CN113393155A - Risk cause identification method and device and storage medium - Google Patents

Risk cause identification method and device and storage medium Download PDF

Info

Publication number
CN113393155A
CN113393155A CN202110751947.6A CN202110751947A CN113393155A CN 113393155 A CN113393155 A CN 113393155A CN 202110751947 A CN202110751947 A CN 202110751947A CN 113393155 A CN113393155 A CN 113393155A
Authority
CN
China
Prior art keywords
risk
wind control
designated
node
wind
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
CN202110751947.6A
Other languages
Chinese (zh)
Inventor
李瑾瑜
张宝华
徐祎
丁凯文
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.)
Industrial and Commercial Bank of China Ltd ICBC
Original Assignee
Industrial and Commercial Bank of China Ltd ICBC
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 Industrial and Commercial Bank of China Ltd ICBC filed Critical Industrial and Commercial Bank of China Ltd ICBC
Priority to CN202110751947.6A priority Critical patent/CN113393155A/en
Publication of CN113393155A publication Critical patent/CN113393155A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • General Physics & Mathematics (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Educational Administration (AREA)
  • Game Theory and Decision Science (AREA)
  • Data Mining & Analysis (AREA)
  • Marketing (AREA)
  • General Engineering & Computer Science (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Computational Linguistics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Bioinformatics & Cheminformatics (AREA)

Abstract

The embodiment of the specification relates to the technical field of big data, and particularly discloses a risk cause identification method, a risk cause identification device and a storage medium, wherein the method comprises the following steps: acquiring a designated wind control object set corresponding to a designated risk level; merging the wind control objects of which the object labels in the designated wind control object set meet the label consistency requirement into a node, and determining edges among the nodes based on the relation among the wind control objects in the designated wind control object set to obtain a designated risk relation map corresponding to the designated risk level; and screening edges or nodes with the importance degree meeting the requirement of the specified importance degree from the specified risk relationship graph based on the edge importance degree factor and the node characteristic importance degree factor, and constructing a risk mode identification graph corresponding to the specified risk grade to identify the risk cause of the wind control object based on the risk mode identification graph, so that the accuracy of risk cause identification is further improved.

