CN117575328A - Industrial chain risk assessment method and device based on graph attention neural network - Google Patents
Industrial chain risk assessment method and device based on graph attention neural network Download PDFInfo
- Publication number
- CN117575328A CN117575328A CN202311699916.6A CN202311699916A CN117575328A CN 117575328 A CN117575328 A CN 117575328A CN 202311699916 A CN202311699916 A CN 202311699916A CN 117575328 A CN117575328 A CN 117575328A
- Authority
- CN
- China
- Prior art keywords
- node
- risk
- industrial chain
- value
- nodes
- 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
Links
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 67
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012502 risk assessment Methods 0.000 title claims abstract description 33
- 238000013210 evaluation model Methods 0.000 claims abstract description 39
- 238000011144 upstream manufacturing Methods 0.000 claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 29
- 239000013598 vector Substances 0.000 claims description 43
- 238000004364 calculation method Methods 0.000 claims description 28
- 230000002776 aggregation Effects 0.000 claims description 21
- 238000004220 aggregation Methods 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 11
- 238000010276 construction Methods 0.000 claims description 10
- 230000004931 aggregating effect Effects 0.000 claims description 9
- 238000010606 normalization Methods 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 6
- 238000010586 diagram Methods 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 6
- 230000007246 mechanism Effects 0.000 claims description 6
- 238000007526 fusion splicing Methods 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 2
- 238000011156 evaluation Methods 0.000 description 6
- 238000007726 management method Methods 0.000 description 6
- 238000011160 research Methods 0.000 description 5
- 238000012544 monitoring process Methods 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- 238000012827 research and development Methods 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000006116 polymerization reaction Methods 0.000 description 3
- 239000002994 raw material Substances 0.000 description 3
- WHXSMMKQMYFTQS-UHFFFAOYSA-N Lithium Chemical compound [Li] WHXSMMKQMYFTQS-UHFFFAOYSA-N 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 229910052744 lithium Inorganic materials 0.000 description 2
- 238000005065 mining Methods 0.000 description 2
- 230000001133 acceleration Effects 0.000 description 1
- 239000010405 anode material Substances 0.000 description 1
- 238000009412 basement excavation Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000010406 cathode material Substances 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000002860 competitive effect Effects 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 239000003792 electrolyte Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 229910052751 metal Inorganic materials 0.000 description 1
- 239000002184 metal Substances 0.000 description 1
- 150000002739 metals Chemical class 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000004451 qualitative analysis Methods 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 230000001502 supplementing effect Effects 0.000 description 1
- 238000013068 supply chain management Methods 0.000 description 1
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
- 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
- G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
-
- 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
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Entrepreneurship & Innovation (AREA)
- Software Systems (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Evolutionary Computation (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Game Theory and Decision Science (AREA)
- Educational Administration (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application relates to the technical field of industrial chain wind control management, discloses an industrial chain risk assessment method and device based on a graph attention neural network, and aims to solve the problem that an existing industrial chain risk assessment mode is poor in accuracy, and the scheme mainly comprises the following steps: constructing a network topology structure at the upstream, the middle and the downstream of the industrial chain according to the data information of the industrial chain to be evaluated; constructing a graph attention neural network evaluation model, performing model training, and calculating a node risk value and an attention relationship coefficient of the industrial chain network to be evaluated by using the trained graph attention neural network evaluation model; and calculating an aggregate risk value of the nodes at the same layer according to the node risk values and the attention relation coefficient, calculating a risk value of the nodes at a cross layer according to the aggregate risk value of the nodes at the same layer to obtain an upper, middle and lower risk value, and calculating according to the upper, middle and lower risk values to obtain an overall risk value of the industrial chain to be evaluated. The method and the device improve the accuracy of risk assessment of the industrial chain, and are particularly suitable for new energy automobile industrial chains.
Description
Technical Field
The application relates to the technical field of industrial chain wind control management, in particular to an industrial chain risk assessment method and device based on a graph attention neural network.
Background
The industrial chain safety is the basis for ensuring the economic high-quality development, the industrial chain is scientifically evaluated, and the national development and safety of an industrial chain safety risk assessment and early warning monitoring system are constructed. The risk of the industrial chain has dependency, the interdependence relationship exists between different links and participants in the industrial chain, and the risk problem of one link can be transmitted to other links, and chain reaction is generated on the whole industrial chain. Industry chain risk assessment can help identify and assess potential risk factors and provide basis for risk management and control for various links and participants.
The related research of the current industrial chain security risk assessment mainly shows the following characteristics: first, there are relatively few studies on risk of the industrial chain that are not deep and focus on concept discrimination and qualitative analysis, and at the same time, there are relatively few quantitative assessment methods for risk of the industrial chain, and more, from the viewpoint of supply chain management, the research methods are not applicable to industrial chain research. See literature: li Zheng, wang Saini Innovative science and technology, 2022, of construction of systems for risk assessment and early warning monitoring of industrial chains, which is merely a qualitative description of the necessity, ideas and key links and policy advice for constructing systems for risk assessment and early warning monitoring of industrial chains. See literature: dai Bin, yang Qian, important industry chain supply chain safety risk feature recognition and management mechanism design [ J ]. Lopa-nations management comment, 46 (1): 1-10, which only qualitatively presents the secondary index system defining the important industry chain supply chain, the features, types and management mechanisms of important industry chain supply chain safety risk. Secondly, quantitative analysis for some industries is mainly simple statistics on current data, and cannot effectively reflect the risk of an industrial chain. The method is characterized in that the current research is mainly carried out on the aspects of competitiveness, control force, dependency and the like of an industrial chain, and an evaluation path of the safety risk of the industrial chain is mostly provided from the aspects of connotation characteristics and macroscopic situation of the industrial chain, and in the construction and evaluation of an index system, a balance integration card, text mining, AHC and other methods are used for establishing the index system for evaluation in some researches, for example, an energy power industrial chain risk early warning method, system, equipment and medium disclosed in application publication No. CN114493078A are used for evaluation. Since expert opinion, experience and expertise are relied on, this method is susceptible to subjective bias and personal experience of the evaluator, resulting in uncertainty and subjectivity of the evaluation result.
