transfer network comprises at least two nodes, the nodes are used for representing user accounts for resource transfer, directed edges are correspondingly arranged between the nodes, and the directed edges are used for representing the resource flow direction in the resource transfer;
segmenting the resource transfer network according to a preset sub-network structure to obtain at least two segmented sub-network structures, wherein different preset sub-network structures are used for representing resource transfer relations of different topology types;
and obtaining the resource characteristics corresponding to the resource transfer network according to the topology type of the segmented sub-network structure.
According to another aspect of the present application, there is provided a resource feature extraction apparatus based on a complex network, the apparatus including:
the acquisition module is used for acquiring the resource transfer relationship;
the construction module is used for constructing a resource transfer network according to the resource transfer relationship, the resource transfer network comprises at least two nodes, the nodes are used for representing user accounts for resource transfer, directed edges are correspondingly arranged between the nodes, and the directed edges are used for representing fund flow direction in the resource transfer;
the segmentation module is used for segmenting the resource transfer network according to a preset sub-network structure to obtain at least two segmented sub-network structures, and different preset sub-network structures are used for representing resource transfer relations of different topology types;
and the feature extraction module is used for obtaining the resource features corresponding to the resource transfer network according to the topology type of the segmented sub-network structure.
In an optional embodiment, the feature extraction module is configured to calculate, in the at least two segmented sub-network structures, an accumulated number of sub-network structures belonging to each topology type; and obtaining the resource characteristics corresponding to the resource transfer network according to the accumulated number of the sub-network structures belonging to each topology type.
In an optional embodiment, the resource transfer network includes n nodes, where n is a positive integer;
the acquisition module is used for acquiring a first set corresponding to an ith node, the first set comprises a first neighbor node set and a first directed edge set, the first neighbor node set represents a set corresponding to a first neighbor node connected with the ith node, the node identification of the first neighbor node is larger than that of the ith node, the first directed edge set represents a set corresponding to a directed edge between the ith node and the first neighbor node, i is less than or equal to n, and i is a positive integer; acquiring a second set corresponding to the first neighbor node, wherein the second set comprises a second neighbor node set and a second directed edge set, the second neighbor node set represents a set corresponding to a second neighbor node connected with the first neighbor node, and the second directed edge set represents a set corresponding to a directed edge between the first neighbor node and the second neighbor node; the feature extraction module is configured to repeat the step of obtaining the first set and the second set until the n nodes are traversed, and calculate the cumulative number of the sub-network structures belonging to each topology type according to the first set and the second set respectively corresponding to the n nodes.
In an optional embodiment, the obtaining module is configured to, in response to that an intersection set of the first set and the second set is a non-empty set and a node identifier of the second neighbor node is greater than a maximum node identifier, obtain a directed edge between the ith node, the first neighbor node, and the second neighbor node, where the maximum node identifier is a largest node identifier of the node identifiers of the ith node and the first neighbor node; acquiring a sub-network structure formed by the ith node, the first neighbor node and the second neighbor node according to directed edges among the ith node, the first neighbor node and the second neighbor node; the feature extraction module is used for calculating the accumulated number of the sub-network structures belonging to each topological type in the sub-network structures formed by the three modules.
In an optional embodiment, the obtaining module is configured to, in response to that an intersection set of the first set and the second set is an empty set, and the second neighbor node is connected to the first neighbor node and not connected to the ith node, obtain a second node identifier of the second neighbor node and a third node identifier of the ith node; the feature extraction module is configured to, in response to that the second node identifier is greater than the third node identifier, obtain a first directed edge between the ith node and the first neighboring node, and a second directed edge between the first neighboring node and the second neighboring node; acquiring a first sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the second directed edge; the feature extraction module is configured to calculate, in the first sub-network structure, an accumulated number of sub-network structures belonging to each topology type with the ith node as a target node; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node.
In an optional embodiment, the obtaining module is configured to obtain, in response to the second node identifier being smaller than the third node identifier, a first directed edge between the ith node and the first neighboring node, and a second directed edge between the first neighboring node and the second neighboring node; acquiring a first sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the second directed edge; the feature extraction module is configured to calculate, in the first sub-network structure, the cumulative number of sub-network structures belonging to each topology type with the ith node as a target node.
In an optional embodiment, the obtaining module is configured to obtain a second node identifier of the second neighboring node and a first node identifier of the first neighboring node in response to that an intersection set of the first set and the second set is an empty set, and the second neighboring node is connected to the ith node and not connected to the first neighboring node; responding to the second node identification larger than the first node identification, and acquiring a first directed edge between the ith node and the first neighbor node and a third directed edge between the ith node and the second neighbor node; acquiring a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the third directed edge; the feature extraction module is configured to calculate, in the second sub-network structure, the cumulative number of sub-network structures belonging to each topology type with the ith node as a target node; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node.
In an optional embodiment, the obtaining module is configured to obtain, in response to the second node identifier being smaller than the first node identifier, a first directed edge between the ith node and the first neighboring node, and a third directed edge between the ith node and the second neighboring node; acquiring a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the third directed edge; the feature extraction module is configured to calculate, in the second sub-network structure, the cumulative number of sub-network structures belonging to each topology type with the first neighboring node as a target node.
In an optional embodiment, the directed edges between the nodes correspond to weight information;
and the feature extraction module is used for obtaining the resource features corresponding to the resource transfer network according to the type of the segmented sub-network structure and the weight information corresponding to the directed edge.
In an optional embodiment, the feature extraction module is configured to calculate, in the at least two segmented sub-network structures, an accumulated number of sub-network structures belonging to each topology type; calculating the weight value corresponding to the sub-network structure belonging to the same type according to the weight information corresponding to the directed edge; and obtaining the resource characteristics corresponding to the resource transfer network according to the accumulated quantity and the weight value.
In an optional embodiment, the resource transfer network includes n nodes, where n is a positive integer;
the obtaining module is configured to obtain a first adjacent triple set corresponding to an ith node, where the first adjacent triple set includes a first weight set, the first weight set represents a weight value corresponding to a directed edge between the ith node and the first neighbor node, i is not greater than n, and i is a positive integer; acquiring a second adjacent triple set corresponding to the first neighbor node, wherein the second adjacent triple set comprises a second weight set, the second weight set represents a weight value corresponding to a directed edge between the first neighbor node and a second neighbor node, and the second neighbor node is connected with the first neighbor node; repeating the steps of obtaining the first adjacent triple set and the second adjacent triple set until the n nodes are traversed, and calculating the corresponding weight values of the sub-network structures belonging to the same type according to the weight values corresponding to the directed edges between the n nodes.
In an alternative embodiment, the apparatus includes a calling module;
the calling module is used for calling a resource classification model to classify the resource characteristics to obtain the prediction probability that the resource transfer belongs to the abnormal resource transfer; and outputting the type of the resource transfer according to the prediction probability.
According to another aspect of the present application, there is provided a computer device comprising a processor and a memory, the memory having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by the processor to implement the complex network based resource feature extraction method according to the above aspect.
According to another aspect of the present application, there is provided a computer-readable storage medium having stored therein at least one instruction, at least one program, code set, or set of instructions that is loaded and executed by a processor to implement the complex network based resource feature extraction method according to the above aspect.
According to another aspect of the application, a computer program product or computer program is provided, comprising computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions to cause the computer device to perform the complex network-based resource extraction method as described above.
The beneficial effects brought by the technical scheme provided by the embodiment of the application at least comprise:
the resource transfer network is constructed, the resource features are extracted according to the type of the preset sub-network structure to which the sub-network structure in the resource transfer network belongs, the resource transfer network is divided according to the preset sub-network structures representing different types of resource transfer relations, and the extracted resource features can accurately represent various resource transfer relations, so that the accuracy of judging whether abnormal resource transfer exists in the resource transfer network by using the extracted resource features is improved, and the efficiency of identifying abnormal resource transfer is improved.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 is a schematic diagram of a default subnetwork structure provided in an exemplary embodiment of the present application;
FIG. 2 is a block diagram of a computer system provided in an exemplary embodiment of the present application;
FIG. 3 is a flowchart of a complex network-based resource feature extraction method according to an exemplary embodiment of the present application;
FIG. 4 is a system framework diagram for extracting resource features provided by an exemplary embodiment of the present application;
FIG. 5 is a flowchart of a complex network-based resource feature extraction method according to another exemplary embodiment of the present application;
FIG. 6 is a schematic illustration of resource features provided by an exemplary embodiment of the present application;
FIG. 7 is a flowchart of a complex network-based resource feature extraction method according to another exemplary embodiment of the present application;
FIG. 8 is a flowchart of a complex network-based resource feature extraction method according to another exemplary embodiment of the present application;
FIG. 9 is a flowchart of a complex network-based resource feature extraction method according to another exemplary embodiment of the present application;
FIG. 10 is a flowchart of a complex network-based resource feature extraction method according to another exemplary embodiment of the present application;
FIG. 11 is a flowchart of a complex network-based resource feature extraction method according to another exemplary embodiment of the present application;
FIG. 12 is a flowchart of a complex network-based resource feature extraction method according to another exemplary embodiment of the present application;
FIG. 13 is a flowchart of a complex network-based resource feature extraction method according to another exemplary embodiment of the present application;
FIG. 14 is a schematic illustration of resource features provided by another exemplary embodiment of the present application;
fig. 15 is a schematic structural diagram of a resource feature extraction apparatus based on a complex network according to an exemplary embodiment of the present application;
fig. 16 is a schematic structural diagram of a server according to an exemplary embodiment of the present application.
Detailed Description
To make the objects, technical solutions and advantages of the present application more clear, embodiments of the present application will be described in further detail below with reference to the accompanying drawings.
First, terms related to embodiments of the present application will be described.
Complex networks (ComplexNetwork): refers to a network exhibiting a high degree of complexity, which is manifested primarily as: 1. the network structure is complex, the number of nodes in the network is huge, and the network structure presents various different characteristics; 2. the connection between nodes may have directionality (the connection between nodes is named as an edge, and the edge with direction is a directed edge), the edges between nodes may have corresponding weights, and the weights corresponding to the edges of different nodes have differences; 3. the nodes can represent anything, for example, the nodes represent individual individuals, and the complex network formed by the nodes is an interpersonal relationship network.
In the embodiment of the application, a complex network is taken as a resource transfer network as an example, nodes in the resource transfer network represent user accounts for resource transfer, directed edges exist among the nodes, and the directed edges represent the flow direction of resources when the resources are transferred among the user accounts. For example, if the user account a transfers money to the user account b, the directed edge points from the user account a to the user account b. Illustratively, the directed edges between the nodes correspond to weight information, which includes a resource transfer object, a resource transfer type (e.g., presenting a virtual gift or a virtual red packet representing a certain resource value, transferring a resource from one user account to another user account, etc.), a resource transfer value, a resource transfer frequency, a resource transfer time, and a resource transfer message (i.e., remark information filled by any one of two parties participating in the resource transfer during the resource transfer).
Artificial Intelligence (AI): the method is a theory, method, technology and application system for simulating, extending and expanding human intelligence by using a digital computer or a machine controlled by the digital computer, sensing the environment, acquiring knowledge and obtaining the best result by using the knowledge. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making.
The artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and the like.
Machine Learning (ML): the method is a multi-field cross discipline and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, complex networks, reinforcement learning, transfer learning, inductive learning, and teaching learning.
With the research and progress of artificial intelligence technology, the artificial intelligence technology develops research and application in a plurality of fields, such as common smart homes, smart wearable devices, virtual assistants, smart speakers, smart marketing, unmanned driving, automatic driving, unmanned aerial vehicles, robots, smart medical care, smart customer service, transaction risk monitoring and the like.
The scheme provided by the embodiment of the application relates to a resource feature extraction method based on a complex network, and is specifically explained by the following embodiment.
Regarding the characteristics of resource transfer, the embodiment of the present application enumerates 33 predetermined sub-network structures (Motif), as shown in fig. 1. The 33 preset sub-network structures represent different types of resource transfer, and the resource transfer network is segmented through the 33 sub-network structures, so that the cumulative number of the sub-network structures belonging to each topology type is calculated to obtain the resource characteristics corresponding to the resource transfer network.
As shown in fig. 1, 33 predetermined sub-network structures (Motif) are numbered, wherein the predetermined sub-network structures No. 1, No. 2 and No. 4 each include two nodes, which represent resource transfer between two user accounts. The black nodes represent the target nodes currently analyzed, and the white nodes represent neighbor nodes connected with the target nodes. The 1 st and 2 nd default sub-network structures both represent unidirectional (different directions) resource transfer, and the 4 th default sub-network structure is bidirectional resource transfer.
The remaining numbered sub-network structures all contain three nodes, i.e., resource transfers between three user accounts. The subnet configurations No. 3 to 9 and No. 15 to 22 include two directed edges, and the subnet configurations No. 10 to 13 and No. 23 to 33 include three directed edges.
The resource feature extraction method based on the complex network can be applied to computer equipment with strong data processing capacity. In a possible implementation manner, the resource feature extraction method based on the complex network provided by the embodiment of the present application may be applied to a personal computer, a workstation, or a server, that is, the resource feature may be extracted from the resource transfer network through the personal computer, the workstation, or the server. Illustratively, the resource feature extraction method based on the complex network is applied to the distributed server.