Description

Risk cause identification method and device and storage medium
Technical Field
The present disclosure relates to the field of big data technologies, and in particular, to a method and an apparatus for identifying risk cause, and a storage medium.
Background
Customer risk level identification is an important element of risk management. The relationship between the client and the client is connected together in a map mode, risk level identification is realized by using a neural network model, the client association information can be more fully utilized, and a better client risk level identification result is obtained. However, in practical application scenarios, the client characteristics and the relationship between clients are complex and changeable, and how to better perform risk cause analysis becomes a technical problem to be solved urgently in the field.
Disclosure of Invention
An object of the embodiments of the present specification is to provide a risk cause identification method, an apparatus, and a storage medium, which can identify influence factors causing wind-controlled objects to be at different risk levels, and further improve accuracy of risk cause identification.
The present specification provides a risk cause identification method, apparatus and storage medium, which are implemented in the following manner:
a risk cause identification method is applied to a server and comprises the following steps: acquiring a designated wind control object set corresponding to a designated risk level; the appointed wind control object set comprises wind control objects corresponding to the appointed risk levels; merging the wind control objects of which the object labels in the designated wind control object set meet the label consistency requirement into a node, and determining edges among the nodes based on the relation among the wind control objects in the designated wind control object set to obtain a designated risk relation map corresponding to the designated risk level; node features corresponding to all nodes in the designated risk relationship graph are determined based on object features of the wind control objects merged into the corresponding nodes; and screening edges or nodes with the importance degree meeting the requirement of the specified importance degree from the specified risk relationship graph based on the edge importance degree factor and the node characteristic importance degree factor, and constructing a risk mode identification graph corresponding to the specified risk grade so as to identify the risk cause of the wind control object based on the risk mode identification graph.
In another aspect, an embodiment of the present specification provides a risk cause identification device, which is applied to a server, and the device includes: the acquisition module is used for acquiring a designated wind control object set corresponding to a designated risk level; the appointed wind control object set comprises wind control objects corresponding to the appointed risk levels; the map construction module is used for merging the wind control objects of which the labels of the objects in the designated wind control object set meet the requirement of label consistency into a node, and determining edges among the nodes based on the relation among the wind control objects in the designated wind control object set to obtain a designated risk relation map corresponding to the designated risk level; node features corresponding to all nodes in the designated risk relationship graph are determined based on object features of the wind control objects merged into the corresponding nodes; and the risk mode determination module is used for screening edges or nodes with the importance degree meeting the requirement of the specified importance degree from the specified risk relationship graph based on the edge importance degree factor and the node characteristic importance degree factor, constructing a risk mode identification graph corresponding to the specified risk grade, and identifying the risk cause of the wind control object based on the risk mode identification graph.
In another aspect, the present specification provides a computer readable storage medium, on which computer instructions are stored, and the instructions, when executed, implement the steps of the method according to any one or more of the above embodiments.
According to the risk cause identification method, the risk cause identification device and the storage medium provided by one or more embodiments of the present specification, by merging the wind control objects with the similar object tags into one node, the object tags, the object characteristics of the wind control objects with the similar object tags and the influence of the relationship among the wind control objects on the wind control objects falling into a certain risk level can be more effectively represented. Furthermore, the edge importance factor and the node feature importance factor in the combined risk relationship graph are determined, edges and nodes which have large influence on the wind control object falling into a certain risk level are extracted based on the edge importance factor and the node feature importance factor, the risk pattern recognition graph is constructed based on the extracted edges and nodes, the object labels, the object features of the wind control objects with the similarity object labels and the influence of the relation between the wind control objects on the wind control object falling into the certain risk level can be represented more simply and accurately, and the accuracy of risk cause recognition is further improved.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments described in the present specification, and for those skilled in the art, other drawings can be obtained according to the drawings without any creative effort. In the drawings:
FIG. 1 is a schematic view of a risk pattern recognition process provided herein;
FIG. 2 is a schematic flow chart of risk cause analysis provided herein;
FIG. 3 is a schematic flow chart of an implementation of the risk cause identification method provided in the present specification;
fig. 4 is a schematic block diagram of a risk cause identification device provided in this specification.
Detailed Description
In order to make those skilled in the art better understand the technical solutions in the present specification, the technical solutions in one or more embodiments of the present specification will be clearly and completely described below with reference to the drawings in one or more embodiments of the present specification, and it is obvious that the described embodiments are only a part of the embodiments of the specification, and not all embodiments. All other embodiments obtained by a person skilled in the art based on one or more embodiments of the present specification without making any creative effort shall fall within the protection scope of the embodiments of the present specification.
In one example scenario provided in this specification, the risk cause identification method may be applied to a single server or a server cluster composed of multiple servers. As shown in fig. 1 to 2, the risk cause identification method may include at least two links of risk pattern identification and risk cause analysis. As shown in fig. 1, the risk prediction model may be constructed by using the following steps S101 to S103:
step S101: and constructing an initial wind control object relation map.
Defining graph structure G ═ V, E, V represents nodes, and E represents edges between nodes. Corresponding to the wind control object risk classification scene, V represents a wind control object, and E represents the relation between the wind control object and the wind control object. The wind-controlled object may be, for example, an individual user, a business, or the like. Correspondingly, the relationship of the wind control objects can be an investment relationship, a guarantee relationship and the like. Wind-controlled object and wind-controlled object (v)i,vj) The relationship between is ei,jDifferent ei,jAdjacent matrix A belonging to R among wind control object relationsn*nInitial state Ai,j＝ei,j. When (v)i,vj) When there is a direct association in the relationship graph ei,j1, otherwisei,j＝0。