The application publication number CN115860474A discloses an industrial chain safety early warning method and system based on ensemble learning, and literature: pan Meng, zhang Jiantong, chen Xiaodong, study on risk assessment of automotive supply chain based on neural networks [ J ]. Shanghai management science, 2019, describes methods for applying machine learning and deep learning, respectively, to risk assessment of industrial chain. The method still needs to rely on the assistance of expert scoring, meanwhile, lacks consideration of risk factors in the industrial chain and risk conduction mechanisms among enterprise nodes, and particularly neglects capture of information of network relation structures at the upstream and downstream in the industrial chain, lacks excavation of information of complex network relation of the industrial chain, cannot accurately analyze propagation paths and influence degrees of risks in the industrial chain, is difficult to comprehensively identify potential risk points and loopholes, and finally can lead to deviation and inaccuracy of an evaluation result. The limitations described above make it difficult for existing methods to accurately and early identify industry chain risks.
Disclosure of Invention
The application aims to solve the problem of poor accuracy of the existing industrial chain risk assessment mode and provides an industrial chain risk assessment method and device based on a graph attention neural network.
The technical scheme adopted for solving the technical problems is as follows:
in a first aspect, the present application provides a method for evaluating risk of an industrial chain based on a graph attention neural network, the method comprising:
constructing a network topology structure of the upstream, the middle and the downstream of the industrial chain according to the data information of the industrial chain to be evaluated, and obtaining the industrial chain network to be evaluated;
constructing a graph attention neural network evaluation model, performing model training, and calculating a node risk value and an attention relationship coefficient of the industrial chain network to be evaluated by using the trained graph attention neural network evaluation model;
and calculating an aggregate risk value of the nodes at the same layer according to the node risk value and the attention relation coefficient, calculating a risk value of a cross-layer node according to the aggregate risk value of the nodes at the same layer to obtain an upper, middle and lower risk value, and calculating according to the upper, middle and lower risk values to obtain an overall risk value of the industrial chain to be evaluated.
Further, the data information of the industrial chain to be evaluated comprises attribute characteristics of enterprises on the industrial chain and relationship information among the enterprises in the industrial chain;
the attribute features comprise enterprise technical innovation capability index data, enterprise profit capability index data, enterprise market competitiveness index data, enterprise technical chain index data, enterprise capital chain index data, enterprise market chain index data, enterprise natural political environment index data, policy environment index data and talent guarantee condition index data;
the relationship information includes supply relationships, investment relationships, collaboration relationships, and competition relationships present in the industry chain upstream, midstream, and downstream sub-chains.
Further, the construction method of the network topology structure comprises the following steps:
and counting all enterprises and association relations among the enterprises on the industrial chain to be evaluated, carrying out hierarchical and class division on all the enterprises, determining upstream, middle and downstream enterprises based on the hierarchical and class division results, taking all the enterprises as network nodes, taking the association relations among the enterprises as edges, and constructing a topological network corresponding to the upstream, middle and downstream of the industrial chain based on the network nodes and the edges.
Further, the drawing attention neural network evaluation model comprises an input layer, a drawing attention layer, a dense full-connection layer and an output layer, and the construction process of the drawing attention neural network evaluation model comprises the following steps:
calculating a central feature vector and a node attribute feature vector of an industrial chain network node, and fusing and splicing the two feature vectors into a central node by using a graph attention mechanism to be embedded into an input layer serving as the graph attention neural network evaluation model;
embedding the fused nodes into a feature input diagram attention layer to obtain output features of the nodes;
inputting the graph annotation force layer output characteristics of the nodes into the dense full-connection layer to obtain the dense full-connection layer output characteristics of the nodes;
and inputting the output characteristics of the dense full-connection layer of the nodes into an output layer to obtain the risk prediction value of the nodes.
Further, the feature vector fusion splicing process includes:
firstly, calculating the central feature vector of a node network nodeAnd node Attribute feature vector->The expression is as follows:
wherein,representing a node network node centre feature vector +.>Weight of->Representing node Attribute feature vector +.>Sigma represents the LeakyReLU activation function, W c And W is X Representing a learnable transformation matrix->Representing a network node centre feature vector->Attention weight vector of corresponding discipline,/->Representing node Attribute feature vector +.>Attention weight vector of corresponding discipline, [ ·]A concatenation operation representing vectors;
the weights were then normalized by a Softmax function, expressed as follows:
finally, the node embedding after fusion is obtainedThe expression is as follows:
wherein,and->And representing the weight normalization result.
Further, the fused node is embedded into a feature input graph attention layer, and the calculation process comprises the following steps:
vector embedding of the attention layer node is calculated, and the expression is as follows: z i ＝Wh i Wherein z is i Vector embedding representing node i, W representing a learnable weight matrix, h i Representing embedded features of node i;
the attention value of the node i and the first-order neighbor node j is calculated, and the expression is as follows: e, e ij ＝σ(α T [z i ||z j ]) Wherein σ represents the LeakyReLU activation function, α T Representing a matrix of learnable parameters, z j Vector embedding representing node j;
normalizing the attention value, and the expression is: alpha ij ＝softmax(e ij )；
Weighting and summing the node i to obtain the final output characteristic of the node iThe computational expression is: /> Wherein N (i) represents all first-order neighbor nodes of the node i, alpha ij Representing the normalization result of the attention value;
the computing expression of the output characteristics of the intensive full-connection layer is as follows:wherein s is i Representing dense full-connection layer output characteristics of node i, x L ＝H L ([x 0 ,x 1 ,...,x L-1 ]) Represents a dense full-connection layer of the L-th layer, H L Is a cascading function;
the calculation expression of the risk prediction value of the node is as follows:wherein (1)>Representing the risk prediction value of node i, +.>Representing the learnable weights.