FIG. 2 illustrates a block diagram of a computer system provided in an exemplary embodiment of the present application. The computer system 100 includes a first computer device 110 and a second computer device 120.
The first computer device 110 and the second computer device 120 are two node servers in a distributed server. The network structure diagram 101 corresponding to the resource transfer network is input to the first computer device 110, and the first computer device 110 stores node information in the resource transfer network. The resource transfer network comprises at least two nodes, and the nodes represent user accounts for transferring resources; directed edges exist among the nodes, and the directed edges represent the resource flow direction when resource transfer is carried out. Illustratively, the first computer device 110 includes a Parameter Server (PS).
The first computer device 110 splits the input network structure diagram 101 according to a predetermined subnetwork structure, which is 33 representative subnetwork structures, that is, 33 types of subnetwork structures, obtained according to a common resource transfer relationship type.
The second computer device 120 is configured to pull the node information in a preset Batch Size (Batch Size), which is the Size of the node information and the directional edge information pulled by the work server per Batch. The second computer device 120 performs sub-network structure (Motif) segmentation on the pulled node information and calculates the cumulative number of sub-network structures belonging to each type. Transferring a node v in a network with resourcesiFor the target node, acquiring a target node viThe sub-network structure corresponding to the target node is obtained according to the adjacent triple set, and the accumulated number of the preset sub-network structures of each type of each topology of the sub-network structure corresponding to the target node is calculated. And in the same way, all the nodes in the resource transfer network are used as target nodes to calculate the accumulated number, so that the accumulated number meeting the structure of each type of sub-network in the resource transfer network is counted, and the resource characteristics of the resource transfer network are determined. Illustratively, the second computer device 120 includes a work server (Worker).
The first computer device 110 and the second computer device 120 are generally referred to as one or more computer devices, and the embodiment of the present application is only described by taking the first computer device 110 and the second computer device 120 as an example, the type of the computer device may be at least one of a laptop computer, a desktop computer, a tablet computer, a server computer, and a workstation computer, and the present application does not limit the type of the computer device.
For convenience of description, the following embodiments are described as examples in which the resource feature extraction method based on a complex network is performed by a parameter server and a work server.
Fig. 3 shows a flowchart of a complex network-based resource feature extraction method according to an exemplary embodiment of the present application. The embodiment is described by taking as an example that the method is used in the computer system 100 shown in fig. 2, and the method includes the following steps:
step 301, obtaining a resource transfer relationship.
The resource transfer relationship refers to a relationship between two parties performing resource transfer, and the resource may be funds (cash), game equipment (such as virtual items), game materials, game pets, game coins, icons (or signs), members, titles, value-added services, points, shoe-shaped gold ingots, gold beans, gift certificates, exchange certificates, coupons, greeting cards, and the like. The application is not limited as to the type of resource.
In the resource transfer network, nodes represent user accounts corresponding to two parties carrying out resource transfer, and directed edges between the nodes represent a resource transfer relation. The resource transfer relationship includes a resource transfer relationship between a user and a user, a resource transfer relationship between a merchant and a consumer, and a resource transfer relationship between an enterprise and an enterprise. For example, the resource transfer relationship is a transfer relationship between users, and when a user account a transfers to a user account b, the directed edge points to a node corresponding to the user account b from the node corresponding to the user account a.
In some embodiments, the resource transfer relationship further includes information such as a resource transfer value, a resource transfer time, a resource transfer message, and the resource transfer message refers to remark information filled by either one of the two parties of the resource transfer when the resource transfer is performed, and in general, the resource transfer message is remark information filled by a resource payer in the resource transfer process.
Step 302, a resource transfer network is constructed according to the resource transfer relationship, the resource transfer network includes at least two nodes, the nodes are used for representing user accounts for resource transfer, directed edges are corresponding between the nodes, and the directed edges are used for representing resource flow direction in resource transfer.
Denote the resource transfer network by G ═ (V, E), by ViE.g. V represents the node corresponding to the user account, and eijE.g. E represents a directed edge between nodes, i.e. a resource transfer relationship, EijRepresenting a node viAnd node vjWith directed edges in between. If node viAnd node vjThe resource flow between is bidirectional, i.e. the resource is by node viFlow direction node vjAnd resources are also from node vjFlow direction node viThen eijRepresenting a bidirectional edge. If node viAnd node vjThe resource flow direction between is unidirectional, i.e. the resource is routed by node viFlow direction node vj(or node v)jFlow direction node vi) Then eijRepresenting a unidirectional edge.
As shown in fig. 2, a network structure diagram 101 corresponding to the resource transfer network includes 10 nodes, and there is a directed edge between the 10 nodes. Illustratively, a directed edge between node 1 and node 2 points from node 1 to node 2, i.e., the resource flows from node 1 to node 2, a directed edge between node 2 and node 4 is bidirectional, i.e., the resource flows from node 2 to node 4, and the resource also flows from node 4 to node 2. Illustratively, there is no resource transfer relationship, i.e., there is no directed edge, between node 6 and node 8.
Step 303, segmenting the resource transfer network according to the preset sub-network structures to obtain at least two segmented sub-network structures, where different preset sub-network structures are used to represent resource transfer relationships of different topology types.
As shown in fig. 1, different predetermined subnetworks represent resource transfer relationships of different topology types, for example, the No. 33 predetermined subnetwork represents that three nodes have a resource transfer relationship with each other, and resources flow between any two nodes.
Illustratively, a resource transfer network is input into the first computer device 110, 33 preset sub-network structures as shown in fig. 1 are stored in the first computer device 110 in advance, when the first computer device 110 receives the input resource transfer network, nodes and resource transfer relations in the resource transfer network are stored, the first computer device 110 further stores the preset sub-network structures, and the second computer device 120 pulls a certain amount of node information from the first computer device 110 with a preset batch size (BatchSize) and splits the resource transfer network according to the preset sub-network structures. I.e. the second computer device 120 determines a predetermined sub-network structure comprised in the resource transfer network. Illustratively, taking the network structure diagram 101 as an example, wherein a sub-network structure composed of the node 1, the node 2, the node 3, and the node 4 is taken as an example for explanation, taking the node 1 as a target node, the network structures corresponding to the node 1 and the node 2 conform to the No. 1 predetermined sub-network structure, the network structures corresponding to the node 1 and the node 3 conform to the No. 1 predetermined sub-network structure, the network structures corresponding to the node 1 and the node 4 also conform to the No. 1 predetermined sub-network structure, the network structures corresponding to the node 1, the node 2, and the node 3 conform to the No. 25 predetermined sub-network structure, and the network structures corresponding to the node 1, the node 2, and the node 4 conform to the No. 25 predetermined sub-network structure. In this sub-network structure, two topology types of sub-network structures (No. 1 and No. 25) exist with the node 1 as a target node.
Illustratively, with the node 2 as a target node, the network structures corresponding to the node 1 and the node 2 conform to the No. 2 predetermined sub-network structure, the network structures corresponding to the node 2 and the node 3 conform to the No. 14 predetermined sub-network structure, and the network structures corresponding to the node 2 and the node 4 also conform to the No. 14 predetermined sub-network structure. In this sub-network structure, two topology types of sub-network structures (No. 2 and No. 14) exist with the node 2 as a target node.
Illustratively, taking the node 3 as a target node, the network structures corresponding to the nodes 1 and 3 conform to the predetermined sub-network structure No. 2, and the network structures corresponding to the nodes 2 and 3 conform to the predetermined sub-network structure No. 14. In this sub-network structure, two topology types of sub-network structures (No. 2 and No. 14) exist with the node 3 as a target node.
Similarly, in the manner described above, with node 4 as the target node, the first computer device 110 determines the topology type present in the sub-network structure. By analogy, each node in the resource transfer network is used as a target node, and the topology type in the sub-network structure corresponding to each target node is determined according to the above manner, so that the topology type existing in the resource transfer network is obtained.
And 304, obtaining the resource characteristics corresponding to the resource transfer network according to the topology types to which the at least two segmented sub-network structures belong.
Illustratively, according to the embodiment shown in step 303, when node 4 is taken as a target node, the network structures corresponding to node 1 and node 4 conform to the predetermined sub-network structure No. 2, and the network structures corresponding to node 2 and node 4 conform to the predetermined sub-network structure No. 14, in which two topology types (No. 2 and No. 14) of sub-network structures exist with node 4 as a target node.
Therefore, in the sub-network structure, the proportion of the sub-network structures belonging to the topology type No. 2 and the topology type No. 14 is high, and therefore, the second computer device 120 determines that the topology type No. 2 and the topology type No. 14 are both the resource features in the sub-network structure, and similarly, the second computer device 120 determines the topology types corresponding to the other sub-network structures after being split in the resource transfer network, thereby determining the resource features corresponding to the whole resource transfer network.
The resource characteristics refer to characteristics used in describing a resource transfer process, and include at least one of the following characteristics: the characteristics corresponding to the user account for resource transfer, the characteristics corresponding to the resource transfer value in the resource transfer process, the characteristics corresponding to the remark information in the resource transfer process and the characteristics corresponding to the resource transfer type.
Illustratively, taking the resource transfer as a fund transaction as an example, transfer characteristics can be extracted according to the topology types to which the at least two segmented sub-network structures belong, the transfer characteristics are classified by the resource classification model, and the transaction type is output as the transfer transaction between users.
In summary, in the method provided in this embodiment, the resource transfer network is constructed, the resource features are extracted according to the types of the preset sub-network structures to which the sub-network structures in the resource transfer network belong, and the resource transfer network is divided according to the preset sub-network structures representing different types of resource transfer relationships, so that the extracted resource features can accurately represent various resource transfer relationships, thereby improving the accuracy of determining whether the resource transfer network has abnormal resource transfer by using the extracted resource features, and improving the efficiency of identifying the abnormal resource transfer.
Fig. 4 shows a system framework diagram for extracting resource features according to an exemplary embodiment of the present application, where a network structure diagram 101 corresponding to a resource transfer network is input into a parameter server 130, nodes in the resource transfer network represent user accounts for performing resource transfer, directed edges between the nodes represent resource flow directions during resource transfer, and node serial numbers represent node identifiers.
The Parameter Server (PS) 130 stores node information and directed edge information in the resource transfer network, and the parameter server 130 also stores a preset sub-network structure as shown in fig. 1 in advance. Different preset sub-network structures represent different types of resource transfer relationships, and any node in the resource transfer network can be used as a target node. Illustratively, the parameter server 130 is a server node in a distributed machine learning framework (Angel) for reasonably partitioning a high-dimensional model into a plurality of parameter server nodes to implement an efficient machine learning algorithm. When a large-scale resource transfer network is input, the resource transfer network is segmented, and the segmented sub-networks are respectively stored on different parameter server nodes.
In the parameter server 130, several grids exist near each node, each grid including a number, and the target node is illustrated as 1. The 6 grids near the node 1 have 6 numbers, the first three numbers represent the node identification of the neighbor node of the node 1, and the last three numbers represent the indication direction of the directed edge between the target node and the neighbor node. The directed edge directed by the target node to the neighbor node is denoted by 0, the directed edge directed by the neighbor node to the target node is denoted by 1, and the bidirectional edge between the target node and the neighbor node is denoted by 2. The neighbor nodes connected with the target node comprise a node 2, a node 3 and a node 4, and the directed edges between the node 1 and the three neighbor nodes all point to the neighbor nodes from the node 1, so that the directed edges corresponding to the node 2, the node 3 and the node 4 are all represented by 0.
Illustratively, the parameter server 130 sends node information and directed side information in the resource transfer network to the computation layer logic 102, and the work server (Worker)140 obtains the node information and the directed side information from the computation logic layer and extracts resource features according to the information. Illustratively, the work server 140 is a server node (Spark) in a distributed machine learning framework.
Illustratively, the parameter server 130 directly sends the node information and the directed side information in the resource transfer network to the work server 140, and the work server 140 extracts the resource features according to the node information and the directed side information.
In some embodiments, the resource transfer network is input to the parameter server 130 and the work server 140 at the same time, the parameter server 130 stores the node information and the directed side information of the resource transfer network, and the work server 140 determines the neighbor node information and the directed side information corresponding to the target node and stores the neighbor node information and the directed side information.
The work server 140 is configured to complete the resource feature extraction process by using a preset batch size (BatchSize) to pull the required node information and directional side information, where the batch size refers to the size of the node information and the directional side information pulled by each batch of the work server. Illustratively, the BatchSize is determined synthetically by balancing the computing performance, data size, and cluster resources among the servers. Illustratively, the work server 140 employs a uniform BatchSize for the entire resource transfer network; illustratively, the work server 140 uses different BatchSize for different types of structures according to the structure of the resource transfer network; illustratively, the work server 140 uses different BatchSize for the resource transfer relationship with different importance according to the importance of the resource transfer relationship in the resource transfer network. For example, in the resource transfer network, the importance of the resource transfer relationship between an enterprise and an enterprise > the importance of the resource transfer relationship between a merchant and a consumer > the importance of the resource transfer relationship between a user and a user, so that when the work server 140 reads the resource transfer relationship between an enterprise and an enterprise, the node information can be pulled through the BatchSize 1; when the work server 140 reads the resource transfer relationship between the user and the user, the node information may be pulled by the BatchSize2, where the BatchSize1 is larger than the BatchSize2, that is, the node information pulled by the work server 140 through the BatchSize1 is larger than the node information pulled through the BatchSize 2.
In the work server 140, several grids exist near each node, each grid including or not including a number, and the node 4 is taken as a target node for explanation. The neighbor nodes of the node 4 are the node 1, the node 2 and the node 6, the working server 140 determines the neighbor nodes whose node identifiers are larger than the node identifier of the node 4 (the node identifiers are represented by the serial numbers of the nodes), the node is the node 6, and the resource transfer relationship between the node 4 and the node 6 is that the resource flow is directed from the node 4 to the node 6, so that a directed edge between the node 4 and the node 6 is represented by 0. The working server 140 adds 1 to the cumulative number of sub-network structures belonging to the topology type No. 1 in the sub-network structures corresponding to the node 4 and the node 6, with the node 4 being the target node.