Record each wind-controlled object viWith K characteristic variables
Figure BDA0003145044390000031
For each wind-controlled object viRecord its object tag
Figure BDA0003145044390000032
Representing wind-controlled objects viHaving the label and vice versa
Figure BDA0003145044390000033
Marking wind-controlled objects viRisk rating
Figure BDA0003145044390000034
When the wind control object viWith a risk class of c is given to,
Figure BDA0003145044390000035
otherwise, the reverse is carried out
Figure BDA0003145044390000036
Each wind-controlled object can only correspond to one risk level.
Step S102: and identifying the risk level of the wind control object.
Let the model be
Figure BDA0003145044390000037
Representing risk level, phi (·) representing a graph neural network model, G (v) representing a graph structure relation of v, X (v) representing a feature vector of v, and B (v) representing a label vector of v.
Step S103: and (4) risk pattern recognition.
And respectively taking N typical wind control objects for the wind control object set of each risk level c. And integrating typical wind control objects in each wind control object set into a big graph to obtain an appointed risk relation graph corresponding to the risk level c. The typical wind control objects may be all wind control objects under the risk level c, or may be part of wind control objects extracted according to a certain requirement. If the wind control objects with more label types and more characteristic variable types and more relationship information which can be acquired by other wind control objects can be extracted, the finally obtained risk pattern recognition result is more accurate and representative by extracting the wind control objects of the type to perform risk pattern recognition. Of course, the service personnel can set the typical wind control object selection mode according to the requirements of the service scene, and the selection mode is not limited here.
For the wind control level c, wind control object points of the same type can be merged into the same node. Wind control objects of the same type are mainly judged according to the labels, and the labels with higher consistency are combined into the same node. And marking the association relation between the nodes according to the wind control objects associated with the nodes. And on the basis of the initial combination and the node association of the server, the service personnel adjusts the point-edge relationship so as to enable the point-edge relationship to be more accurate and obtain an appointed risk relationship map corresponding to the risk level c. The mean value of the corresponding characteristic variables of the wind-controlled objects merged into a node may be used as the value of the characteristic variable corresponding to the node, and used as the node characteristic vector of the node. And information such as the object label related to each node and the occurrence frequency of each object label in the designated risk relationship graph can be recorded as the node label of the corresponding node.
The relevant symbols are defined as follows:
marking the designated risk relation map corresponding to the wind control grade c as Gc(v) The adjacent junction matrix is Ac(v) The node feature vector is Xc(v)＝{xi|vi∈Gc(v) }, label vector Bc(v)＝{bi|vi∈Gc(v) }, prediction result PΦ(Y|Gc,Xc)。
Sub-graphs of the assigned risk relationship graph can be extracted, so that the calculation data volume is reduced, and the resource consumption of the system during calculation is reduced. E.g. at Gc(v) Within range, sub-graph G is extractedc,S(v) In that respect The system defaults to wind the object viAnd 3, extending the relation of steps outwards as a starting point. Other ways of sub-graph extraction may also include: defining a number of label-gated objects, defining a number of nodes in the sub-graph, defining a number of edges in the sub-graph, etc. The subgraph extraction mode can be set according to actual business requirements. Of course, if the data size is not large, sub-graph extraction may not be performed, and subsequent processing is directly performed based on the designated risk relationship graph, and the processing mode is the same as the processing mode performed based on the extracted sub-graph, which is not described herein again.
Suppose that the extracted subgraph is labeled Gc,S(v) The adjacent junction matrix is Ac,S(v) The node feature vector is Xc,S(v)＝{xi|vi∈Gc,S(v) }, node label vector Bc,S(v)＝{bi|vi∈Gc,S(v) }, prediction result PΦ(Y|Gc,S,Xc,S). Of course, if sub-graph extraction is not performed, and subsequent analysis is directly performed based on the designated risk relationship graph, it can be understood that the sub-graph mark is identical to the G of the designated risk relationship graphc(v)、Ac(v)、Xc(v)＝{xi|vi∈Gc(v)}、Bc(v)＝{bi|vi∈Gc(v)}、PΦ(Y|Gc,Xc) Have the same meaning, but are shown in different ways.
The recognition pattern is obtained by solving the following first objective function as follows.
Figure BDA0003145044390000051
Figure BDA0003145044390000052
Wherein,
Figure BDA0003145044390000053
is an n-dimensional vector, and the value of each element is 1; y represents the target variable, Y represents the vector formed by the target variable, C represents the risk level, C represents the upper limit value of the risk level, PΦRepresenting the result of risk prediction, G represents a map, Ac,SAn adjacency matrix representing the assigned risk relationship graph,. point multiplication representing the matrix,. sigma (M) represents the sigmoid function, and. sigma (M) belongs to [0,1 ]]n*nN represents the adjacent junction matrix Ac,SM represents the masking function of the edge, H (·) represents the entropy, X represents the feature vector, Xc,SRepresenting a feature vector formed by node features of the designated risk relationship graph; gc,SRepresenting a designated risk relationship graph; f represents a masking function of the node characteristics, and F is equal to {0,1}KAnd K represents the feature variable number of the node feature.
By the above objective function, can pass through Ac,Sσ (M) will Ac,SAnd the important edge relation influencing the prediction result is reserved, and the secondary edge relation is weakened. F is corresponding to {0,1}KThe method is a mask function of the features, and retains the features which have important influence on the prediction result and weakens the secondary features through the value of 0-1. Of course, if a feature or edge is important and must be preserved, the M or F middle pair can also be fixedThe value of the element.
The mask function of the edge and the mask function of the node feature obtained by solving the objective function can be respectively marked as M*、F*. Correspondingly, the relation mode of the risk mode recognition map can be obtained
Figure BDA0003145044390000054
And characteristic patterns
Figure BDA0003145044390000055
Figure BDA0003145044390000056
Each element of (1)
Figure BDA0003145044390000057
Representing a node viAnd vjThe strength of the relationship. The larger the value is, the more important the relationship corresponding to the edge is to the prediction result. By setting a certain threshold, the edges with lower importance are screened out, the edges with high importance are reserved, and the method obtains
Figure BDA0003145044390000058
Figure BDA0003145044390000061
Indicating how important the node characteristics of a certain node are. Reserving the node indicates importance to the prediction result, and nodes that are not reserved indicate lower importance. Eliminating nodes with lower importance, reserving nodes with high importance, and obtaining
Figure BDA0003145044390000062
By using
Figure BDA0003145044390000063
Construction of Risk Pattern recognition maps
Figure BDA0003145044390000064
Wherein,
Figure BDA0003145044390000065
namely the neighborhood matrix of the risk pattern recognition map.
Figure BDA0003145044390000066
Representing labels on reserved nodes and corresponding relation marks of edges, and describing the wind control object v by wordsiAnd factors influencing different risk levels are included, so that the risk interpretation is closer to the business scene and is convenient to understand.