Further, the training method of the graph meaning neural network evaluation model comprises the following steps:
selecting training nodes in the industrial chain network to be evaluated, calculating a risk prediction value of each node in the training nodes by using the graph-annotation-force neural network evaluation model, calculating a loss function, calculating the loss value according to the loss function, and updating the learnable parameters of the graph-annotation-force neural network evaluation model through back propagation;
repeating the steps until the loss value converges, and finishing training.
Further, calculating an aggregate risk value of the nodes at the same layer according to the node risk value and the attention relation coefficient, wherein the aggregate risk value specifically comprises:
respectively calculating an aggregation risk value represented by each relation information, and calculating the aggregation risk value of the nodes at the same layer according to the aggregation risk value represented by each relation information;
the calculation expression of the risk value aggregate value expressed by the competition relationship is as follows:wherein (1)>Aggregating risk values for competition relationships of the ith subdivision industry of the first layer, M (i) represents all first-order neighbor nodes under the competition relationship of the node i, M represents the number of the first-order neighbor nodes under the competition relationship of the node i, and +.>Representing the risk prediction value, alpha, of node i ij Representing the normalization result of the attention value;
the risk value aggregate value calculation expression expressed by the investment relation is as follows:wherein (1)>Aggregating risk values for investment relations of the ith subdivision industry of the first layer, wherein N (i) represents all first-order neighbor nodes under the investment relation of the node i, and M represents the number of the first-order neighbor nodes under the investment relation of the node i;
the risk value aggregate value calculation expression of the supply relation expression is:wherein (1)>Aggregating risk values for supply relationships for a layer i, subdivision industry, P (i) representing node i supply relationshipsAll first-order neighbor nodes under the system, P represents the number of first-order neighbor nodes under the node i supply relation;
aggregation risk value of the same-layer nodeThe calculated expression of (2) is: />
Further, the calculation expression of the upstream, middle and downstream risk values is as follows:
wherein R represents an upstream, midstream or downstream risk value,representing the aggregation risk value of the corresponding same-layer node;
the overall risk value R of the industrial chain to be evaluated i The calculated expression of (2) is:
wherein R is 1 Represents an upstream risk value, R 2 Represents midstream risk value, R 3 Representing a risk value downstream.
In a second aspect, the present application provides an apparatus for risk assessment of an industrial chain based on a graph attention neural network, the apparatus comprising:
the construction module is used for constructing a network topology structure of the upstream, the middle and the downstream of the industrial chain according to the data information of the industrial chain to be evaluated to obtain an industrial chain network to be evaluated;
the prediction module is used for constructing a graph attention neural network evaluation model and carrying out model training, and calculating a node risk value and an attention relationship coefficient of the industrial chain network to be evaluated by using the trained graph attention neural network evaluation model;
the calculation module is used for calculating the aggregate risk value of the nodes at the same layer according to the node risk value and the attention relation coefficient, calculating the risk value of the nodes at the cross layer according to the aggregate risk value of the nodes at the same layer to obtain an upper, middle and lower risk value, and calculating the overall risk value of the industrial chain to be evaluated according to the upper, middle and lower risk values.
The beneficial effects of this application are: according to the industrial chain risk assessment method and device based on the graph attention neural network, the originally isolated nodes are connected into the network subgraph by excavating various complex relation network structure information existing in each link of the upper, middle and lower links of the industrial chain, the central node risk value is predicted by utilizing the network subgraph structure characteristics and the node attribute characteristics which are fused with various relations, the risk value of each link of the industrial chain and the overall risk value are obtained according to the aggregation method, and various topological relations of the industrial chain are fully utilized to identify the industrial chain risk, so that the problem that the effect is poor when only node attribute data are used in the industrial chain risk assessment field is solved, the high efficiency and risk assessment precision are achieved, and a decision maker can be helped to better manage and control the safety risk in the industrial chain.
Drawings
Fig. 1 is a schematic flow chart of an industrial chain risk assessment method based on a graph attention neural network according to an embodiment of the present application;
FIG. 2 is an exemplary diagram of an industrial chain network structure to be evaluated according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of a schematic neural network model according to an embodiment of the present application;
fig. 4 is a schematic structural diagram of an industrial chain risk assessment device based on a graph attention neural network according to an embodiment of the present application;
fig. 5 is a schematic structural diagram of another industrial chain risk assessment device based on a graph attention neural network according to an embodiment of the present application.
Detailed Description
In order to enable those skilled in the art to better understand the present application, the following description will make clear and complete descriptions of the technical solutions in the embodiments of the present application with reference to the accompanying drawings in the embodiments of the present application.
In some of the flows described in the specification of this application and the foregoing figures, a plurality of operations are included that occur in a particular order, but it should be understood that the operations may be performed in other than the order in which they occur or in parallel, the order numbers of the operations, such as 101, 102, etc., are merely used to distinguish between the various operations, and the order numbers themselves do not represent any order of execution.
The technical scheme of the embodiment of the application is suitable for application scenes needing risk assessment of the industrial chain, such as a new energy automobile industrial chain.
Because the current industrial chain risk assessment application only uses node attribute data, capturing of the information of the relationship structures of the upstream and downstream networks in the industrial chain is ignored, mining of the complex relationship information of the industrial chain network is lacking, the propagation path and the influence degree of risks in the industrial chain cannot be accurately analyzed, potential risk points and loopholes are difficult to comprehensively identify, and finally deviation and inaccuracy of an assessment result can be caused.