And repeating the steps until all nodes in the resource transfer network are traversed, and counting the accumulated number of the sub-network structures belonging to each topology type to obtain a sub-network (Motif) calculation result 103 corresponding to the resource transfer network.
Fig. 5 shows a complex network-based resource feature extraction method according to another exemplary embodiment of the present application, which is described as an example of the method used in the computer system 100 shown in fig. 2, and the method includes the following steps:
step 501, obtaining a resource transfer relationship.
Illustratively, the system framework shown in fig. 4 is connected to a background server of an application program, the application program is an application program supporting resource transfer, and the application program includes: at least one of an instant messaging application, a payment application, a shopping application, and a take-away application. And acquiring a resource transfer relationship through a background server of the application program, wherein the resource transfer relationship comprises a resource transfer relationship between users, a resource transfer relationship between merchants and consumers and a resource transfer relationship between enterprises. The resource transfer relationship between the users comprises a transfer relationship and a donation relationship, such as the way of sending a virtual commodity package or a virtual red package to donate funds to others.
Step 502, a resource transfer network is constructed according to the resource transfer relationship, the resource transfer network includes at least two nodes, the nodes are used for representing user accounts for resource transfer, directed edges are corresponding between the nodes, and the directed edges are used for representing resource flow direction in resource transfer.
Illustratively, the background server of the application sends the resource transfer relationship to a server for constructing the resource transfer network, and the server constructs the resource transfer network.
Illustratively, the background server of the application program constructs a resource transfer network according to the acquired resource transfer relationship.
A resource transfer network constructed based on the resource transfer relationship is shown as a network structure diagram 101 in fig. 4.
Step 503, segmenting the resource transfer network according to the preset sub-network structures to obtain at least two segmented sub-network structures, where different preset sub-network structures are used to represent resource transfer relationships of different topology types.
In some embodiments, the system framework shown in fig. 4 includes a plurality of parameter server nodes and a plurality of work server nodes, and because the resource transfer network is large in scale, before the resource transfer network is input to the parameter server and the work server, the resource transfer network needs to be split, each parameter server node stores a part of the split sub-network structure, each work server node processes a part of the split sub-network structure, for example, the parameter server nodes store node information of nodes 1 to 1000, and the work server nodes process nodes of nodes 1500 to 2000.
And each working server node divides the divided sub-network structures according to the preset sub-network structures, namely determining the sub-network structures belonging to each topological type in the divided sub-network structures.
In step 504, the cumulative number of sub-network structures belonging to each topology type is calculated among the at least two sub-network structures after segmentation.
The operation process of one operation server node is taken as an example for explanation, as shown in fig. 6. The sub-network structure in fig. 6 is a partial network structure in the resource transfer network, the sub-network structure 21 is divided according to the preset sub-network structure shown in fig. 1, and the cumulative number of the sub-network structures after being divided, which belong to each topology type, is shown in the table in fig. 6. The table comprises the predetermined sub-network structures 22 and the target node 23 and also the cumulative number of sub-network structures belonging to each topology type.
Illustratively, the target node is taken as the node 1, the nodes connected to the node 1 include a node 2, a node 3 and a node 4, and in the sub-network structures corresponding to the node 1, the node 2, the node 3 and the node 4, it can be determined that the sub-network structure contains the topology type of the No. 10 predetermined sub-network structure, and there is no other topology type conforming to the predetermined sub-network structure. The number of sub-network structures corresponding to the number 10 is therefore 1.
And in the same way, each node is taken as a target node, and the accumulated number of the sub-network structures which are consistent with the topological type of the preset sub-network structures is determined. As can be seen from this, in the sub-network structures formed by the nodes 1 to 6, the cumulative number of sub-network structures belonging to the topology type No. 10 is 3, and the cumulative number of sub-network structures of this topology type is the largest.
As shown in step 503, the split sub-network structure still has more nodes, and therefore, the working server needs to pull the required node information in batch when determining the cumulative number of sub-network structures belonging to each topology type.
Illustratively, the resource transfer network includes n nodes, where n is a positive integer. And taking the ith node in the resource transfer network as a target node, and calculating the accumulated number of the sub-network structures belonging to the topology type through the ith node and the neighbor nodes connected with the ith node.
Step 504 may be replaced with the following steps:
step 5041, a first set corresponding to the ith node is obtained, the first set includes a first neighbor node set and a first directed edge set, the first neighbor node set represents a set corresponding to a first neighbor node connected to the ith node, a node identifier of the first neighbor node is larger than a node identifier of the ith node, the first directed edge set represents a set corresponding to a directed edge between the ith node and the first neighbor node, i is not larger than n, and i is a positive integer.
The resource transfer network is denoted by G ═ (V, E), including all nodes ViE.g. V, the working server obtains the node Vi(ith node) corresponding first set
Figure BDA0002782822860000161
The first set includes a first set of neighboring nodes
Figure BDA0002782822860000162
And a first set of directed edges
Figure BDA0002782822860000163
Wherein the first neighbor node is aggregated with
Figure BDA0002782822860000164
Representing, at a first set of directed edges
Figure BDA0002782822860000165
In (1),
Figure BDA0002782822860000166
dijrepresenting a node viAnd a neighbor node vjDirected edges between (first neighbor nodes), d ij0 denotes a directed edge formed by node viPointing to neighbor node vj，d ij1 denotes a directed edge formed by a node viPointing to neighbor node vj，dijDenotes node v by 2iAnd a neighbor node vjA bidirectional edge in between.
Step 5042, obtaining a second set corresponding to the first neighbor node, the second set including a second set of neighbor nodes and a second set of directed edges, the second set of neighbor nodes characterizing the set corresponding to the second neighbor nodes connected to the first neighbor node, the second set of directed edges characterizing the set corresponding to the directed edges between the first neighbor node and the second neighbor node.
With the first neighbor node vjFor a target node, a first neighbor node v is acquiredjCorresponding second set
Figure BDA0002782822860000171
The second set includes a second set of neighboring nodes
Figure BDA0002782822860000172
And a second set of directed edges
Figure BDA0002782822860000173
The second neighbor node is a node connected to the first neighbor node and has a second set of directed edges
Figure BDA0002782822860000174
In (1),
Figure BDA0002782822860000175
it is noted that step 5041 may be performed prior to step 5042 or concurrently with step 5042.
And step 5043, repeating the steps of obtaining the first set and the second set until n nodes are traversed, and calculating the cumulative number of the sub-network structures belonging to each topology type according to the first set and the second set corresponding to the n nodes respectively.
The cumulative number of sub-network structures belonging to each topology type can be calculated according to the relationship between the first set and the second set, and the steps 5041 and 5042 are repeated, all nodes in the resource transfer network are traversed, and the cumulative number of sub-network structures belonging to each topology type is calculated through the first set and the second set corresponding to all nodes.
And 505, obtaining the resource characteristics corresponding to the resource transfer network according to the accumulated number of the sub-network structures belonging to each topology type.
Schematically, as can be seen from fig. 6, when the cumulative number of the sub-network structures belonging to the topology type No. 10 is the largest, the topology type No. 10 is used as the resource feature corresponding to the sub-network structure 21. By using the method, the accumulated number of the topology types to which the sub-network structures 21 belong corresponding to all the nodes in the whole resource transfer network is calculated, so that the resource characteristics corresponding to the resource transfer network are obtained.
Illustratively, if the cumulative number of the sub-network structures belonging to the topology types No. 3 and No. 5 is 2, the resource characteristics corresponding to the sub-network structure 21 are determined by integrating the topology types No. 3 and No. 5. The resource characteristics corresponding to the sub-network structures 21 corresponding to all the nodes are obtained by using the method, so that the resource characteristics corresponding to all the sub-network structures are integrated to obtain the resource characteristics corresponding to the resource transfer network.
Step 506, calling a resource classification model to classify the resource features to obtain the prediction probability that the resource transfer transaction belongs to the abnormal resource transfer.
The resource classification model is a machine learning model with the capability of classifying the resource transfer types, and is a model trained in advance, and illustratively, the resource classification model can be trained by using the resource features extracted in the above embodiments.
And inputting the resource characteristics into the resource classification model, and outputting the prediction probability that the resource transfer belongs to the abnormal resource transfer. Illustratively, a resource classification model is called to classify the resource donation characteristics, the resource classification model judges that the resource is transferred to the user account A according to the resource donation characteristics, a certain number of resources are sent to the user account B through a virtual red packet, and the prediction probability that the donation resource behavior belongs to abnormal resource transfer is output.
And step 507, outputting the resource type of the resource transfer according to the prediction probability.
Illustratively, a probability threshold is set for the prediction probability, for example, the probability threshold is 0.6, and when the prediction probability output by the resource classification model is 0.75, the resource transfer belongs to an abnormal resource transfer; when the prediction probability output by the resource classification model is 0.4, the resource transfer belongs to the normal resource transfer. An abnormal resource transfer refers to a resource transfer that is at risk, such as a fraudulent transaction in a financial scenario.
In summary, in the method of this embodiment, the resource transfer network is constructed, the resource features are extracted according to the types of the preset sub-network structures to which the sub-network structures in the resource transfer network belong, and the resource transfer network is divided according to the preset sub-network structures representing resource transaction relations of different types, so that the extracted transaction resource features can accurately represent various resource transfer relations, thereby improving the accuracy of determining whether abnormal resource transfer exists in the resource transfer network by using the extracted resource features, and improving the efficiency of identifying abnormal resource transfer.
The resource characteristics are determined according to the accumulated number of the sub-network structures distributed under the different types of preset sub-network structures by calculating the accumulated number of the sub-network structures belonging to each topological type in the segmented sub-network structures, so that accurate and comprehensive resource characteristics are ensured to be extracted from the resource transfer network.
By acquiring a first set corresponding to the ith node and a second set corresponding to the first neighbor node, the accumulated quantity of each topology type in the sub-network structure is calculated according to the first set and the second set, so that comprehensive resource characteristics are extracted from the resource transfer network according to the accumulated quantity of the sub-network structures meeting the conditions.
The process of calculating the cumulative number of nodes belonging to each topology type in the sub-network structure is explained below.
In an alternative embodiment based on fig. 5, when the intersection set of the first set and the second set is a non-empty set, the process of calculating the cumulative number by the work server includes the following steps, as shown in fig. 7:
with viRepresenting the ith node, for node viE.g., V, for each node in V, respectively initializing feature vectors corresponding to 33-dimensional (corresponding to 33 preset sub-network structures) sub-network structures
Figure BDA0002782822860000191
(zero matrix).
Work byServer determining node viEach neighbor node of
Figure BDA0002782822860000192
With the first neighbor node as vjFor example, the working server pulls node v from the parameter serverjSecond set of
Figure BDA0002782822860000193
Obtaining a node viAnd node vjThere are common neighbor nodes (i.e., both with node v)iAnd node vjConnected), i.e. taking the intersection of the first set and the second set
Figure BDA0002782822860000194
Step 701, in response to that the intersection set of the first set and the second set is a non-empty set and the node identifier of the second neighbor node is greater than the maximum node identifier, obtaining a directed edge between the ith node, the first neighbor node and the second neighbor node, where the maximum node identifier is the maximum node identifier in the node identifiers of the ith node and the first neighbor node.
Step 702, according to the directed edges among the ith node, the first neighbor node and the second neighbor node, obtaining a sub-network structure formed by the ith node, the first neighbor node and the second neighbor node.
In step 703, the cumulative number of sub-network structures belonging to each topology type is calculated among the three sub-network structures.
Inputting the node identification of the ith node, the first neighbor node and the second neighbor node, and acquiring a first directed edge d between the ith node and the first neighbor node by the work serverijA second directed edge d between the first neighbor node and the second neighbor nodejkThird directed edge d between the ith node and a third neighboring nodeik。
Taking the network structure diagram 101 corresponding to the resource transfer network in fig. 4 as an example, taking the node 1 as the node viThen the first neighbor node vj(i.e., the first set of neighboring nodes)
Figure BDA0002782822860000195
Node 2) includes node 2, node 3 and node 4, when node 2 is taken as node vjThen, the second neighbor node vkComprising node 1, node 3 and node 4 (i.e. the second set of neighboring nodes)
Figure BDA0002782822860000196
) When node 3 is taken as node vjThen, the second neighbor node vkComprises a node 2 and a node 8, when the node 4 is taken as a node vjThen, the second neighbor node vkIncluding node 1, node 2, and node 6.
Representing node v by table oneiAnd the corresponding relation between the common neighbor nodes.
Watch 1
Figure BDA0002782822860000197
As can be seen from table one, the intersection set of the first set and the second set is a non-empty set, i.e., there is a common neighbor node between the ith node and the first neighbor node.
If commmon Neighbors! If phi, then the following logic is performed:
for node vk∈commonNeighborsijIf v isk＞max(vi，vj) Node vkFor the second neighboring node, the following logic is performed:
Figure BDA0002782822860000201
when node vkIf the node is the node 3 or the node 4, the logic is satisfied, and the working server is according to the node viNode vjAnd node vkThe directed edge therebetween outputs the accumulated number of topology types conforming to the predetermined sub-network structure in the sub-network structure formed by the three nodes. In the embodiment of the application, the serial number of the node is used for representing the node identifier, and the node with the larger serial number isThe larger the identification.
When node vkFor the node 3, the working server obtains directed edges among the nodes 1, 2 and 3, and if the node 1 is taken as a target node, and the sub-network structure formed by the nodes 1, 2 and 3 conforms to the topology type of the No. 25 preset sub-network structure in the graph 1, the cumulative number of the No. 25 topology type is added by 1; with the node 2 as a target node, if a sub-network structure formed by the node 1, the node 2 and the node 3 conforms to the topology type of the 26 th preset sub-network structure in fig. 1, adding 1 to the accumulated number of the 26 th topology type; with the node 3 as a target node, if the sub-network structure formed by the nodes 1, 2 and 3 conforms to the topology type of the 26 th predetermined sub-network structure in fig. 1, then 1 is added to the accumulated number of the 26 th topology type.
When node vkFor the node 4, the working server obtains directed edges among the node 1, the node 2 and the node 4, takes the node 1 as a target node, and adds 1 to the cumulative number of the No. 25 topology type if a sub-network structure formed by the node 1, the node 2 and the node 4 conforms to the topology type of the No. 25 preset sub-network structure in the graph 1; with the node 2 as a target node, if a sub-network structure formed by the node 1, the node 2 and the node 4 conforms to the topology type of the 26 th preset sub-network structure in fig. 1, adding 1 to the accumulated number of the 26 th topology type; with the node 4 as a target node, if the sub-network structure formed by the node 1, the node 2 and the node 4 conforms to the topology type of the 26 th preset sub-network structure in fig. 1, then 1 is added to the accumulated number of the 26 th topology type.
Therefore, when the node vkIn the case of node 3 or node 4, the cumulative number belonging to topology type No. 25 is 2, and the cumulative number belonging to topology type No. 26 is 4.