Correspondingly, the server can identify the map according to the risk mode to generate the risk description of the wind control object. The wind control object risk description at least comprises one of a node label of a node where the wind control object is located, node characteristics and a risk conduction mode. The risk conduction mode can be characterized by edges between the node where the wind control object is located and other nodes and edge importance factors of the edges. For example, the risk transmission pattern may be expressed as:
"× × × (risk level) wind control object viThe first-level risk propagation model of (1) is the edge relationship A1 (e.g., loan) (influence importance, taking the values of the corresponding elements of the adjacency matrix, the same below), A2 (e.g., guarantee) (influence importance), A3 (influence importance). Secondary risk conduction patterns are edge relationships B1 (influence significance), B2 (influence significance), B3 (influence significance), B4 (influence significance). "wherein A1, A2 and A3 are related to the wind-controlled object viThe node relation corresponding to the edge directly connected with the node Q, B1, B2, B3 and B4 are the node relation corresponding to the wind control object viThe node Q is separated by the node relation corresponding to the edge connected with one node. Of course, a third-level risk transmission mode, a fourth-level risk transmission mode, etc. may be included, but generally, the farther apart from the node Q, the smaller the risk influence of the edge relation on the node Q, and the corresponding risk transmission progression may be selected for analysis as required.
According to the embodiment, the wind control objects with the similar object labels are combined to one node, so that the influence of the object labels, the object characteristics of the wind control objects with the similar object labels and the relation among the wind control objects on the wind control objects falling into a certain risk level can be more effectively represented. Furthermore, edges and nodes which have large influence on the wind control object falling into a certain risk level are extracted based on the edge importance factor and the node feature importance factor in the combined risk relationship graph, and the risk pattern recognition graph is constructed based on the extracted edges and nodes, so that the object labels, the object features of the wind control objects with the similarity object labels and the influence of the relation between the wind control objects on the wind control object falling into the certain risk level can be represented more simply and accurately.
Risk influence factor analysis can be performed according to the risk pattern recognition maps corresponding to the risk grades. The risk pattern recognition graph is constructed based on typical wind control objects, and the typical wind control objects may not fully represent the characteristics of all the wind control objects at each risk level.
FIG. 2 is a flow diagram of a risk cause analysis module.
Step S201: risk level cause analysis of wind-controlled objects
For the set of wind-controlled objects of risk level c, for each wind-controlled object v thereofiAnd extracting subgraph, wherein the subgraph extraction rule is the same as the steps. The wind control object relation subgraph extracted by the marks is Gi,c,S(v) The adjacent junction matrix is Ai,c,S(v) The feature vector is Xi,c,S(v)＝{xi|vi∈Gi,c,S(v)}。
Solving the following second objective function:
Figure BDA0003145044390000071
obtaining each wind control object v under the risk level ciCorresponding Pc TAi,c,SPc. Wherein, PcIndicating the adjustment coefficientThe matrix is a matrix of a plurality of matrices,
Figure BDA00031450443900000717
represents a pair PcThe transposing of (1). Can be used for each wind control object v under the risk level ciCorresponding to
Figure BDA00031450443900000718
Taking the mean or median to obtain
Figure BDA0003145044390000072
Or
Figure BDA0003145044390000073
Can be combined with
Figure BDA0003145044390000074
As a reference neighbor matrix for risk class c, use is made of
Figure BDA0003145044390000075
To characterize the risk transmission pattern of the risk class c.
As can be seen from the above-described processing procedure,
Figure BDA0003145044390000076
risk transmission patterns of risk class c may also be characterized, but
Figure BDA0003145044390000077
The risk objects on which the calculations are based may not be comprehensive enough in
Figure BDA0003145044390000078
On the basis, the calculation is further carried out in the way
Figure BDA0003145044390000079
By using
Figure BDA00031450443900000710
The risk conduction mode of the risk level c is characterized, so that the risk conduction characteristics of the risk level c can be more accurately and comprehensively characterized.Of course, by first basing on typical wind-controlled objects
Figure BDA00031450443900000711
And then the wind control object set based on the risk level c is calculated
Figure BDA00031450443900000712
By using
Figure BDA00031450443900000713
The risk conduction mode of the risk level c is characterized, the whole calculation amount can be reduced, and the whole processing efficiency is improved.
Of course, each wind-controlled object v under the risk level c can also be obtained through the objective functioniCorresponding to
Figure BDA00031450443900000719
Or for each wind-controlled object v under the risk level ciCorresponding to
Figure BDA00031450443900000720
Taking the mean or median to obtain
Figure BDA00031450443900000714
Can be combined with
Figure BDA00031450443900000715
As reference node characteristics of risk level c, use is made of
Figure BDA00031450443900000716
To characterize the impact of the object features on the wind control object falling into the risk level c.
Correspondingly, the server can also take any object label of the risk level c as a designated object label, and extract any wind control object v of the designated object label under the risk level ciCorresponding wind-controlled object relation subgraph Gi,c,S(v) The frequency of occurrence of (1) as a designated label frequency; calculating a relationship sub-graph G corresponding to each wind-controlled object under the risk level ci,c,S(v) The mean value or the median of the assigned label frequency of the assigned object label is obtained, and the reference label frequency of the assigned object label under the risk level c is obtained, so that the risk level cause of the risk level c is analyzed according to the reference label frequency. By calculating the frequency of the reference label in the above manner, the influence of the object label on the wind control object falling into the risk level c can be more accurately and comprehensively represented by using the frequency of the reference label.
Step S202: wind control object risk level change scenario analysis
For a target wind control object to be identified by a risk cause, a server can obtain a target wind control object relation subgraph corresponding to the target wind control object; calculating a neighbor matrix of the target wind control object relation subgraph as a target neighbor matrix; and comparing the target adjacent joint matrix with the reference adjacent joint matrix under each risk level to determine the influence factors of the target wind control object falling into each risk level.
For example, to wind a target object viAnd extracting according to a sub-graph extraction mode to obtain a target wind control object relation sub-graph corresponding to the target wind control object. And can calculate to obtain a target wind control object viCorresponding adjacent junction matrix Ai,S(vi). Can be combined with Ai,S(vi) With A at different risk levelscAnd comparing and analyzing the difference of the two adjacent junction matrixes. The service personnel can adjust the association degree between any nodes in the target wind control object relation subgraph, and the server can calculate the adjacency matrix of the adjusted target wind control object relation subgraph and compare Ai,S(vi) To the corresponding element of the adjusted node
Figure BDA0003145044390000081