Based on the above, the technical scheme of the application is provided, and in the embodiment of the application, a network topology structure of the upstream, the middle and the downstream of the industrial chain is constructed according to the data information of the industrial chain to be evaluated, so as to obtain the industrial chain network to be evaluated; constructing a graph attention neural network evaluation model, performing model training, and calculating a node risk value and an attention relationship coefficient of the industrial chain network to be evaluated by using the trained graph attention neural network evaluation model; and calculating an aggregate risk value of the nodes at the same layer according to the node risk value and the attention relation coefficient, calculating a risk value of a cross-layer node according to the aggregate risk value of the nodes at the same layer to obtain an upper, middle and lower risk value, and calculating according to the upper, middle and lower risk values to obtain an overall risk value of the industrial chain to be evaluated. According to the embodiment of the application, based on the topological relation of the industrial chain network, the industrial chain node risk value is calculated through the graph attention neural network, further, the risk values of all links of the industrial chain and the whole industrial chain are obtained through the same-layer and cross-layer risk aggregation method, various topological relations of the industrial chain are fully utilized to identify the industrial chain risk, and the problem that the effect is poor when the conventional industrial chain risk assessment application only uses node attribute information is solved.
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application.
Referring to fig. 1, an industrial chain risk assessment method based on a graph attention neural network provided in an embodiment of the present application includes the following steps:
step 101, constructing a network topology structure of the upstream, the middle and the downstream of the industrial chain according to data information of the industrial chain to be evaluated, and obtaining the industrial chain network to be evaluated;
in this embodiment of the present application, the data information of the industrial chain to be evaluated includes attribute features of enterprises on the industrial chain and relationship information between enterprises in the industrial chain.
The attribute features include enterprise technical innovation capability index data, enterprise profitability index data, enterprise market competitiveness index data, enterprise technical chain index data, enterprise capital chain index data, enterprise market chain index data, enterprise natural political environment index data, policy environment index data and talent guarantee condition index data.
The enterprise technical innovation capability index data is used for measuring performances of enterprises in technical research and development and innovation capability, and corresponding subdivision indexes such as research and development investment acceleration, research and development investment occupation and sales income proportion, capital research and development investment proportion, technician proportion, staff proportion of the family and the schools, patent application quantity, patent intensity, invention patent proportion and the like are used for measuring the performance of the enterprises; the enterprise profitability index data reflects the profitability level and financial status of the enterprise, corresponding subdivision indexes such as net profit growth rate, cost profit rate, net asset profit rate, and the like; the enterprise market competitiveness index data evaluates the competitive position and capacity of an enterprise on the market, and corresponds to the market share of a subdivision index domestic market, the international market share, the sales profit margin, the market growth rate, the free brand value, the financing scale and the like; the enterprise technical chain index data describes the technical control force of an enterprise on an industrial chain, and corresponds to subdivision indexes such as the national concentration of suppliers, the concentration of supply chains, the ratio of foreign patent applications, the control rate of foreign core technical patents, the number of times that patents are cited by foreign patents, the number of articles published by foreign journals and the like; the enterprise capital chain index data describes the capital control force of the enterprise on the industrial chain, and corresponds to subdivision indexes such as overseas financing scale duty ratio and overseas asset share right control rate; the enterprise market chain index data describes the market control capacity of enterprises on an industrial chain, and corresponds to subdivision indexes such as the occupancy of foreign enterprises on domestic markets, overseas business income proportion of marketing companies and the like; the enterprise natural politics environment index data is used for examining whether extreme conditions such as major natural disasters, epidemic diseases, politics fluctuation, war and the like occur in key input product supply places required by key enterprises; the policy environment index data evaluates the supporting strength of government policies on enterprises, and corresponds to subdivision indexes such as financial fund supporting amount, average level of tariffs collected by main export products, cross country purchase or investment number and the like; talent guarantee condition index data describe international talent introduction quantity and speed increasing of industries of enterprises, chinese personnel duty ratio of key technical inventors of industries of enterprises, chinese personnel duty ratio of leading edge technical literature publishers, and the like.
The relationship information includes supply relationships, investment relationships, collaboration relationships, and competition relationships present in the industry chain upstream, midstream, and downstream sub-chains.
Referring to fig. 2, an investment relationship (upper left) formed by an enterprise a by a equity investment to an enterprise C or an enterprise B by an enterprise Cd bond investment, a competing relationship (upper right) formed by an enterprise B and an enterprise C by being homogeneous or similar to a product or service being generated or provided by belonging to a subdivision industry, a supply relationship (lower left) formed by an upstream enterprise a providing a product or service to a downstream enterprise B, and a cooperating relationship (lower right) formed by an enterprise a and an enterprise B providing a product 1 and a service 2, respectively, can complete a certain product or service in an enterprise C.
In this embodiment of the present application, the process for constructing the network topology includes: and counting all enterprises and association relations among the enterprises on the industrial chain to be evaluated, carrying out hierarchical and class division on all the enterprises, determining upstream, middle and downstream enterprises based on the hierarchical and class division results, taking all the enterprises as network nodes, taking the association relations among the enterprises as edges, and obtaining a topology network corresponding to the upstream, middle and downstream of the industrial chain to be evaluated based on the network nodes and the edges.
Step 102, constructing a graph attention neural network evaluation model, performing model training, and calculating a node risk value and an attention relationship coefficient of an industrial chain network to be evaluated by using the trained graph attention neural network evaluation model;
referring to fig. 3, the schematic neural network evaluation model in the embodiment of the present application includes an input layer, a schematic attention layer, a dense full connection layer and an output layer, and the construction process includes the following steps:
step 1021, calculating a central feature vector (such as a local attribute of a node, a global attribute of the node, a position attribute of the node, node embedding and the like) of the node of the industrial chain network and a node attribute feature vector, and fusing and splicing the two feature vectors into a central node embedded as an input layer of the graph attention neural network evaluation model by utilizing a graph attention mechanism;
in the embodiment of the application, the feature vector fusion splicing process includes the following steps:
step 10211, firstly calculating the node network node center feature vectorAnd node Attribute feature vector->The expression is as follows:
wherein,representing a node network node centre feature vector +.>Weight of->Representing node Attribute feature vector +.>Sigma represents the LeakyReLU activation function, W c And W is X Representing a learnable transformation matrix->Representing a network node centre feature vector->Attention weight vector of corresponding discipline,/->Representing node Attribute feature vector +.>Attention weight vector of corresponding discipline, [ ·]A concatenation operation representing vectors;
step 10212, normalizing the weights by Softmax function, and the expression is as follows:
step 10213, finally obtaining the node embedding after fusionThe expression is as follows:
wherein,and->And representing the weight normalization result.