Calculating the current node v according to the modeiIs a target node, node vjWhen the nodes are other nodes, the cumulative number of the sub-network structures belonging to each topology type in the corresponding sub-network structures is calculated, and then the cumulative number of the sub-network structures belonging to each topology type in the sub-network structures corresponding to each node in the resource transfer network is calculated, so that the resource transfer network is obtainedAnd moving the resource characteristics corresponding to the network.
In summary, in the method of this embodiment, when the intersection set of the first set and the second set is a non-empty set, and the node identifier of the second neighbor node is greater than the maximum node identifier of the ith node and the first neighbor node, in this case, a certain node in the resource transfer network is taken as a target node, and the target node, the first neighbor node connected to the target node, and the second neighbor node connected to the first neighbor node calculate the cumulative number of the sub-network structures under the topology type according to the topology type to which the sub-network structure formed by the target node, the first neighbor node connected to the target node, and the second neighbor node connected to the first neighbor node belongs, so as to accurately calculate the cumulative number of the sub-network structures belonging to each topology type in the resource transfer network.
As can be seen from the above embodiments, when the intersection set of the first set and the second set is an empty set, i.e. the node viAnd node vjCommon neighbor nodes not existing in between, i.e. common neighborsijPhi is given. There are two cases: 1. node vkIs node vjBut not node viThe neighbor node of (2); 2. node vkIs node viBut not node vjOf the neighboring node.
1. Node vkIs node vjBut not node viOf the neighboring node. The process of the work server calculating the cumulative number includes the following steps, as shown in fig. 8:
step 801a, in response to that the intersection set of the first set and the second set is an empty set, and the second neighboring node is connected to the first neighboring node and not connected to the ith node, acquiring a second node identifier of the second neighboring node and a third node identifier of the ith node.
Step 802a, in response to the second node identifier being larger than the third node identifier, a first directed edge between the ith node and the first neighboring node and a second directed edge between the first neighboring node and the second neighboring node are obtained.
And 803a, acquiring a first sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the second directed edge.
Step 804a, in the first sub-network structure, taking the ith node as a target node, and calculating the accumulated number of the sub-network structures belonging to each topology type; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node.
Inputting the node identification of the ith node, the first neighbor node and the second neighbor node, wherein the second neighbor node vkWith the first neighbor node vjConnected to but connected to the ith node viDisconnected, the working server obtains the node viA first directed edge d with a first neighboring nodeijA second directed edge d between the first neighboring node and the second neighboring nodejk。
Taking the network structure diagram 101 corresponding to the resource transfer network in fig. 4 as an example, taking the node 7 as the node viFirst neighbor node vjIs node 9, a second neighbor node vkIs node 8.
For the
Figure BDA0002782822860000223
If v isk＞viThe following logic is executed:
indexi，indexj＝motifTwoEdgesLookupTable(dij，djk)
Figure BDA0002782822860000221
according to the logic, the working server acquires the node identifications of the node 7, the node 8 and the node 9, the first directed edge between the node 7 and the node 9, and the second directed edge between the node 8 and the node 9, and outputs the accumulated number which is in accordance with the topology type of the preset sub-network structure in the sub-network structure formed by the three nodes.
If the target node is taken as the node 7, the sub-network structure formed by the node 7, the node 8 and the node 9 conforms to the topology type of the No. 3 preset sub-network structure in the graph 1, and 1 is added to the accumulated number of the No. 3 topology type; if the target node is the node 9, the sub-network structure formed by the nodes 7, 8 and 9 conforms to the topology type of the No. 4 predetermined sub-network structure in fig. 1, and 1 is added to the cumulative number of the No. 4 topology type.
Step 801b, in response to the second node identifier being smaller than the third node identifier, obtains a first directed edge between the ith node and the first neighboring node, and a second directed edge between the first neighboring node and the second neighboring node.
And 802b, acquiring a first sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the second directed edge.
Step 803b, in the first sub-network structure, the cumulative number belonging to each topology type is calculated with the ith node as the target node.
Taking the network structure diagram 101 corresponding to the resource transfer network in fig. 4 as an example, taking the node 8 as the node viNode 9 is node vjNode 7 is node vk。
If v isk＜viThe following logic is executed:
indexi，indexj＝motifTwoEdgesLookupTable(dij，djk)
Figure BDA0002782822860000222
the working server obtains node identifiers of the nodes 7, 8 and 9 and a first directed edge between the nodes 8 and 9, and a second directed edge between the nodes 7 and 9, and takes the node 8 as a target node, so that a sub-network structure formed by the nodes 7, 8 and 9 conforms to the No. 5 preset sub-network structure in fig. 1, and 1 is added to the cumulative number of the No. 5 topology type.
When the node 7 and the node 9 are respectively target nodes, the topology types respectively calculated by the node 7 and the node 9 as the target nodes are repeated in the above steps 801a to 804a, so that the cumulative number of the corresponding topology types when the target nodes are the node 7 and the node 9 is not calculated any more in order to avoid the repetition count.
To sum up, when the intersection of the first set and the second set is an empty set, the second neighbor node is connected to the first neighbor node, and is not connected to the ith node, under the circumstance, the second node identifier of the second neighbor node, the third node identifier of the ith node, and the directed edge between the connected nodes are obtained to determine the sub-network structure formed between the three nodes, the ith node is taken as a target node, and the first neighbor node is taken as a target node to respectively calculate the cumulative number of the sub-network structures belonging to each topology type, so that the result of the cumulative number of each topology type is accurately extracted, and the comprehensive and accurate resource characteristics from the resource transfer network according to the accurate cumulative number are ensured.
When the second node identification is smaller than the third node identification, the cumulative number of the sub-network structures belonging to each topology type is calculated in the sub-network structure corresponding to the ith node as the target node to avoid repeated counting, so that the result of the cumulative number of each topology type is accurate, and the comprehensive and accurate resource characteristics are extracted from the resource transfer network according to the accurate cumulative number.
2. Node vkIs node viBut not node vjOf the neighboring node. The method of calculating the cumulative amount includes the following steps, as shown in fig. 9:
step 901a, in response to that the intersection set of the first set and the second set is an empty set, and the second neighboring node is connected to the ith node and not connected to the first neighboring node, acquiring a second node identifier of the second neighboring node and a first node identifier of the first neighboring node.
Step 902a, in response to the second node identifier being greater than the first node identifier, a first directed edge between the ith node and a first neighboring node and a third directed edge between the ith node and a second neighboring node are obtained.
Step 903a, according to the first directed edge and the third directed edge, obtaining a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node.
Step 904a, in the second sub-network structure, taking the ith node as a target node, and calculating the accumulated number of the sub-network structures belonging to each topology type; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node.
Inputting the node identification of the ith node, the first neighbor node and the second neighbor node, wherein the second neighbor node vkAnd the ith node viConnected but with the first neighbor node vjDisconnected, the working server obtains the node viA first directed edge d with a first neighboring nodeijThird directed edge d between the ith node and a second neighboring nodeki。
Taking the network structure diagram 101 corresponding to the resource transfer network in fig. 4 as an example, the node 7-bit node v is taken as an exampleiFirst neighbor node vjIs node 9, a second neighbor node vkIs node 10.
For the
Figure BDA0002782822860000242
If v isk＞vjThen the following logic is executed:
indexi，indexj＝motifTwoEdgesLookupTable(dij，dki)
Figure BDA0002782822860000241
according to the logic, the working server obtains the node identifications of the node 7, the node 9 and the node 10, the first directed edge between the node 7 and the node 9, and the third directed edge between the node 7 and the node 10, and outputs the accumulated number which accords with the topological type of the preset sub-network structure in the sub-network structure formed by the three nodes.
If the target node is taken as the node 7, the sub-network structure formed by the node 7, the node 9 and the node 10 conforms to the topology type of the No. 4 preset sub-network structure in the graph 1, and 1 is added to the cumulative number of the No. 4 topology type; if the target node is the node 9, the sub-network structure formed by the nodes 7, 9 and 10 conforms to the topology type of the preset sub-network structure No. 5 in fig. 1, and 1 is added to the cumulative number of the topology type No. 5.
When the second neighbor node vkIs smaller than the first neighbor node vjThe method of calculating the cumulative number comprises the following steps. As shown in fig. 9:
step 901b, in response to the second node identifier being smaller than the first node identifier, obtaining a first directed edge between the ith node and a first neighboring node, and a third directed edge between the ith node and a second neighboring node.
And 902b, acquiring a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the third directed edge.
Step 903b, in the second sub-network structure, the first neighbor node is taken as a target node, and the accumulated number of the sub-network structures belonging to each topology type is calculated.
Taking the network structure diagram 101 corresponding to the resource transfer network in fig. 4 as an example, taking the node 7 as the node viNode 10 is node vjNode 9 is node vk。
If v isk＜vjThen the following logic is executed:
indexi，indexj＝motifTwoEdgesLookupTable(dij，dki)
Figure BDA0002782822860000251
the working server acquires node identifiers of the nodes 7, 9 and 10, a first directed edge between the nodes 7 and 10, and a third directed edge between the nodes 7 and 9, and takes the node 10 as a target node, so that a sub-network structure formed by the nodes 7, 9 and 10 conforms to the No. 3 preset sub-network structure in fig. 1, and 1 is added to the cumulative number of the No. 3 topology type.
When the node 7 and the node 9 are respectively target nodes, the topology types respectively calculated by using the node 7 and the node 9 as the target nodes are repeated in the above steps 901a to 904a, so that the cumulative number of the corresponding topology types when the target nodes are the node 7 and the node 9 is not calculated again in order to avoid the repetition count.
To sum up, when the intersection of the first set and the second set is an empty set, the second neighbor node is connected to the ith node and is not connected to the first neighbor node, under the circumstance, the second node identifier of the second neighbor node, the first node identifier of the first neighbor node and the directed edge between the connected nodes are obtained to determine the sub-network structure formed among the three nodes, and the first neighbor node is used as the target node to calculate the cumulative number of the sub-network structures belonging to each topology type, so that the result of the cumulative number of each topology type is accurate, and the comprehensive and accurate resource features are extracted from the resource transfer network according to the accurate cumulative number.
When the second node identification is smaller than the first node identification, the accumulated quantity of the sub-network structures belonging to each topology type is calculated to avoid repeated counting in the sub-network structure corresponding to the target node by taking the first neighbor node as the target node, so that the result of the accumulated quantity of each topology type is accurate, and the extraction of comprehensive and accurate resource characteristics from the resource transfer network according to the accurate accumulated quantity is ensured.
In some embodiments, the directed edges between the nodes correspond to weight information, and the work server obtains resource characteristics corresponding to the resource transfer network according to the type to which the segmented sub-network structure belongs and the weight information corresponding to the directed edges.
Fig. 10 shows a flowchart of a complex network resource feature extraction based method according to another exemplary embodiment of the present application. The embodiment is described by taking as an example that the method is used in the computer system 100 shown in fig. 2, and the method includes the following steps:
step 1001, acquiring a resource transfer relationship.
Step 1002, a resource transfer network is constructed according to the resource transfer relationship, the resource transfer network includes at least two nodes, the nodes are used for representing user accounts for resource transfer, directed edges are corresponding between the nodes, and the directed edges are used for representing resource flow direction in the resource transfer.
Step 1003, segmenting the resource transfer network according to the preset sub-network structures to obtain at least two segmented sub-network structures, wherein different preset sub-network structures are used for representing resource transfer relations of different topology types.
And 1004, calculating the accumulated number of the sub-network structures belonging to each topological type in at least two segmented sub-network structures.
The implementation of steps 1001 to 1004 is the same as the implementation of steps 501 to 504 in the embodiment shown in fig. 5, and is not described here again.
Step 1005, calculating the weight value corresponding to the sub-network structures belonging to the same type according to the weight information corresponding to the directed edge.
When calculating the weight information corresponding to the directed edge, step 1005 may be replaced by the following steps:
s1, obtaining a first adjacent triple set corresponding to the ith node, wherein the first adjacent triple set comprises a first weight set, the first weight set represents a weight value corresponding to a directed edge between the ith node and a first neighbor node, i is not more than n, and i is a positive integer.
The work server pulls the ith node v from the parameter serveriAdjacent triple set of
Figure BDA0002782822860000261
As can be seen from the above-described embodiments,
Figure BDA0002782822860000262
representing a first set of neighboring nodes that are,
Figure BDA0002782822860000263
a first set of directed edges is represented,
Figure BDA0002782822860000264
representing the ith node and the first neighbor node vjThe weight values corresponding to the directed edges in between,
Figure BDA0002782822860000265
s2, acquiring a second adjacent triple set corresponding to the first neighbor node, wherein the second adjacent triple set comprises a second weight set, the second weight set represents a weight value corresponding to a directed edge between the first neighbor node and the second neighbor node, and the second neighbor node is connected with the first neighbor node.
For node viFirst neighbor node of (i-th node)
Figure BDA0002782822860000266
The work server pulls the first neighbor node v from the parameter serverjAdjacent triple set of
Figure BDA0002782822860000267
Figure BDA0002782822860000268
Representing a second set of neighboring nodes, the second set of neighboring nodes being a set corresponding to nodes connected to the first neighboring node,
Figure BDA0002782822860000269
representing a second set of directed edges, the second set of directed edges being a first neighbor node vjWith a second neighboring node vkThe corresponding set of directed edges in between,
Figure BDA00027828228600002610
a set of weight values representing the correspondence of directed edges between a first neighbor node and a second neighbor node,
Figure BDA00027828228600002611
and S3, repeating the step of obtaining the adjacent triple set until the n nodes are traversed, and calculating the corresponding weight values of the sub-network structures belonging to the same type according to the weight values corresponding to the directed edges among the n nodes.
And repeating the step S1 and the step S2, obtaining an adjacent triple set corresponding to each node in the resource transfer network, and calculating a weight value corresponding to a directed edge of the same type of sub-network structure in the sub-network structure corresponding to each node according to the weight value in the adjacent triple set.
And step 1006, obtaining the resource characteristics corresponding to the resource transfer network according to the accumulated quantity and the weight value.
And the work server outputs the resource characteristics corresponding to the resource transfer network by combining the accumulated number of the sub-network structures belonging to each topology type and the weight value.
In summary, in the method of this embodiment, the weight information corresponding to the directed edge in the sub-network structure is obtained, and the resource characteristics corresponding to the resource transfer network are obtained by combining the topology number, so that the obtained resource characteristics are more comprehensive and accurate and representative, and it is ensured that the abnormal resource transfer in the resource transfer network can be accurately identified according to the resource characteristics in the following process.
And calculating the weight values corresponding to the sub-network structures belonging to the same type according to the weight information corresponding to the directed edges, so that comprehensive resource characteristics are obtained by combining the topological quantity. The obtained resource characteristics are more comprehensive and accurate and representative, and the abnormal resource transfer transaction in the resource transfer network can be accurately identified according to the resource characteristics in the follow-up process.