If the element values are closer, the adjustment operation of the association degree between the two nodes can be explained so that the wind-controlled object v is controlled by the windiThe likelihood of falling into risk class c is greater.
For example, Ai,S(vi)[vp,vq]＝1,
Figure BDA0003145044390000082
Description of viIf the associated node v in its subgraph is loweredpAnd vqDegree of association of (d), that viThe risk level of (c) may become c. Further analysis, if the object v is controlled by windiWhen the risk level (assuming that the calculation result is B) is changed into B +/B-, the conduction factors influencing the change are certain, and the association degree between the nodes which needs to be changed can be improved/reduced to generate a risk analysis report.
Therefore, risk cause identification is carried out in the mode, accuracy and comprehensiveness of risk analysis and investigation can be greatly improved, business personnel can determine risk influence factors of the wind control object falling into a certain risk grade clearly and intuitively, and accuracy, comprehensiveness and efficiency of risk cause identification in a complex business scene are greatly improved.
Based on the above scenario example, the present specification further provides a risk cause identification method. Fig. 3 is a schematic flow chart of an embodiment of the risk cause identification method provided in this specification. As shown in fig. 3, in one embodiment of the risk cause identification method provided by the present specification, the method may be applied to a server. The method may comprise the following steps.
S30: acquiring a designated wind control object set corresponding to a designated risk level; and the appointed wind control object set comprises wind control objects corresponding to the appointed risk level.
Wherein the specified risk level may be any risk level. The designated set of wind control objects may include all wind control objects under the designated risk level, or may include only extracted typical wind control objects, which is not limited herein. The typical extraction manner of the wind control object can refer to the scene example. Through the mode of extracting the typical wind control object, the data volume of subsequent processing can be greatly reduced, and the processing efficiency is improved.
S32: merging the wind control objects of which the object labels in the designated wind control object set meet the label consistency requirement into a node, and determining edges among the nodes based on the relation among the wind control objects in the designated wind control object set to obtain a designated risk relation map corresponding to the designated risk level; and determining node characteristics corresponding to each node in the appointed risk relationship graph based on the object characteristics of the wind control object merged to the corresponding node.
The object tag meeting the tag consistency requirement can be configured as required, which is not limited herein. By setting the requirement of tag consistency, the object tags with higher object tag consistency can be extracted, so that the part of the wind control objects is combined into one node. Other embodiments of the construction of the designated risk relationship graph may refer to the above scenario example, which is not described herein again.
S34: and screening edges or nodes with the importance degree meeting the requirement of the specified importance degree from the specified risk relationship graph based on the edge importance degree factor and the node characteristic importance degree factor, and constructing a risk mode identification graph corresponding to the specified risk grade so as to identify the risk cause of the wind control object based on the risk mode identification graph.
In some embodiments, the edge importance factor and the node feature importance factor may be determined in the following manner. Solving a first objective function based on the designated risk relationship graph:
Figure BDA0003145044390000091
wherein,
Figure BDA0003145044390000092
is an n-dimensional vector, and the value of each element is 1; y represents the target variable, Y represents the vector formed by the target variable, C represents the risk level, C represents the upper limit value of the risk level, PΦRepresenting the result of risk prediction, G represents a map, Ac,SAn adjacency matrix representing the assigned risk relationship graph,. point multiplication representing the matrix,. sigma (M) represents the sigmoid function, and. sigma (M) belongs to [0,1 ]]n*nN represents the adjacent junction matrix Ac,SM represents the masking function of the edge, H (·) represents the entropy, X represents the feature vector, Xc,SIndicating a designated risk gateFeature vectors formed by the node features of the family graph; gc,SRepresenting a designated risk relationship graph; f represents a masking function of the node characteristics, and F is equal to {0,1}KK represents the feature variable number of the node feature;
respectively marking mask functions of edges and node characteristics obtained by solving the objective function as M*、F*And is expressed by σ (M) respectively*)、F*As edge importance factors and node feature importance factors.
Then, the edges or nodes with the importance degree meeting the specified importance degree requirement can be screened based on the edge importance degree factor and the node characteristic importance degree factor. If the extracted relation mode is
Figure BDA0003145044390000101
Characteristic patterns
Figure BDA0003145044390000102
And the extracted relation mode and the extracted characteristic mode also indirectly represent the influence importance of the edge relation and the node characteristics on the wind control object falling into the risk level c. Meanwhile, each node is also correspondingly provided with a node label, and the influence importance of the object label on the wind control object falling into the risk level c can be visually determined through the occurrence frequency of the label in the node label, so that the influence factor of the wind control object falling into a certain risk level can be accurately and visually determined based on the risk pattern recognition map.
Of course, the edge importance factor and the node feature importance factor are merely examples, and when the determination process is implemented, suitable modifications to the determination process are also included in the protection scope of the embodiments of the present specification. In addition to the masking function, of course, other functions with similar characteristics may be used to characterize the impact of edges or features on the prediction,
in some embodiments, the risk pattern recognition graph may further include a wind-controlled object risk description; the risk description of the wind control object at least comprises one of a node label, a node characteristic and a risk conduction mode of a node where the wind control object is located; and the risk conduction mode is characterized by utilizing edges between the node where the wind control object is located and other nodes and edge importance factors of the edges.
In other embodiments, the following processing is further performed based on the determination of the risk pattern recognition map based on the typical wind control object. Solving a second objective function based on the risk pattern recognition map to obtain
Figure BDA0003145044390000107
：
Figure BDA0003145044390000103
Wherein, PcA matrix of adjustment coefficients is represented by,
Figure BDA0003145044390000108
represents a pair PcThe transpose of (a) is performed,
Figure BDA0003145044390000104
a relational schema representing a risk pattern recognition graph,
Figure BDA0003145044390000105
a characteristic pattern representing a risk pattern recognition profile,
Figure BDA0003145044390000106
Ai,c,Srepresenting wind-controlled objects v at a risk level ciAdjacent matrix, X, of the corresponding wind-controlled object relationship subgraphi,c,SRepresenting wind-controlled objects viThe object characteristics of (1). According to the risk class c, corresponding to each wind-controlled object
Figure BDA0003145044390000109