Step 1022, embedding the fused nodes into a feature input graph attention layer to obtain output features of the nodes;
in the embodiment of the application, the fused node is embedded into a feature input graph attention layer, and the calculation process comprises the following steps:
step 10221, calculating vector embedding of the attention layer node, wherein the expression is: z i ＝Wh i Wherein z is i Vector embedding representing node i, W representing a learnable weight matrix, h i Representing embedded features of node i;
step 10222, calculating the attention value of the node i and the first-order neighbor node j, wherein the expression is as follows: e, e ij ＝σ(α T [z i ||z j ]) Wherein σ represents the LeakyReLU activation function, α T Representing a matrix of learnable parameters, z j Vector embedding representing node j;
step 10223, normalizing the attention value, and the expression is: alpha ij ＝softmax(e ij )；
Step 10224, weighting and summing the node i to obtain the final output characteristic of the node iThe computational expression is: /> Wherein N (i) represents all first-order neighbor nodes of the node i, alpha ij The attention value normalization result is shown.
Step 1023, inputting the graph annotation force layer output characteristics of the nodes into the dense full-connection layer to obtain the dense full-connection layer output characteristics of the nodes;
the computing expression of the output characteristics of the intensive full-connection layer is as follows:
wherein d i Representing dense full-connection layer output characteristics of node i, x L ＝H L ([x 0 ,x 1 ,...,x L-1 ]) Represents a dense full-connection layer of the L-th layer, H L As a cascading function.
And step 1024, inputting the output characteristics of the dense full-connection layer of the nodes into an output layer to obtain the risk prediction value of the nodes.
The calculation expression of the risk prediction value of the node is as follows:
wherein,representing the risk prediction value of node i, +.>Representing the learnable weights.
After the graph attention neural network evaluation model is constructed and obtained, training is needed, and in the embodiment of the application, the training process comprises the following steps:
selecting training nodes in the industrial chain network to be evaluated, calculating risk prediction values of each node in the training nodes by using the graph-annotation force neural network evaluation model, and calculating a loss functionCalculating a loss value from a loss function, in particular, < >>The risk prediction value of the node is y, which is the true risk value of the node; updating the learnable parameters of the graph annotation meaning neural network evaluation model through back propagation; repeating the steps until the loss value converges, and finishing training.
After training is completed, the node risk value and the attention relationship coefficient of the industrial chain network to be evaluated can be calculated by using the trained graph annotation intention network evaluation model. Specifically, a trained graph attention neural network evaluation model is used for predicting the node risk value, the predicted value can be used for supplementing the risk value of the node with the missing industrial chain network risk value on one hand, and on the other hand, can be used for correcting the existing node risk value, so that the method provided by the application is also suitable for semi-supervised learning, and the risk values of all nodes can be obtained through training and prediction only by knowing the risk values of part of the nodes. And after the nodes are predicted in sequence, all risk values of the industrial chain nodes and corresponding attention relation coefficients among the nodes can be obtained.
And 103, calculating an aggregate risk value of the nodes at the same layer according to the node risk value and the attention relation coefficient, calculating a risk value of the nodes at a cross layer according to the aggregate risk value of the nodes at the same layer to obtain an upper, middle and lower risk value, and calculating according to the upper, middle and lower risk values to obtain an overall risk value of the industrial chain to be evaluated.
The peer nodes represent all enterprises in the same subdivision industry category in the industry chain network, for example, all enterprises in the lithium mine industry at the upstream of the new energy automobile industry chain belong to the peer nodes, and the enterprises have a competition relationship due to the product homogeneity or the similarity. Also, in the same-layer node, there are investment relations between enterprises in the industry chain network due to the forms of equity investment, liability financing, etc., and supply relations formed by one enterprise (such as a manufacturer or a supplier) in the industry chain network providing products, raw materials, or services to another enterprise. Then, the relationship level risk aggregation formula is a weighted average of the node risk values formed by the corresponding relationships and the corresponding attention relationship coefficients.
Specifically, the aggregate risk value represented by each relationship information may be calculated separately, and the aggregate risk value of the nodes at the same level may be calculated from the aggregate risk value represented by each relationship information.
In the embodiment of the present application, the calculation expression of the risk value aggregate value represented by the competition relationship is:
wherein,aggregating risk values for competition relationships of the ith subdivision industry of the first layer, M (i) represents all first-order neighbor nodes under the competition relationship of the node i, M represents the number of the first-order neighbor nodes under the competition relationship of the node i, and +.>Representing the risk prediction value, alpha, of node i ij Representing the normalization result of the attention value;
the risk value aggregate value calculation expression expressed by the investment relation is as follows:
wherein,aggregating risk values for investment relations of the ith subdivision industry of the first layer, wherein N (i) represents all first-order neighbor nodes under the investment relation of the node i, and M represents the number of the first-order neighbor nodes under the investment relation of the node i;
the risk value aggregate value calculation expression of the supply relation expression is:
wherein,aggregating risk values for the supply relations of the ith subdivision industry of the first layer, wherein P (i) represents all first-order neighbor nodes under the supply relation of the node i, and P represents the number of the first-order neighbor nodes under the supply relation of the node i;
aggregation risk value of the peer nodeCalculation of (2)The expression is:
it should be noted that, in the embodiment of the present application, only three polymerization modes of relationships are demonstrated, and more polymerization of relationships can be performed by referring to the polymerization method.