The method comprises the steps of obtaining a first adjacent triple set of the ith node and a second adjacent triple set corresponding to the first neighbor node, calculating the weight values corresponding to the directed edges between the nodes according to the weight sets, repeating the calculating step, and obtaining the weight values corresponding to the directed edges in the sub-network structures belonging to the same type in the resource transfer network, so that the calculated weight values in the resource transfer network are more accurate, and the fact that comprehensive and accurate resource characteristics are obtained according to the accurate weight values is guaranteed.
In an alternative embodiment based on fig. 10, when the intersection set of the first set and the second set is a non-empty set, the process of extracting resource features by the work server includes the following steps, as shown in fig. 11:
with viRepresenting the ith node, for node viE.g., V, for each node in V, respectively initializing feature vectors corresponding to 33-dimensional (corresponding to 33 preset sub-network structures) sub-network structures
Figure BDA0002782822860000271
(zero matrix), weighted eigenvectors corresponding to directed edges
Figure BDA0002782822860000272
(zero matrix).
Step 1101, in response to that the intersection set of the first set and the second set is a non-empty set and the node identifier of the second neighbor node is greater than the maximum node identifier, obtaining a directed edge between the ith node, the first neighbor node and the second neighbor node, wherein the maximum node identifier is the maximum node identifier in the node identifier of the ith node and the node identifier of the first neighbor node.
Step 1102, a sub-network structure formed by the ith node, the first neighbor node and the second neighbor node is obtained according to the directed edges among the ith node, the first neighbor node and the second neighbor node.
Step 1103, calculating the cumulative number of the sub-network structures belonging to each topology type and the weight value corresponding to each directed edge in the sub-network structures formed by the three.
In step 1101 and step 1103, the embodiment of calculating the cumulative number of sub-network structures belonging to each topology type in the sub-network structures is the same as the embodiment of step 701 to step 703 shown in fig. 7, and will not be described again here.
For the situation, a sub-network structure formed by the ith node, the first neighbor node and the second neighbor node has a directed edge between any two nodes, and weight values corresponding to the directed edges of the sub-network structures belonging to the same type are calculated. Calculating a weight value corresponding to a directed edge in a sub-network structure by taking the ith node as a sub-network structure corresponding to a target node; calculating a weight value corresponding to a directed edge in a sub-network structure by taking a first neighbor node as a sub-network structure corresponding to a target node; and calculating the weight value corresponding to the directed edge in the sub-network structure by taking the second neighbor node as the sub-network structure corresponding to the target node.
If commmon Neighbors! For node v ═ phik∈commonNeighborsijAnd v isk＞max(vi，vj) Node vkFor the second neighboring node, the following logic is performed:
Figure BDA0002782822860000281
in an alternative embodiment based on fig. 10, when the intersection set of the first set and the second set is an empty set, the process of extracting resource features by the work server includes the following steps, as shown in fig. 12:
1. node vkIs node vjBut not node viOf the neighboring node.
Step 1201a, in response to that the intersection set of the first set and the second set is an empty set, and the second neighbor node is connected to the first neighbor node and not connected to the ith node, acquiring a second node identifier of the second neighbor node and a third node identifier of the ith node.
Step 1202a, in response to the second node identifier being greater than the third node identifier, a first directed edge between the ith node and the first neighboring node and a second directed edge between the first neighboring node and the second neighboring node are obtained.
Step 1203a, obtaining a first sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the second directed edge.
Step 1204a, in the first sub-network structure, taking the ith node as a target node, calculating the cumulative number of the sub-network structures belonging to each topology type, and calculating the weight values corresponding to the sub-network structures belonging to the same type; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node, and calculating the weight values corresponding to the sub-network structures belonging to the same type.
For the
Figure BDA0002782822860000292
If v isk＞viThe following logic is executed:
indexi，indexj＝motifTwoEdgesLookupTable(dij，djk)
Figure BDA0002782822860000291
in step 1201a and step 1204a, the embodiment for calculating the cumulative number of sub-network structures belonging to each topology type in the sub-network structures is the same as the embodiment of step 801a to step 804a shown in fig. 8, and will not be described again here.
For the situation, the sub-network structure formed by the ith node, the first neighbor node and the second neighbor node is arranged at the node viAnd node vjThere is a directed edge between, at node vjAnd node vkThere is a directed edge between them, and the weighted value corresponding to the directed edge belonging to the same type of sub-network structure is calculated. Calculating a weight value corresponding to a directed edge in a sub-network structure by taking the ith node as a sub-network structure corresponding to a target node; and calculating the weight value corresponding to the directed edge in the sub-network structure by taking the first neighbor node as the sub-network structure corresponding to the target node.
When the second neighbor node vkIs less than node viThe method for extracting resource features includes the following steps, as shown in fig. 12:
step 1201b, in response to the second node identifier being smaller than the third node identifier, acquiring a first directed edge between the ith node and the first neighboring node, and a second directed edge between the first neighboring node and the second neighboring node.
And 1202b, acquiring a first sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the second directed edge.
Step 1203b, in the first sub-network structure, taking the ith node as a target node, calculating the accumulated number of the sub-network structures belonging to each topology type, and calculating the weight values corresponding to the sub-network structures belonging to the same type.
If v isk＜viThe following logic is executed:
indexi，indexj＝motifTwoEdgesLookupTable(dij，djk)
Figure BDA0002782822860000301
in step 1201b and step 1203a, the embodiment for calculating the cumulative number of sub-network structures belonging to each topology type in the sub-network structures is the same as the embodiment of step 801b to step 803b shown in fig. 8, and will not be described again here.
For the situation, the sub-network structure formed by the ith node, the first neighbor node and the second neighbor node is arranged at the node viAnd node vjThere is a directed edge between, at node vjAnd node vkThere is a directed edge between them, and the weighted value corresponding to the directed edge belonging to the same type of sub-network structure is calculated. And calculating the weight value corresponding to the directed edge in the sub-network structure by taking the ith node as the sub-network structure corresponding to the target node.
2. Node vkIs node viBut not node vjOf the neighboring node. The process of extracting resource features by the work server includes the following steps, as shown in fig. 13:
step 1301a, in response to that the intersection set of the first set and the second set is an empty set, and the second neighbor node is connected to the ith node and not connected to the first neighbor node, acquiring a second node identifier of the second neighbor node and a first node identifier of the first neighbor node.
Step 1302a, in response to the second node identifier being greater than the first node identifier, obtains a first directed edge between the ith node and a first neighboring node, and a third directed edge between the ith node and a second neighboring node.
And step 1303a, obtaining a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the third directed edge.
Step 1304a, in the second sub-network structure, with the ith node as a target node, calculating the cumulative number of sub-network structures belonging to each topology type, and calculating the weight values of the sub-network structures belonging to the same type; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node, and calculating the weight value of the sub-network structures belonging to the same type.
For the
Figure BDA0002782822860000302
If v isk＞vjThen the following logic is executed:
indexi，indexj＝motifTwoEdgesLookupTable(dij，dki)
Figure BDA0002782822860000311
in step 1301a and step 1304a, the embodiment for calculating the cumulative number of sub-network structures belonging to each topology type in the sub-network structures is the same as the embodiment of step 901a to step 904a shown in fig. 9, and will not be described again here.
For the situation, the sub-network structure formed by the ith node, the first neighbor node and the second neighbor node is arranged at the node viAnd node vjThere is a directed edge between, at node viAnd node vkThere is a directed edge between them, and the weighted value corresponding to the directed edge belonging to the same type of sub-network structure is calculated. Calculating the sub-network structure corresponding to the i-th node as the target nodeWeighted values corresponding to the directed edges; and calculating the weight value corresponding to the directed edge in the sub-network structure by taking the first neighbor node as the sub-network structure corresponding to the target node.
When the second neighbor node vkIs smaller than the first neighbor node vjWhen the first node is identified, the method for extracting the resource features includes the following steps, as shown in fig. 13:
step 1301b, in response to the fact that the second node identification is smaller than the first node identification, a first directed edge between the ith node and the first neighbor node and a third directed edge between the ith node and the second neighbor node are obtained.
Step 1302b, according to the first directed edge and the third directed edge, a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node is obtained.
And step 1303b, in the second sub-network structure, the first neighbor node is taken as a target node, the accumulated number of the sub-network structures belonging to each topology type is calculated, and the weight values corresponding to the sub-network structures of the same type are calculated.
If v isk＜vjThen the following logic is executed:
indexi，indexj＝motifTwoEdgesLookupTable(dij，dki)
Figure BDA0002782822860000312
in step 1301b and step 1303b, the embodiment of calculating the cumulative number of sub-network structures belonging to each topology type in the sub-network structures is the same as the embodiment of step 901b to step 903b shown in fig. 9, and is not described again here.
For the situation, the sub-network structure formed by the ith node, the first neighbor node and the second neighbor node is arranged at the node viAnd node vjThere is a directed edge between, at node viAnd node vkThere is a directed edge between them, and the directed edge pair belonging to the same type of sub-network structure is calculatedThe weight value should be weighted. And calculating the weight value corresponding to the directed edge in the sub-network structure by taking the first neighbor node as the sub-network structure corresponding to the target node.
In summary, in the method of this embodiment, by calculating the weight value corresponding to the sub-network structure formed by the target node (i-th node), the first neighboring node corresponding to the target node, and the second neighboring node, and combining the previously calculated cumulative number of sub-network structures belonging to each topology type, the resource characteristics corresponding to the resource transfer network are output comprehensively, so that the abnormal resource transfer can be accurately identified subsequently according to the resource characteristics, and the risky resource transfer can be controlled at the time.
A method of calculating a weight value corresponding to a subnet structure is described as a specific example.
As shown in fig. 14, a sub-network structure 31 is formed among a node a, a node B, a node C, a node D, and a node E, weight information (weight value) is associated with directed edges between the nodes, the sub-network structure 31 is divided according to a preset sub-network structure, and the divided sub-network structures are respectively calculated with corresponding weight values, so that the weight values corresponding to the sub-network structure 31 are obtained.
With node a as the target node, the sub-network structure 31 is divided into a first sub-network structure 32, a second sub-network structure 33, a third sub-network structure 34 and a fourth sub-network structure 35 according to the predetermined sub-network structure shown in fig. 1.
First sub-network structure 32: consisting of node a, node B and node C, the corresponding weight values in the first sub-network structure 32 are:
Figure BDA0002782822860000321
as can be seen from the above embodiments, from the angle of the directional edge indicating the direction, the value corresponding to the directional edge AC is 0, the value corresponding to the directional edge BC is 0, and the value corresponding to the directional edge BA is 1.
The second sub-network structure 33: consisting of node a, node D and node C, the corresponding weight values in the second sub-network structure 33 are:
Figure BDA0002782822860000322
as can be seen from the above embodiments, from the angle of the directional edge indicating the direction, the value corresponding to the directional edge CD is 0, the value corresponding to the directional edge AC is 0, and the value corresponding to the directional edge DA is 1.
Third sub-network structure 34: the sub-network structure consisting of the node A, the node D and the node E and the sub-network structure consisting of the node A, the node E and the node B which satisfy the sub-network structure belong to the same topological type, so the weighted values calculated in the two cases are added. Since the third sub-network structure 34 includes two directed edges, the weight values corresponding to the third sub-network structure 34 are:
Figure BDA0002782822860000331
as can be seen from the above embodiment, from the angle in which the directional edge indicates the direction, the numerical value corresponding to the directional edge AE is 2, the numerical value corresponding to the directional edge DA is 1, and the numerical value corresponding to the directional edge BA is 1.
Fourth sub-network structure 35: the fourth sub-network structure 35 includes two directed edges, and the corresponding weight values in the fourth sub-network structure 35 are: 12.25
Figure BDA0002782822860000332
As can be seen from the above embodiments, the value corresponding to the directed edge AE is 2 and the value corresponding to the directed edge AC is 0, from the perspective of the directed edge indicating the direction.
It should be noted that, in the present embodiment, the sub-network structure 31 can also be split into sub-network structures between two nodes, and the present embodiment is described by taking a sub-network structure of three nodes as an example.
The resource feature extraction method based on the complex network is suitable for large-scale resource transfer networks, such as transaction fraud, financial credit investigation, merchant portrait, user (consumer) portrait, risk estimation and other scenes.
Fig. 15 is a block diagram illustrating a structure of an apparatus for extracting resource features based on a complex network according to an exemplary embodiment of the present application, where the apparatus includes the following components:
an obtaining module 1510, configured to obtain a resource transfer relationship;
a building module 1520, configured to build a resource transfer network according to the resource transfer relationship, where the resource transfer network includes at least two nodes, the nodes are used to represent user accounts for resource transfer, directed edges are corresponding between the nodes, and the directed edges are used to represent resource flow directions in resource transfer;
the segmentation module 1530 is configured to segment the resource transfer network according to a preset sub-network structure to obtain at least two segmented sub-network structures, where different preset sub-network structures are used to represent resource transfer relationships of different topology types;
the feature extraction module 1540 is configured to obtain the resource feature corresponding to the resource transfer network according to the topology type to which the segmented sub-network structure belongs.
In an alternative embodiment, the feature extraction module 1540 is configured to calculate, in at least two segmented sub-network structures, the cumulative number of sub-network structures belonging to each topology type; and obtaining the resource characteristics corresponding to the resource transfer network according to the accumulated number of the sub-network structures belonging to each topology type.
In an optional embodiment, the resource transfer network includes n nodes, where n is a positive integer;
the obtaining module 1510 is configured to obtain a first set corresponding to an ith node, where the first set includes a first neighbor node set and a first directed edge set, the first neighbor node set represents a set corresponding to a first neighbor node connected to the ith node, and a node identifier of the first neighbor node is greater than a node identifier of the ith node, the first directed edge set represents a set corresponding to a directed edge between the ith node and the first neighbor node, i is not greater than n, and i is a positive integer; acquiring a second set corresponding to the first neighbor node, wherein the second set comprises a second neighbor node set and a second directed edge set, the second neighbor node set represents a set corresponding to a second neighbor node connected with the first neighbor node, and the second directed edge set represents a set corresponding to a directed edge between the first neighbor node and the second neighbor node; the feature extraction module 1540 is configured to repeat the above steps of obtaining the first set and the second set until n nodes are traversed, and calculate the cumulative number of the sub-network structures belonging to each topology type according to the first set and the second set corresponding to the n nodes, respectively.