And determining a reference adjacent node matrix of the risk grade c so as to identify the risk cause of the wind control object according to the reference adjacent node matrix. E.g. to calculate the correspondence of each wind-controlled object
Figure BDA0003145044390000111
The mean or median, etc., to obtain a baseline neighbor matrix for risk level c.
Correspondingly, in other embodiments, the server may obtain, for a target wind control object to be risk cause identified, a target wind control object relationship subgraph corresponding to the target wind control object; calculating a neighbor matrix of the target wind control object relation subgraph as a target neighbor matrix; and comparing the target adjacent joint matrix with the reference adjacent joint matrix under each risk level to determine the influence factors of the target wind control object falling into each risk level.
In other embodiments, the server may further use any object tag of the risk level c as a designated object tag, and extract any wind control object v of the designated object tag under the risk level ciCorresponding wind-controlled object relation subgraph Gi,c,S(v) As the designated label frequency. According to the risk level c, corresponding to each wind control object relation subgraph Gi,c,S(v) Determining the reference label frequency of the designated object label under the risk level c, and analyzing the risk level cause of the risk level c according to the reference label frequency.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. For details, reference may be made to the description of the related embodiments of the related processing, and details are not repeated herein.
The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
As shown in fig. 4, based on the method provided by the foregoing embodiment, an embodiment of this specification further provides a risk cause identification device, applied to a server, where the device includes: the obtaining module 40 may be configured to obtain a designated wind control object set corresponding to a designated risk level; and the appointed wind control object set comprises wind control objects corresponding to the appointed risk level. The map building module 42 may be configured to combine the wind control objects whose object labels in the designated wind control object set meet the requirement of label consistency into one node, and determine edges between the nodes based on a relationship between the wind control objects in the designated wind control object set, so as to obtain a designated risk relationship map corresponding to the designated risk level; and determining node characteristics corresponding to each node in the appointed risk relationship graph based on the object characteristics of the wind control object merged to the corresponding node. The risk mode determination module 44 may be configured to screen, from the designated risk relationship graph, edges or nodes whose importance degrees meet a designated importance degree requirement based on the edge importance degree factor and the node feature importance degree factor, and construct a risk mode identification graph corresponding to the designated risk level, so as to identify a risk cause of the wind control object based on the risk mode identification graph.
In other embodiments, the risk pattern determination module is further configured to determine the edge importance factor and the node feature importance factor based on: solving a first objective function based on the designated risk relationship graph:
Figure BDA0003145044390000121
wherein,
Figure BDA0003145044390000122
is an n-dimensional vector, and the value of each element is 1; y represents the target variable, Y represents the vector formed by the target variable, C represents the risk level, C represents the upper limit value of the risk level, PΦRepresenting the result of risk prediction, G represents a map, Ac,SA neighborhood matrix representing the assigned risk relationship graph,. a point product representing the matrix,. sigma (M) represents the sigmoid function,
Figure BDA0003145044390000123
n represents the adjacent junction matrix Ac,SM represents the masking function of the edge, H (·) represents the entropy, X represents the feature vector, Xc,SRepresenting a feature vector formed by node features of the designated risk relationship graph; gc,SRepresenting a designated risk relationship graph; f represents a masking function of the node characteristics, and F is equal to {0,1}KK represents the feature variable number of the node feature; respectively marking mask functions of edges and node characteristics obtained by solving the objective function as M*、F*And is expressed by σ (M) respectively*)、F*As edge importance factors and node feature importance factors.
In other embodiments, the apparatus further comprises a reference risk model determining module, configured to solve a second objective function based on the risk pattern recognition map to obtain
Figure BDA0003145044390000128
Figure BDA0003145044390000124
Wherein, PcA matrix of adjustment coefficients is represented by,
Figure BDA0003145044390000129
represents a pair PcThe transpose of (a) is performed,
Figure BDA0003145044390000125
a relational schema representing a risk pattern recognition graph,
Figure BDA0003145044390000126
a characteristic pattern representing a risk pattern recognition profile,
Figure BDA0003145044390000127
Ai,c,Srepresenting wind-controlled objects v at a risk level ciAdjacent matrix, X, of the corresponding wind-controlled object relationship subgraphi,c,SRepresenting wind-controlled objects viThe object characteristics of (1); and also for each wind-controlled object under risk level c
Figure BDA00031450443900001210
And determining a reference adjacent node matrix of the risk grade c so as to identify the risk cause of the wind control object according to the reference adjacent node matrix.
It should be noted that the above-mentioned apparatus may also include other embodiments according to the description of the above-mentioned embodiments. The specific implementation manner may refer to the description of the related method embodiment, and is not described in detail herein.
The present specification also provides a computer readable storage medium having stored thereon computer instructions which, when executed, implement steps of a method comprising any one or more of the embodiments described above. The storage medium may include a physical device for storing information, and typically, the information is digitized and then stored using an electrical, magnetic, or optical media. The storage medium may include: devices that store information using electrical energy, such as various types of memory, e.g., RAM, ROM, etc.; devices that store information using magnetic energy, such as hard disks, floppy disks, tapes, core memories, bubble memories, and usb disks; devices that store information optically, such as CDs or DVDs. Of course, there are other ways of storing media that can be read, such as quantum memory, graphene memory, and so forth.
It should be noted that the embodiments of the present disclosure are not limited to the cases where the data model/template is necessarily compliant with the standard data model/template or the description of the embodiments of the present disclosure. Certain industry standards, or implementations modified slightly from those described using custom modes or examples, may also achieve the same, equivalent, or similar, or other, contemplated implementations of the above-described examples. The embodiments using these modified or transformed data acquisition, storage, judgment, processing, etc. may still fall within the scope of the alternative embodiments of the present description.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment. In the description of the specification, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the specification. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
The above description is only an example of the present specification, and is not intended to limit the present specification. Various modifications and alterations to this description will become apparent to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification should be included in the scope of the claims of the present specification.