It can be understood that the cross-layer node represents different industries and sub-industry categories in the industrial chain topology network, for example, upstream raw materials of the industrial chain of the new energy automobile comprise lithium ores, electrolyte, anode materials, cathode materials, diaphragms and other metals, the sub-industries under the upstream raw materials form a network node, namely the cross-layer node, and the risk value of the cross-layer node comprises an upstream risk value, a middle risk value and a downstream risk value.
In this embodiment of the present application, the calculation expression of the upstream, middle and downstream risk values is:
wherein R represents an upstream, midstream or downstream risk value,representing the aggregation risk value of the corresponding same-layer node;
the overall risk value R of the industrial chain to be evaluated i The calculated expression of (2) is:
wherein R is 1 Represents an upstream risk value, R 2 Represents midstream risk value, R 3 Representing a risk value downstream.
In summary, according to the industrial chain risk assessment method based on the graph attention neural network provided by the embodiment of the application, the industrial chain node risk value is calculated through the graph attention neural network based on the industrial chain network topological relation, and further, the risk values of all links of the industrial chain and the whole industrial chain are obtained through the same-layer and cross-layer risk aggregation method, so that various topological relations of the industrial chain are fully utilized to identify the industrial chain risk, the problem that the effect is poor when the conventional industrial chain risk assessment application only uses node attribute information is solved, and a decision maker can be helped to manage and control the safety risk in the industrial chain better.
Referring to fig. 4, based on the above technical solution, an embodiment of the present application further provides an industrial chain risk assessment device based on a graph attention neural network, where the device includes:
the construction module is used for constructing a network topology structure of the upstream, the middle and the downstream of the industrial chain according to the data information of the industrial chain to be evaluated to obtain an industrial chain network to be evaluated;
the prediction module is used for constructing a graph attention neural network evaluation model and carrying out model training, and calculating a node risk value and an attention relationship coefficient of the industrial chain network to be evaluated by using the trained graph attention neural network evaluation model;
the calculation module is used for calculating the aggregate risk value of the nodes at the same layer according to the node risk value and the attention relation coefficient, calculating the risk value of the nodes at the cross layer according to the aggregate risk value of the nodes at the same layer to obtain an upper, middle and lower risk value, and calculating the overall risk value of the industrial chain to be evaluated according to the upper, middle and lower risk values.
Referring to fig. 5, in the embodiment of the present application, the construction module is responsible for data collection and processing, mainly collecting network node attribute feature data related to an industrial chain to be evaluated and relationship information data of network nodes, and performing data preprocessing operation on the related data, so that the data meets the requirement of training a graph attention neural network model.
The prediction module comprises a graph attention neural network model training unit, a node risk value calculation unit and a graph attention force weight coefficient calculation unit. The drawing meaning neural network model training module unit is used for selecting training nodes in the industrial chain network to be evaluated, calculating risk prediction values of each node in the training nodes by using the drawing meaning neural network evaluation model, calculating a loss function, and calculating a loss value according to the loss function; updating the learnable parameters of the graph annotation neural network evaluation model through back propagation; repeating the steps until the loss value converges to complete the model training. And the node risk value calculation unit is used for predicting the risk value of the industrial chain network node to be predicted by using the trained graph attention neural network evaluation model to obtain all risk values of the industrial chain node. The graph attention weight coefficient calculation unit is used for predicting the risk value of the industrial chain network node by using the trained graph attention neural network evaluation model to obtain the attention relation coefficient between the nodes corresponding to the industrial chain link point.
The computing module comprises a same-layer risk value aggregation unit, a cross-layer risk value computing unit and an industrial chain risk value computing unit. The same-layer risk value aggregation unit is used for calculating an aggregation risk value of the same-layer nodes according to the node risk values and the attention relationship coefficients. The same-layer aggregation risk value is essentially the risk value of the nodes of the industry chain network major industry and the subdivision industry, and all main links of the industry chain can be monitored in real time, once the warning state is generated, namely the fluctuation of the risk value is abnormal, the problem can be found out in time, namely which link has the problem, so that a decision maker can make an emergency plan in time. The cross-layer risk value calculation unit is used for calculating the risk value of the cross-layer node according to the aggregate risk value of the same-layer node to obtain the upstream, middle and downstream risk values. The cross-layer aggregation risk value is essentially the overall risk value of the upstream, midstream and downstream of the industrial chain, and by the value, the overall risk of the upstream, midstream and downstream of the industrial chain can be monitored in real time, and once an alert state occurs, namely the fluctuation of the risk value is abnormal, the problem of the downstream link of the industrial chain can be timely found out, so that a decision maker can make an emergency plan timely. The industrial chain risk value calculation unit is used for calculating according to the upstream, middle and downstream aggregate risk values to finally obtain the overall risk value of the industrial chain to be evaluated. The whole risk value of the industrial chain is provided, real-time monitoring can be carried out on the whole risk of the industrial chain, once an alert state occurs, namely the fluctuation of the risk value is abnormal, a decision maker can search the reason through tracing, and the decision maker can conveniently and timely make an emergency plan.
It can be understood that, since the apparatus for risk assessment of an industrial chain based on a graph attention neural network according to the embodiments of the present application is an apparatus for implementing the method for risk assessment of an industrial chain based on a graph attention neural network according to the embodiments, for the apparatus disclosed in the embodiments, the description is simpler, and the relevant points will only be referred to the part of the description of the method, and will not be repeated here.
The apparatus embodiments described above are merely illustrative, wherein the elements illustrated as separate elements may or may not be physically separate, and the elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.
From the above description of the embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented by means of software plus necessary general hardware platforms, or of course may be implemented by means of hardware. Based on this understanding, the foregoing technical solution may be embodied essentially or in a part contributing to the prior art in the form of a software product, which may be stored in a computer readable storage medium, such as ROM/RAM, a magnetic disk, an optical disk, etc., including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the method described in the respective embodiments or some parts of the embodiments.