In an optional embodiment, the obtaining module 1510 is configured to, in response to that the intersection set of the first set and the second set is a non-empty set and the node identifier of the second neighboring node is greater than the maximum node identifier, obtain a directed edge between the ith node, the first neighboring node, and the second neighboring node, where the maximum node identifier is a largest node identifier of the node identifiers of the ith node and the first neighboring node; acquiring a sub-network structure formed by the ith node, the first neighbor node and the second neighbor node according to directed edges among the ith node, the first neighbor node and the second neighbor node; the feature extraction module 1540 is configured to calculate the cumulative number of sub-network structures belonging to each topology type in the three formed sub-network structures.
In an optional embodiment, the obtaining module 1510 is configured to, in response to that the intersection set of the first set and the second set is an empty set, and the second neighboring node is connected to the first neighboring node and not connected to the ith node, obtain a second node identifier of the second neighboring node and a third node identifier of the ith node; responding to the second node identification larger than the third node identification, and acquiring a first directed edge between the ith node and a first neighbor node and a second directed edge between the first neighbor node and a second neighbor node; acquiring a first sub-network structure corresponding to the ith node, a first neighbor node and a second neighbor node according to the first directed edge and the second directed edge; the feature extraction module 1540 is configured to calculate, in the first sub-network structure, the cumulative number of sub-network structures belonging to each topology type by using the ith node as a target node; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node.
In an optional embodiment, the obtaining module 1510 is configured to, in response to the second node identifier being smaller than the third node identifier, obtain a first directed edge between the ith node and a first neighboring node, and a second directed edge between the first neighboring node and a second neighboring node; acquiring a first sub-network structure corresponding to the ith node, a first neighbor node and a second neighbor node according to the first directed edge and the second directed edge; the feature extraction module 1540 is configured to calculate, in the first sub-network structure, the cumulative number of sub-network structures belonging to each topology type by using the ith node as a target node.
In an optional embodiment, the obtaining module 1510 is configured to, in response to that the intersection set of the first set and the second set is an empty set, and the second neighboring node is connected to the ith node and not connected to the first neighboring node, obtain a second node identifier of the second neighboring node and a first node identifier of the first neighboring node; responding to the second node identification larger than the first node identification, and acquiring a first directed edge between the ith node and a first neighbor node and a third directed edge between the ith node and a second neighbor node; acquiring a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the third directed edge; the feature extraction module 1540 is configured to calculate, in the second sub-network structure, the cumulative number of sub-network structures belonging to each topology type by using the ith node as a target node; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node.
In an optional embodiment, the obtaining module 1510 is configured to, in response to the second node identifier being smaller than the first node identifier, obtain a first directed edge between the ith node and a first neighboring node, and a third directed edge between the ith node and a second neighboring node; acquiring a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the third directed edge; the feature extraction module 1540 is configured to calculate, in the second sub-network structure, the cumulative number of sub-network structures belonging to each topology type by using the first neighboring node as the target node.
In an optional embodiment, the directed edges between the nodes correspond to weight information;
the feature extraction module 1540 is configured to obtain the resource feature corresponding to the resource transfer network according to the type of the segmented sub-network structure and the weight information corresponding to the directed edge.
In an alternative embodiment, the feature extraction module 1540 is configured to calculate, in at least two segmented sub-network structures, the cumulative number of sub-network structures belonging to each topology type; calculating the weight value corresponding to the sub-network structure belonging to the same type according to the weight information corresponding to the directed edge; and obtaining the resource characteristics corresponding to the resource transfer network according to the accumulated quantity and the weight value.
In an optional embodiment, the resource transfer network includes n nodes, where n is a positive integer;
the obtaining module 1510 is configured to obtain a first adjacent triple set corresponding to the ith node, where the first adjacent triple set includes a weight set, the first weight set represents a weight value corresponding to a directed edge between the ith node and a first neighboring node, i is not greater than n, and i is a positive integer; acquiring a second adjacent triple set corresponding to the first neighbor node, wherein the second adjacent triple set comprises a second weight set, the second weight set represents a weight value corresponding to a directed edge between the first neighbor node and a second neighbor node, and the second neighbor node is connected with the first neighbor node; the feature extraction module 1540 is configured to repeat the steps of obtaining the first adjacent triple set and the second adjacent triple set until the n nodes are traversed, and calculate the weight values corresponding to the sub-network structures belonging to the same type according to the weight values corresponding to the directed edges between the n nodes.
In an alternative embodiment, the apparatus includes a calling module 1550;
the calling module 1550 is configured to call the resource classification model to classify the resource features, so as to obtain a prediction probability that the resource transfer belongs to the abnormal resource transfer; and outputting the type of the resource transfer according to the prediction probability.
In summary, the device provided in this embodiment extracts the resource features by constructing the resource transfer network, and divides the resource transfer network by the preset sub-network structures representing different types of resource transfer relationships, so that the extracted resource features can accurately represent various resource transfer relationships, thereby improving the accuracy of determining whether there is an abnormal resource transfer in the resource transfer network by using the extracted resource features, and improving the efficiency of identifying the abnormal resource transfer.
The resource transfer characteristics are determined according to the accumulated number of the sub-network structures distributed under the different types of preset sub-network structures by calculating the accumulated number of the sub-network structures belonging to each topological type in the segmented sub-network structures, so that accurate and comprehensive resource characteristics are ensured to be extracted from the resource transfer network.
By acquiring a first set corresponding to the ith node and a second set corresponding to the first neighbor node, the accumulated quantity of each topology type in the sub-network structure is calculated according to the first set and the second set, so that comprehensive resource characteristics are extracted from the resource transfer network according to the accumulated quantity of the sub-network structures meeting the conditions.
When the intersection set of the first set and the second set is a non-empty set, and the node identification of the second neighbor node is greater than the maximum node identification of the ith node and the first neighbor node, under the condition, taking a certain node in the resource transfer network as a target node, and taking the target node, the first neighbor node connected with the target node and the second neighbor node connected with the first neighbor node, according to the topology type of the sub-network structure formed by the target node, the first neighbor node and the second neighbor node, the cumulative number of the sub-network structures under the topology type is calculated, so that the cumulative number of the sub-network structures belonging to each topology type in the resource transfer network is accurately calculated.
When the intersection of the first set and the second set is an empty set, the second neighbor node is connected with the first neighbor node and is not connected with the ith node, under the condition, the second node identification of the second neighbor node and the third node identification of the ith node are obtained, when the third node identification is smaller than the second node identification, the sub-network structure formed by the ith node, the first neighbor node and the second neighbor node is determined by obtaining the first directed edge and the second directed edge, and then the cumulative number of the sub-network structures belonging to each type in the sub-network structure is calculated, so that the comprehensive resource characteristics in the resource transfer network are extracted according to the accurate cumulative number.
And when the third node identification is larger than the second node identification, in order to avoid repeated counting, determining a sub-network structure formed by the first neighbor node and the second neighbor node by acquiring a second directed edge, and further calculating the cumulative number of the sub-network structures belonging to each topology type in the sub-network structure, so that comprehensive resource characteristics are extracted from the resource transfer network according to the accurate cumulative number.
When the intersection of the first set and the second set is an empty set, the second neighbor node is connected with the ith node and is not connected with the first neighbor node, under the condition, the second node identification of the second neighbor node and the first node identification of the first neighbor node are obtained, when the second node identification is larger than the first node identification, the sub-network structure formed by the ith node, the first neighbor node and the second neighbor node is determined by obtaining the first directed edge and the third directed edge, the cumulative number of the sub-network structures belonging to each type in the sub-network structure is further calculated, and the comprehensive resource characteristics in the resource transfer network are extracted according to the accurate cumulative number.
And when the second node identification is smaller than the first node identification, in order to avoid repeated counting, determining a sub-network structure formed by the ith node and the second neighbor node by acquiring a third directed edge, and further calculating the cumulative number of the sub-network structures belonging to each topological type in the sub-network structure, so that the comprehensive resource characteristics are extracted from the resource transfer network according to the accurate cumulative number.
By acquiring the weight information corresponding to the directed edges in the sub-network structure and combining the topological quantity to obtain the resource characteristics corresponding to the resource transfer network, the obtained resource characteristics are more comprehensive and accurate and representative, and the abnormal resource transfer in the resource transfer network can be accurately identified according to the resource characteristics in the follow-up process.
And calculating the weight values corresponding to the sub-network structures belonging to the same type according to the weight information corresponding to the directed edges, so that comprehensive resource characteristics are obtained by combining the topological quantity. The obtained resource characteristics are more comprehensive and accurate and representative, and the abnormal resource transfer in the resource transfer network can be accurately identified according to the resource characteristics in the follow-up process.
By obtaining the adjacent triple set of the ith node, wherein the set comprises the weight set, the weight value corresponding to the directed edge between the ith node and the neighbor node is calculated according to the weight set, and the step is repeated to obtain the weight value corresponding to the directed edge in the sub-network structure belonging to the same type in the resource transfer network, so that the calculated weight value in the resource transfer network is more accurate, and the comprehensive resource characteristics can be obtained according to the accurate weight value subsequently.
It should be noted that: the resource feature extraction device based on the complex network provided in the foregoing embodiment is only illustrated by the division of the functional modules, and in practical applications, the function distribution may be completed by different functional modules according to needs, that is, the internal structure of the device is divided into different functional modules, so as to complete all or part of the functions described above. In addition, the complex network-based resource feature extraction device provided by the above embodiment and the complex network-based resource feature extraction method embodiment belong to the same concept, and specific implementation processes thereof are described in the method embodiment and are not described herein again.
Fig. 16 shows a schematic structural diagram of a server provided in an exemplary embodiment of the present application. The server may be a server of the first computer device 110 and the second computer device 120 in the computer system 100 shown in fig. 1.
The server 1600 includes a Central Processing Unit (CPU) 1601, a system Memory 1604 including a Random Access Memory (RAM) 1602 and a Read Only Memory (ROM) 1603, and a system bus 1605 connecting the system Memory 1604 and the Central Processing Unit 1601. The server 1600 also includes a basic Input/Output System (I/O System)1606, which facilitates information transfer between various devices within the computer, and a mass storage device 1607 for storing an operating System 1613, application programs 1614, and other program modules 1615.
The basic input/output system 1606 includes a display 1608 for displaying information and an input device 1609 such as a mouse, keyboard, etc. for user input of information. Wherein a display 1608 and an input device 1609 are connected to the central processing unit 1601 by way of an input-output controller 1610 which is connected to the system bus 1605. The basic input/output system 1606 may also include an input-output controller 1610 for receiving and processing input from a number of other devices, such as a keyboard, mouse, or electronic stylus. Similarly, input-output controller 1610 may also provide output to a display screen, a printer, or other type of output device.
The mass storage device 1607 is connected to the central processing unit 1601 by a mass storage controller (not shown) connected to the system bus 1605. The mass storage device 1607 and its associated computer-readable media provide non-volatile storage for the server 1600. That is, the mass storage device 1607 may include a computer-readable medium (not shown) such as a hard disk or Compact disk Read Only Memory (CD-ROM) drive.
Computer-readable media may include computer storage media and communication media. Computer storage media includes volatile and nonvolatile, removable and non-removable media implemented in any method or technology for storage of information such as computer readable instructions, data structures, program modules or other data. Computer storage media includes RAM, ROM, Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable Programmable Read-Only Memory (EEPROM), flash Memory or other Solid State Memory technology, CD-ROM, Digital Versatile Disks (DVD), or Solid State Drives (SSD), other optical storage, magnetic cassettes, magnetic tape, magnetic disk storage, or other magnetic storage devices. The Random Access Memory may include a resistive Random Access Memory (ReRAM) and a Dynamic Random Access Memory (DRAM). Of course, those skilled in the art will appreciate that computer storage media is not limited to the foregoing. The system memory 1604 and mass storage device 1607 described above may be collectively referred to as memory.
According to various embodiments of the application, the server 1600 may also operate with remote computers connected to a network, such as the Internet. That is, the server 1600 may be connected to the network 1612 through the network interface unit 1611 coupled to the system bus 1605 or may be connected to other types of networks or remote computer systems (not shown) using the network interface unit 1611.
The memory further includes one or more programs, and the one or more programs are stored in the memory and configured to be executed by the CPU.
In an alternative embodiment, a computer device is provided, which includes a processor and a memory, wherein at least one instruction, at least one program, set of codes, or set of instructions is stored in the memory, and the at least one instruction, at least one program, set of codes, or set of instructions is loaded and executed by the processor to implement the complex network based resource feature extraction method as described above.
In an alternative embodiment, a computer-readable storage medium is provided that has at least one instruction, at least one program, set of codes, or set of instructions stored therein, which is loaded and executed by a processor to implement the complex network based resource feature extraction method as described above.
Optionally, the computer-readable storage medium may include: a Read Only Memory (ROM), a Random Access Memory (RAM), a Solid State Drive (SSD), or an optical disc. The Random Access Memory may include a resistive Random Access Memory (ReRAM) and a Dynamic Random Access Memory (DRAM). The above-mentioned serial numbers of the embodiments of the present application are for description only and do not represent the merits of the embodiments.
Embodiments of the present application also provide a computer program product or a computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and executes the computer instructions to cause the computer device to perform the complex network-based resource feature extraction method as described above.
It will be understood by those skilled in the art that all or part of the steps for implementing the above embodiments may be implemented by hardware, or may be implemented by a program instructing relevant hardware, where the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, etc.
The above description is intended to be exemplary only, and not to limit the present application, and any modifications, equivalents, improvements, etc. made within the spirit and scope of the present application are intended to be included therein.