Claims (10)

1. A risk cause identification method is applied to a server, and comprises the following steps:
acquiring a designated wind control object set corresponding to a designated risk level; the appointed wind control object set comprises wind control objects corresponding to the appointed risk levels;
merging the wind control objects of which the object labels in the designated wind control object set meet the label consistency requirement into a node, and determining edges among the nodes based on the relation among the wind control objects in the designated wind control object set to obtain a designated risk relation map corresponding to the designated risk level; node features corresponding to all nodes in the designated risk relationship graph are determined based on object features of the wind control objects merged into the corresponding nodes;
and screening edges or nodes with the importance degree meeting the requirement of the specified importance degree from the specified risk relationship graph based on the edge importance degree factor and the node characteristic importance degree factor, and constructing a risk mode identification graph corresponding to the specified risk grade so as to identify the risk cause of the wind control object based on the risk mode identification graph.
2. The method of claim 1, wherein the risk pattern recognition graph further comprises a wind-controlled object risk description; the risk description of the wind control object at least comprises one of a node label, a node characteristic and a risk conduction mode of a node where the wind control object is located; and the risk conduction mode is characterized by utilizing edges between the node where the wind control object is located and other nodes and edge importance factors of the edges.
3. The method of claim 1, wherein the edge importance factor and the node feature importance factor are determined as follows:
solving a first objective function based on the designated risk relationship graph:
Figure FDA0003145044380000011
wherein,
Figure FDA0003145044380000012
is an n-dimensional vector, and the value of each element is 1; y represents the target variable, Y represents the vector formed by the target variable, C represents the risk level, C represents the upper limit value of the risk level, PΦRepresenting the result of risk prediction, G represents a map, Ac,SAn adjacency matrix representing the assigned risk relationship graph,. point multiplication representing the matrix,. sigma (M) represents the sigmoid function, and. sigma (M) belongs to [0,1 ]]n*nN represents the adjacent junction matrix Ac,SThe number of rows or columns of (c), M represents the masking function of the edge, H (-) represents the entropy, X represents the feature vector,Xc,Srepresenting a feature vector formed by node features of the designated risk relationship graph; gc,SRepresenting a designated risk relationship graph; f represents a masking function of the node characteristics, and F is equal to {0,1}KK represents the feature variable number of the node feature;
respectively marking mask functions of edges and node characteristics obtained by solving the objective function as M*、F*And is expressed by σ (M) respectively*)、F*As edge importance factors and node feature importance factors.
4. The method of claim 3, wherein identifying risk causes of the wind-controlled object based on the risk pattern recognition graph comprises:
solving a second objective function based on the risk pattern recognition map to obtain Pc TAi,c,SPc：
Figure FDA0003145044380000021
Wherein, PcA matrix of adjustment coefficients is represented by,
Figure FDA0003145044380000022
represents a pair PcThe transpose of (a) is performed,
Figure FDA0003145044380000023
a relational schema representing a risk pattern recognition graph,
Figure FDA0003145044380000024
Figure FDA0003145044380000025
a characteristic pattern representing a risk pattern recognition profile,
Figure FDA0003145044380000026
Ai,c,Srepresenting wind-controlled objects v at a risk level ciAdjacent matrix, X, of the corresponding wind-controlled object relationship subgraphi,c,SRepresenting wind-controlled objects viThe object characteristics of (1);
according to the risk class c, corresponding to each wind-controlled object
Figure FDA0003145044380000027
And determining a reference adjacent node matrix of the risk grade c so as to identify the risk cause of the wind control object according to the reference adjacent node matrix.
5. The method of claim 4, wherein the identifying risk causes of the wind-controlled object according to the reference neighbor matrix comprises:
for a target wind control object to be identified by a risk cause, acquiring a target wind control object relation subgraph corresponding to the target wind control object;
calculating a neighbor matrix of the target wind control object relation subgraph as a target neighbor matrix;
and comparing the target adjacent joint matrix with the reference adjacent joint matrix under each risk level to determine the influence factors of the target wind control object falling into each risk level.
6. The method of claim 1, further comprising:
taking any object label of the risk level c as a designated object label, and extracting any wind control object v of the designated object label under the risk level ciCorresponding wind-controlled object relation subgraph Gi,c,S(v) The frequency of occurrence of (1) as a designated label frequency;
according to the risk level c, corresponding to each wind control object relation subgraph Gi,c,S(v) Determining the reference label frequency of the designated object label under the risk level c, and analyzing the risk level cause of the risk level c according to the reference label frequency.
7. A risk cause identification device applied to a server, the device comprising:
the acquisition module is used for acquiring a designated wind control object set corresponding to a designated risk level; the appointed wind control object set comprises wind control objects corresponding to the appointed risk levels;
the map construction module is used for merging the wind control objects of which the labels of the objects in the designated wind control object set meet the requirement of label consistency into a node, and determining edges among the nodes based on the relation among the wind control objects in the designated wind control object set to obtain a designated risk relation map corresponding to the designated risk level; node features corresponding to all nodes in the designated risk relationship graph are determined based on object features of the wind control objects merged into the corresponding nodes;
and the risk mode determination module is used for screening edges or nodes with the importance degree meeting the requirement of the specified importance degree from the specified risk relationship graph based on the edge importance degree factor and the node characteristic importance degree factor, constructing a risk mode identification graph corresponding to the specified risk grade, and identifying the risk cause of the wind control object based on the risk mode identification graph.
8. The apparatus of claim 7, wherein the risk pattern determination module is further configured to determine the edge importance factor and node feature importance factor based on:
solving a first objective function based on the designated risk relationship graph:
Figure FDA0003145044380000031
wherein,
Figure FDA0003145044380000032
is an n-dimensional vector, and the value of each element is 1; y represents the target variable, Y represents the vector formed by the target variable, C represents the risk level, C represents the windUpper limit of risk class, PΦRepresenting the result of risk prediction, G represents a map, Ac,SAn adjacency matrix representing the assigned risk relationship graph,. point multiplication representing the matrix,. sigma (M) represents the sigmoid function, and. sigma (M) belongs to [0,1 ]]n*nN represents the adjacent junction matrix Ac,SM represents the masking function of the edge, H (·) represents the entropy, X represents the feature vector, Xc,SRepresenting a feature vector formed by node features of the designated risk relationship graph; gc,SRepresenting a designated risk relationship graph; f represents a masking function of the node characteristics, and F is equal to {0,1}KK represents the feature variable number of the node feature;
respectively marking mask functions of edges and node characteristics obtained by solving the objective function as M*、F*And is expressed by σ (M) respectively*)、F*As edge importance factors and node feature importance factors.
9. The apparatus of claim 8, further comprising a reference risk model determination module for solving a second objective function based on the risk pattern recognition graph to obtain
Figure FDA0003145044380000041
Figure FDA0003145044380000042
Wherein, PcA matrix of adjustment coefficients is represented by,
Figure FDA0003145044380000043
represents a pair PcThe transpose of (a) is performed,
Figure FDA0003145044380000044
a relational schema representing a risk pattern recognition graph,
Figure FDA0003145044380000045
Figure FDA0003145044380000046
a characteristic pattern representing a risk pattern recognition profile,
Figure FDA0003145044380000047
Ai,c,Srepresenting wind-controlled objects v at a risk level ciAdjacent matrix, X, of the corresponding wind-controlled object relationship subgraphi,c,SRepresenting wind-controlled objects viThe object characteristics of (1);
and also for each wind-controlled object under risk level c
Figure FDA0003145044380000048
And determining a reference adjacent node matrix of the risk grade c so as to identify the risk cause of the wind control object according to the reference adjacent node matrix.
10. A computer-readable storage medium having stored thereon computer instructions, wherein the instructions, when executed, implement the steps of the method of any one of claims 1-6.
CN202110751947.6A 2021-07-02 2021-07-02 Risk cause identification method and device and storage medium Pending CN113393155A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110751947.6A CN113393155A (en) 2021-07-02 2021-07-02 Risk cause identification method and device and storage medium