Claims (10)
1. An industrial chain risk assessment method based on a graph attention neural network, which is characterized by comprising the following steps:
constructing a network topology structure of the upstream, the middle and the downstream of the industrial chain according to the data information of the industrial chain to be evaluated, and obtaining the industrial chain network to be evaluated;
constructing a graph attention neural network evaluation model, performing model training, and calculating a node risk value and an attention relationship coefficient of the industrial chain network to be evaluated by using the trained graph attention neural network evaluation model;
and calculating an aggregate risk value of the nodes at the same layer according to the node risk value and the attention relation coefficient, calculating a risk value of a cross-layer node according to the aggregate risk value of the nodes at the same layer to obtain an upper, middle and lower risk value, and calculating according to the upper, middle and lower risk values to obtain an overall risk value of the industrial chain to be evaluated.
2. The method for evaluating the risk of an industrial chain based on a graph attention neural network according to claim 1, wherein the data information of the industrial chain to be evaluated comprises attribute characteristics of enterprises on the industrial chain and relationship information among the enterprises in the industrial chain;
the attribute features comprise enterprise technical innovation capability index data, enterprise profit capability index data, enterprise market competitiveness index data, enterprise technical chain index data, enterprise capital chain index data, enterprise market chain index data, enterprise natural political environment index data, policy environment index data and talent guarantee condition index data;
the relationship information includes supply relationships, investment relationships, collaboration relationships, and competition relationships present in the industry chain upstream, midstream, and downstream sub-chains.
3. The method for evaluating risk of an industrial chain based on a graph attention neural network according to claim 1, wherein the method for constructing a network topology comprises:
and counting all enterprises and association relations among the enterprises on the industrial chain to be evaluated, carrying out hierarchical and class division on all the enterprises, determining upstream, middle and downstream enterprises based on the hierarchical and class division results, taking all the enterprises as network nodes, taking the association relations among the enterprises as edges, and constructing a topological network corresponding to the upstream, middle and downstream of the industrial chain based on the network nodes and the edges.
4. The method for evaluating the risk of an industrial chain based on a graph attention neural network according to claim 1, wherein the graph attention neural network evaluation model comprises an input layer, a graph attention layer, a dense full connection layer and an output layer, and the construction process of the graph attention neural network evaluation model comprises the following steps:
calculating a central feature vector and a node attribute feature vector of an industrial chain network node, and fusing and splicing the two feature vectors into a central node by using a graph attention mechanism to be embedded into an input layer serving as the graph attention neural network evaluation model;
embedding the fused nodes into a feature input diagram attention layer to obtain output features of the nodes;
inputting the graph annotation force layer output characteristics of the nodes into the dense full-connection layer to obtain the dense full-connection layer output characteristics of the nodes;
and inputting the output characteristics of the dense full-connection layer of the nodes into an output layer to obtain the risk prediction value of the nodes.
5. The method for evaluating risk of an industrial chain based on a graph attention neural network according to claim 4, wherein the feature vector fusion splicing process comprises:
firstly, calculating the central feature vector of a node network nodeAnd node Attribute feature vector->The expression is as follows:
wherein,representing a node network node centre feature vector +.>Weight of->Representing node Attribute feature vector +.>Sigma represents the LeakyReLU activation function, W c And W is X Representing a learnable transformation matrix->Representing network node centric feature vectorsAttention weight vector of corresponding discipline,/->Representing node Attribute feature vector +.>Attention weight vector of corresponding discipline, [ ·]A concatenation operation representing vectors;
the weights were then normalized by a Softmax function, expressed as follows:
finally, the node embedding after fusion is obtainedThe expression is as follows:
wherein,and->And representing the weight normalization result.
6. The method for evaluating risk of an industrial chain based on a graph attention neural network according to claim 4, wherein the feature input graph attention layer is embedded with the fused node, and the calculation process comprises:
vector embedding of the attention layer node is calculated, and the expression is as follows: z i ＝Wh i Wherein z is i Vector embedding representing node i, W representing a learnable weight matrix, h i Representing embedded features of node i;
the attention value of the node i and the first-order neighbor node j is calculated, and the expression is as follows: e, e ij ＝σ(α T [z i ||z j ]) Wherein σ represents the LeakyReLU activation function, α T Representing a matrix of learnable parameters, z j Vector embedding representing node j;
normalizing the attention value, and the expression is: alpha ij ＝softmax(e ij )；
Weighting and summing the node i to obtain the final output characteristic of the node iThe computational expression is: /> Wherein N (i) represents all first-order neighbor nodes of the node i, alpha ij Representing the normalization result of the attention value;
the computing expression of the output characteristics of the intensive full-connection layer is as follows:wherein d i Representing dense full-connection layer output characteristics of node i, x L ＝H L ([x 0 ,x 1 ,...,x L-1 ]) Represents a dense full-connection layer of the L-th layer, H L Is a cascading function;
the calculation expression of the risk prediction value of the node is as follows:wherein (1)>Representing the risk prediction value of node i,representing the learnable weights.
7. The method for evaluating the risk of an industrial chain based on a graph attention neural network according to claim 1, wherein the training method of the graph attention neural network evaluation model comprises the following steps:
selecting training nodes in the industrial chain network to be evaluated, calculating a risk prediction value of each node in the training nodes by using the graph-annotation-force neural network evaluation model, calculating a loss function, calculating the loss value according to the loss function, and updating the learnable parameters of the graph-annotation-force neural network evaluation model through back propagation;
repeating the steps until the loss value converges, and finishing training.