Claims (15)

1. A resource feature extraction method based on a complex network is characterized by comprising the following steps:
acquiring a resource transfer relationship;
constructing a resource transfer network according to the resource transfer relationship, wherein the resource transfer network comprises at least two nodes, the nodes are used for representing user accounts for resource transfer, directed edges are correspondingly arranged between the nodes, and the directed edges are used for representing the resource flow direction in the resource transfer;
segmenting the resource transfer network according to a preset sub-network structure to obtain at least two segmented sub-network structures, wherein different preset sub-network structures are used for representing resource transfer relations of different topology types;
and obtaining the resource characteristics corresponding to the resource transfer network according to the topology type of the segmented sub-network structure.
2. The method according to claim 1, wherein the obtaining the resource characteristics corresponding to the resource transfer network according to the topology type to which the segmented sub-network structure belongs comprises:
calculating the accumulated number of the sub-network structures belonging to each topological type in the at least two segmented sub-network structures;
and obtaining the resource characteristics corresponding to the resource transfer network according to the accumulated number of the sub-network structures belonging to each topology type.
3. The method of claim 2, wherein the resource transfer network comprises n nodes, n being a positive integer;
the calculating the cumulative number of the sub-network structures belonging to each topology type in the at least two divided sub-network structures includes:
acquiring a first set corresponding to an ith node, wherein the first set comprises a first neighbor node set and a first directed edge set, the first neighbor node set represents a set corresponding to a first neighbor node connected with the ith node, the node identification of the first neighbor node is larger than that of the ith node, the first directed edge set represents a set corresponding to a directed edge between the ith node and the first neighbor node, i is not more than n, and i is a positive integer;
acquiring a second set corresponding to the first neighbor node, wherein the second set comprises a second neighbor node set and a second directed edge set, the second neighbor node set represents a set corresponding to a second neighbor node connected with the first neighbor node, and the second directed edge set represents a set corresponding to a directed edge between the first neighbor node and the second neighbor node;
repeating the steps of obtaining the first set and the second set until the n nodes are traversed, and calculating the cumulative number of the sub-network structures belonging to each topology type according to the first set and the second set corresponding to the n nodes respectively.
4. The method according to claim 3, wherein said calculating a cumulative number of sub-network structures belonging to each topology type according to the first set and the second set respectively corresponding to the n nodes comprises:
in response to that the intersection set of the first set and the second set is a non-empty set and the node identifier of the second neighbor node is greater than a maximum node identifier, acquiring a directed edge between the ith node, the first neighbor node and the second neighbor node, wherein the maximum node identifier is the largest node identifier among the node identifier of the ith node and the node identifier of the first neighbor node;
acquiring a sub-network structure formed by the ith node, the first neighbor node and the second neighbor node according to directed edges among the ith node, the first neighbor node and the second neighbor node;
in the sub-network structure formed by the three, the accumulated number of the sub-network structures belonging to each topological type is calculated.
5. The method according to claim 3, wherein said calculating a cumulative number of sub-network structures belonging to each topology type according to the first set and the second set respectively corresponding to the n nodes comprises:
responding to that the intersection set of the first set and the second set is an empty set, the second neighbor node is connected with the first neighbor node and is not connected with the ith node, and acquiring a second node identifier of the second neighbor node and a third node identifier of the ith node;
responding to the second node identification larger than the third node identification, and acquiring a first directed edge between the ith node and the first neighbor node and a second directed edge between the first neighbor node and the second neighbor node;
acquiring a first sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the second directed edge;
in the first sub-network structure, taking the ith node as a target node, and calculating the accumulated number of the sub-network structures belonging to each topology type; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node.
6. The method of claim 5, further comprising:
responding to the second node identification smaller than the third node identification, and acquiring a first directed edge between the ith node and the first neighbor node and a second directed edge between the first neighbor node and the second neighbor node;
acquiring a first sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the second directed edge;
and in the first sub-network structure, the cumulative number of the sub-network structures belonging to each topological type is calculated by taking the ith node as a target node.
7. The method according to claim 3, wherein said calculating a cumulative number of sub-network structures of which the split sub-network structures belong to each topology type according to the first set and the second set respectively corresponding to the n nodes comprises:
responding to that the intersection set of the first set and the second set is an empty set, the second neighbor node is connected with the ith node and is not connected with the first neighbor node, and acquiring a second node identifier of the second neighbor node and a first node identifier of the first neighbor node;
responding to the second node identification larger than the first node identification, and acquiring a first directed edge between the ith node and the first neighbor node and a third directed edge between the ith node and the second neighbor node;
acquiring a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the third directed edge;
in the second sub-network structure, taking the ith node as a target node, and calculating the accumulated number of the sub-network structures belonging to each topology type; and calculating the accumulated number of the sub-network structures belonging to each topology type by taking the first neighbor node as a target node.
8. The method of claim 7, further comprising:
responding to the second node identification smaller than the first node identification, and acquiring a first directed edge between the ith node and the first neighbor node and a third directed edge between the ith node and the second neighbor node;
acquiring a second sub-network structure corresponding to the ith node, the first neighbor node and the second neighbor node according to the first directed edge and the third directed edge;
and in the second sub-network structure, the accumulated number of the sub-network structures belonging to each topology type is calculated by taking the first neighbor node as a target node.
9. The method according to any one of claims 1 to 8, wherein the directed edge correspondence between the nodes has weight information;
the method further comprises the following steps:
and obtaining the resource characteristics corresponding to the resource transfer network according to the type of the segmented sub-network structure and the weight information corresponding to the directed edge.
10. The method according to claim 9, wherein obtaining the resource characteristics corresponding to the resource transfer network according to the type to which the segmented sub-network structure belongs and the weight information corresponding to the directed edge includes:
calculating the accumulated number of the sub-network structures belonging to each topological type in the at least two segmented sub-network structures;
calculating the weight value corresponding to the sub-network structure belonging to the same type according to the weight information corresponding to the directed edge;
and obtaining the resource characteristics corresponding to the resource transfer network according to the accumulated quantity and the weight value.
11. The method of claim 10, wherein the resource transfer network comprises n nodes, n being a positive integer;
the calculating the weight values corresponding to the sub-network structures belonging to the same type according to the weight information corresponding to the directed edges includes:
acquiring a first adjacent triple set corresponding to an ith node, wherein the first adjacent triple set comprises a first weight set, the first weight set represents a weight value corresponding to a directed edge between the ith node and a first neighbor node, i is not more than n, and i is a positive integer;
acquiring a second adjacent triple set corresponding to the first neighbor node, wherein the second adjacent triple set comprises a second weight set, the second weight set represents a weight value corresponding to a directed edge between the first neighbor node and a second neighbor node, and the second neighbor node is connected with the first neighbor node;
repeating the steps of obtaining the first adjacent triple set and the second adjacent triple set until the n nodes are traversed, and calculating the corresponding weight values of the sub-network structures belonging to the same type according to the weight values corresponding to the directed edges between the n nodes.
12. The method according to any one of claims 1 to 9, further comprising:
calling a resource classification model to classify the resource characteristics to obtain the prediction probability that the resource transfer belongs to abnormal resource transfer;
and outputting the type of the resource transfer according to the prediction probability.
13. A resource feature extraction device based on a complex network is characterized in that the device comprises:
the acquisition module is used for acquiring the resource transfer relationship;
a building module, configured to build a resource transfer network according to the resource transfer relationship, where the resource transfer network includes at least two nodes, where the nodes are used to represent user accounts for resource transfer, and directed edges are corresponding to the nodes and used to represent resource flow directions in the resource transfer;
the segmentation module is used for segmenting the resource transfer network according to a preset sub-network structure to obtain at least two segmented sub-network structures, and different preset sub-network structures are used for representing resource transfer relations of different topology types;
and the feature extraction module is used for obtaining the resource features corresponding to the resource transfer network according to the topology type of the segmented sub-network structure.
14. A computer device comprising a processor and a memory, the memory having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, the at least one instruction, the at least one program, the set of codes, or the set of instructions being loaded and executed by the processor to implement the complex network based resource feature extraction method as claimed in any one of claims 1 to 12.
15. A computer-readable storage medium, having stored therein at least one instruction, at least one program, a set of codes, or a set of instructions, which is loaded and executed by a processor to implement the method for extracting resource features based on a complex network according to any one of claims 1 to 12.
CN202011287407.9A 2020-11-17 2020-11-17 Resource feature extraction method, device, equipment and medium based on complex network Pending CN112245937A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011287407.9A CN112245937A (en) 2020-11-17 2020-11-17 Resource feature extraction method, device, equipment and medium based on complex network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011287407.9A CN112245937A (en) 2020-11-17 2020-11-17 Resource feature extraction method, device, equipment and medium based on complex network