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110751947.6A CN113393155A (en) 2021-07-02 2021-07-02 Risk cause identification method and device and storage medium

Publications (1)

Publication Number Publication Date
CN113393155A true CN113393155A (en) 2021-09-14

Family

ID=77624952

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110751947.6A Pending CN113393155A (en) 2021-07-02 2021-07-02 Risk cause identification method and device and storage medium

Country Status (1)

Country Link
CN (1) CN113393155A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114155086A (en) * 2021-11-22 2022-03-08 北京字节跳动网络技术有限公司 Data processing method and device

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180285795A1 (en) * 2017-03-30 2018-10-04 Wipro Limited Method and device for risk analysis using risk relationship mapping in supply chain networks
CN110363449A (en) * 2019-07-25 2019-10-22 中国工商银行股份有限公司 A kind of Risk Identification Method, apparatus and system
CN112463981A (en) * 2020-11-26 2021-03-09 福建正孚软件有限公司 Enterprise internal operation management risk identification and extraction method and system based on deep learning

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20180285795A1 (en) * 2017-03-30 2018-10-04 Wipro Limited Method and device for risk analysis using risk relationship mapping in supply chain networks
CN110363449A (en) * 2019-07-25 2019-10-22 中国工商银行股份有限公司 A kind of Risk Identification Method, apparatus and system
CN112463981A (en) * 2020-11-26 2021-03-09 福建正孚软件有限公司 Enterprise internal operation management risk identification and extraction method and system based on deep learning

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114155086A (en) * 2021-11-22 2022-03-08 北京字节跳动网络技术有限公司 Data processing method and device
CN114155086B (en) * 2021-11-22 2024-02-02 抖音视界有限公司 Data processing method and device

Similar Documents

Publication Publication Date Title
Unnikrishnan et al. Toward objective evaluation of image segmentation algorithms
Liu et al. Effective clustering and boundary detection algorithm based on Delaunay triangulation
CN110570111A (en) Enterprise risk prediction method, model training method, device and equipment
Cruz et al. Community detection and visualization in social networks: Integrating structural and semantic information
Fushing et al. Time, temperature, and data cloud geometry
CN101901251B (en) Method for analyzing and recognizing complex network cluster structure based on markov process metastability
CN102622753A (en) Semi-supervised spectral clustering synthetic aperture radar (SAR) image segmentation method based on density reachable measure
Li et al. Localizing and quantifying infrastructure damage using class activation mapping approaches
CN112818162A (en) Image retrieval method, image retrieval device, storage medium and electronic equipment
Ip et al. A 3D object classifier for discriminating manufacturing processes
CN113343920A (en) Method and device for classifying face recognition photos, electronic equipment and storage medium
CN117710666A (en) Dividing method and device for slope units
CN113393155A (en) Risk cause identification method and device and storage medium
Inyang et al. Unsupervised Characterization and Visualization of Students' Academic Performance Features.
CN113553326A (en) Spreadsheet data processing method, device, computer equipment and storage medium
CN112597399A (en) Graph data processing method and device, computer equipment and storage medium
CN112488236B (en) Integrated unsupervised student behavior clustering method
CN114491841A (en) Machining feature recognition method based on NX secondary development and graph neural network
CN113518010A (en) Link prediction method, device and storage medium
Zhang et al. Photo aesthetic quality assessment via label distribution learning
Cabanes et al. On the use of Wasserstein metric in topological clustering of distributional data
CN117055818B (en) Client information storage management method and system based on block chain
CN109472370B (en) Method and device for classifying maintenance plants
Liu et al. [Retracted] An Accurate Method of Determining Attribute Weights in Distance‐Based Classification Algorithms
CN113449921A (en) Risk cause identification method and device and storage medium

Legal Events

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