8. The method for evaluating the risk of an industrial chain based on a graph attention neural network according to claim 2, wherein the calculating the aggregate risk value of nodes at the same layer according to the node risk value and the attention relation coefficient specifically comprises:
respectively calculating an aggregation risk value represented by each relation information, and calculating the aggregation risk value of the nodes at the same layer according to the aggregation risk value represented by each relation information;
the calculation expression of the risk value aggregate value expressed by the competition relationship is as follows:wherein (1)>Aggregating risk values for competition relationships of the ith subdivision industry of the first layer, M (i) represents all first-order neighbor nodes under the competition relationship of the node i, M represents the number of the first-order neighbor nodes under the competition relationship of the node i, and +.>Representing the risk prediction value, alpha, of node i ij Representing the normalization result of the attention value;
the risk value aggregate value calculation expression expressed by the investment relation is as follows:wherein (1)>Aggregating risk values for investment relations of the ith subdivision industry of the first layer, wherein N (i) represents all first-order neighbor nodes under the investment relation of the node i, and M represents the number of the first-order neighbor nodes under the investment relation of the node i;
the risk value aggregate value calculation expression of the supply relation expression is:wherein (1)>Aggregating risk values, P, for supply relationships of a layer I, subdivision industry(i) Representing all first-order neighbor nodes in the node i supply relationship, and P represents the number of the first-order neighbor nodes in the node i supply relationship;
aggregation risk value of the same-layer nodeThe calculated expression of (2) is: />
9. The method for evaluating the risk of an industrial chain based on a graph attention neural network according to claim 1, wherein the calculation expression of the upstream, middle and downstream risk values is:
wherein R represents an upstream, midstream or downstream risk value,representing the aggregation risk value of the corresponding same-layer node;
the overall risk value R of the industrial chain to be evaluated i The calculated expression of (2) is:
wherein R is 1 Represents an upstream risk value, R 2 Represents midstream risk value, R 3 Representing a risk value downstream.
10. An apparatus for risk assessment of an industrial chain based on a graph attention neural network, the apparatus comprising:
the construction module is used for constructing a network topology structure of the upstream, the middle and the downstream of the industrial chain according to the data information of the industrial chain to be evaluated to obtain an industrial chain network to be evaluated;
the prediction module is used for constructing a graph attention neural network evaluation model and carrying out model training, and calculating a node risk value and an attention relationship coefficient of the industrial chain network to be evaluated by using the trained graph attention neural network evaluation model;
the calculation module is used for calculating the aggregate risk value of the nodes at the same layer according to the node risk value and the attention relation coefficient, calculating the risk value of the nodes at the cross layer according to the aggregate risk value of the nodes at the same layer to obtain an upper, middle and lower risk value, and calculating the overall risk value of the industrial chain to be evaluated according to the upper, middle and lower risk values.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311699916.6A CN117575328A (en) | 2023-12-12 | 2023-12-12 | Industrial chain risk assessment method and device based on graph attention neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311699916.6A CN117575328A (en) | 2023-12-12 | 2023-12-12 | Industrial chain risk assessment method and device based on graph attention neural network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117575328A true CN117575328A (en) | 2024-02-20 |
Family
ID=89864244
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311699916.6A Pending CN117575328A (en) | 2023-12-12 | 2023-12-12 | Industrial chain risk assessment method and device based on graph attention neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117575328A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117786131A (en) * | 2024-02-23 | 2024-03-29 | 广东省投资和信用中心(广东省发展和改革事务中心) | Industrial chain safety monitoring analysis method, medium and equipment |
-
2023
- 2023-12-12 CN CN202311699916.6A patent/CN117575328A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117786131A (en) * | 2024-02-23 | 2024-03-29 | 广东省投资和信用中心(广东省发展和改革事务中心) | Industrial chain safety monitoring analysis method, medium and equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Barak et al. | A novel hybrid fuzzy DEA-Fuzzy MADM method for airlines safety evaluation | |
| CN110110981B (en) | Credit rating default probability measure and risk early warning method | |
| Sheu | Dynamic relief-demand management for emergency logistics operations under large-scale disasters | |
| Ko et al. | Dynamic prediction of project success using artificial intelligence | |
| CN110298574B (en) | Convolutional neural network-based electricity user payment risk rating method | |
| Hajek et al. | Intuitionistic fuzzy grey cognitive maps for forecasting interval-valued time series | |
| CN117575328A (en) | Industrial chain risk assessment method and device based on graph attention neural network | |
| Luo et al. | Bayesian belief network-based project complexity measurement considering causal relationships | |
| Agami et al. | An innovative fuzzy logic based approach for supply chain performance management | |
| CN111429016A (en) | Small and micro enterprise financing wind control method and system based on industrial internet platform | |
| Guo et al. | Automatic design for shop scheduling strategies based on hyper-heuristics: A systematic review | |
| Morozko et al. | Applying the theory of fuzzy logic in the financial management of small companies | |
| CN114154866A (en) | Marketing enterprise financial risk early warning method and system | |
| KR102499181B1 (en) | Loan regular auditing system using artificia intellicence | |
| CN117495094A (en) | Comprehensive evaluation and early warning method and system for safety risk of industrial chain | |
| CN116911994B (en) | External trade risk early warning system | |
| Zhang | A deep learning model for ERP enterprise financial management system | |
| Hermansah et al. | Automatic time series forecasting using nonlinear autoregressive neural network model with exogenous input | |
| CN110827134A (en) | Power grid enterprise financial health diagnosis method | |
| CN117455681A (en) | Service risk prediction method and device | |
| Ying et al. | Research on tax inspection case selection model based on Bayesian network | |
| Sumo et al. | Performance prediction of a textile reverse logistics system using DEA and ANFIS hybrid models | |
| Seabrook et al. | Structural importance and evolution: An application to financial transaction networks | |
| Mashinchi et al. | Constructing a customer’s satisfactory evaluator system using GA-based fuzzy artificial neural networks | |
| Chen et al. | A ranking of software reliability evaluation based on intuitionistic fuzzy aggregation technique |
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 |