Publications (1)

Publication Number Publication Date
CN112245937A true CN112245937A (en) 2021-01-22

Family

ID=74265997

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011287407.9A Pending CN112245937A (en) 2020-11-17 2020-11-17 Resource feature extraction method, device, equipment and medium based on complex network

Country Status (1)

Country Link
CN (1) CN112245937A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112597399A (en) * 2021-03-08 2021-04-02 腾讯科技（深圳）有限公司 Graph data processing method and device, computer equipment and storage medium

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112597399A (en) * 2021-03-08 2021-04-02 腾讯科技（深圳）有限公司 Graph data processing method and device, computer equipment and storage medium
CN112597399B (en) * 2021-03-08 2021-07-16 腾讯科技（深圳）有限公司 Graph data processing method and device, computer equipment and storage medium
WO2022188646A1 (en) * 2021-03-08 2022-09-15 腾讯科技（深圳）有限公司 Graph data processing method and apparatus, and device, storage medium and program product

Similar Documents

Publication Publication Date Title
CN110363449B (en) Risk identification method, device and system
US10824674B2 (en) Label propagation in graphs
CN111737546B (en) Method and device for determining entity service attribute
Rong et al. A model for software defect prediction using support vector machine based on CBA
CN112396189A (en) Method and device for multi-party construction of federal learning model
CN112785005B (en) Multi-objective task assistant decision-making method and device, computer equipment and medium
CN103699541A (en) Interactive visible data mining for increasing classification accuracy
CN111339818A (en) Face multi-attribute recognition system
CN113822315A (en) Attribute graph processing method and device, electronic equipment and readable storage medium
Li et al. Research on QoS service composition based on coevolutionary genetic algorithm
CN117156442B (en) Cloud data security protection method and system based on 5G network
Vennila et al. Hybrid parallel linguistic fuzzy rules with canopy mapreduce for big data classification in cloud
EP2913760A1 (en) Efficient link management for graph clustering
CN115545103A (en) Abnormal data identification method, label identification method and abnormal data identification device
CN112245937A (en) Resource feature extraction method, device, equipment and medium based on complex network
CN113852204A (en) Three-dimensional panoramic monitoring system and method for transformer substation with digital twin
CN112836742A (en) System resource adjusting method, device and equipment
CN114049161B (en) E-commerce big data feedback-based push optimization method and E-commerce big data system
CN115412401A (en) Method and device for training virtual network embedding model and virtual network embedding
CN111984637B (en) Missing value processing method and device in data modeling, equipment and storage medium
CN110851655B (en) Method and system for simplifying complex network
Salam et al. A novel framework for web service composition
CN113779116A (en) Object sorting method, related equipment and medium
CN111459990A (en) Object processing method, system, computer readable storage medium and computer device
Sharifi et al. A queueing model for video analytics applications of smart cities

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
REG Reference to a national code

Ref country code: HK

Ref legal event code: DE

Ref document number: 40038195

Country of ref document: HK

SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination