Disclosure of Invention
In order to solve the problem of inaccurate risk identification of an automobile supply chain set forth in the background art, the present invention provides the following aspects.
In a first aspect, the present invention provides a knowledge graph-based risk identification method for an automobile supply chain, including:
The method comprises the steps of obtaining entity types of an automobile supply chain, dividing each entity type into a hierarchy of the automobile supply chain, wherein each entity type comprises a plurality of entities, each two entities comprise one or more relations, taking the entities as nodes, the relation between the entities as edges, constructing a knowledge graph of the automobile supply chain, calculating the nodes, and obtaining the knowledge graph of the automobile supply chainRisk to the supply chain of a motor vehicle:
;
In the formula,Representing nodesTo the nodeIs used for the side-weight of (1),Representing nodesAt the time ofThe amount of initial risk change at the time of the event,Representation and nodeA set of connected nodes that are connected to each other,Representing nodesAt the level of (a) the (c) is,Representing nodesAt the time ofInitial risk variable quantity when the edge weight and the nodeSum nodeThe method comprises the steps of performing positive correlation on entropy weight and initial index value of the relation, responding to low risk of nodes with risks smaller than a set threshold value, responding to high risk of nodes with risks larger than or equal to the set threshold value, inputting initial risk variation of each node and corresponding risk levels into a machine model for training to obtain a trained machine model, inputting initial risk variation of nodes to be detected into the trained machine model, and outputting risk levels of the nodes to be detected.
According to the technical scheme, the knowledge graph of the automobile supply chain is constructed, the entities and the relations thereof in the supply chain are represented in a structured mode, and the risk assessment algorithm based on the relations among the nodes is introduced, so that the risk level of each node in the supply chain is assessed dynamically, the complex relations among the entities in the supply chain can be represented more accurately, and the comprehensiveness and accuracy of risk identification of the automobile supply chain are improved.
Further, the nodeTo the nodeSide weight of (2)The method comprises the following steps:
;
In the formula,Representing nodesAnd nodeIs the first of (2)The entropy weight of the individual relationship,Representing nodesAnd nodeFirst, theThe initial index value of the individual relationship is,Representing nodesAnd nodeTotal number of relationships between.
According to the technical scheme, the entropy weight and the initial index value are introduced, so that the strength of the relation between the nodes can be measured more accurately, the influence of a plurality of relation dimensions can be reflected by the edge weights, and the accuracy and the flexibility of risk calculation are improved. The multidimensional weighting mode can capture complex relation structures in the supply chain, and further improves the accuracy of risk identification and early warning of the supply chain.
Further, the method also comprises the step of calculating the node by adopting an entropy weight methodAnd nodeIs the first of (2)Entropy weight of personal relationshipThe method specifically comprises the following steps:
;
In the formula,Representing nodesAnd nodeBetween the firstThe entropy of the information of the individual relations,Representing nodesAnd nodeTotal number of relationships between.
According to the technical scheme, the entropy weight of the relation between the nodes is calculated by introducing the entropy weight method, and the uncertainty is measured by utilizing the information entropy, so that the contribution of each relation to the side weight is dynamically adjusted. By adjusting the relation weight of the information entropy, the role of the key and stable relation in risk identification is enhanced, so that the rationality and the accuracy of risk assessment are improved, and the risk identification is more in line with the actual situation of a supply chain.
Further, the method also comprises the step of preprocessing the initial index value by adopting Z-score standardization.
According to the technical scheme, the Z-score standardized pretreatment is adopted for the initial index values, so that the initial index values of different relations are comparable, the influence of different dimensions on calculation is eliminated, and the stability of data processing and the accuracy of calculation results are effectively improved. This approach further enhances the accuracy of supply chain risk assessment by converting each index value into a standard normal distribution.
Further, the entity types include suppliers, manufacturers, distributors, retailers, logistics service providers, and consumers.
Further, the entities of the supplier include financial status, primary service and part quality of the supplier;
The entities of the distributor comprise a distribution network, market coverage rate and logistics capability of the distributor, the entities of the retailer comprise sales amount and supply chains of the retailer, the entities of the logistics service provider comprise transportation capability, storage capability and transportation cost of the logistics service provider, and the entities of the consumer comprise purchasing power, consuming habit and purchasing behavior of the consumer.
Further, the relationships include a supply relationship, a manufacturing relationship, a distribution relationship, a retail relationship, a logistics relationship, and a purchasing relationship.
Further, the machine model is a random forest model.
According to the technical scheme, the random forest model is adopted, and the accuracy and the robustness of the risk prediction of the supply chain can be improved by utilizing the high-efficiency processing capacity and the strong generalization capacity of the random forest model to complex data. By integrating the results of a plurality of decision trees, the random forest model can reduce the risk of overfitting, and simultaneously improve the adaptability of the model to different types of data and the accurate judging capability of potential risks.
Further, the initial risk variation is the node difference between the current time point and the previous time point, and the initial index value is the association degree of the relationship between the nodes.
The invention has the beneficial effects that:
According to the invention, by constructing an automobile supply chain model based on a knowledge graph, the risk influence of each node on the supply chain is accurately estimated by using the relation between the nodes and the edges, the entropy weight and the initial risk variation, and the stability of data processing is improved through Z-score standardization. The risk classification is carried out by combining a random forest model, so that the automation and the high efficiency of the risk identification of the supply chain are realized, the comprehensive analysis can be carried out aiming at the entity relations of different levels, and the accuracy and the reliability of the risk early warning of the supply chain are obviously improved.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.
An embodiment of an automobile supply chain risk identification method based on a knowledge graph.
As shown in fig. 1, a flow chart of an automobile supply chain risk identification method based on a knowledge graph according to an embodiment of the invention includes the following steps:
And S1, acquiring entity types of the automobile supply chain, dividing the levels according to the entity types, and constructing an automobile supply chain knowledge graph.
The automobile supply chain entity types comprise suppliers, manufacturers, distributors, retailers, logistics service providers and consumers, the manufacturers can be automobile host factories, component factories, automobile body factories and the like, the suppliers can be primary component suppliers, secondary component suppliers, tertiary component suppliers and the like, the distributors can be regional distributors, private distributors and independent distributors, the retailers can be second-hand car distributors, component retailers and the like, the logistics service providers can be railway freight companies, airlines, ship companies and the like, the consumers can be personal consumers, enterprise clients, public institutions and the like, the entities of the suppliers comprise financial conditions, main business, production cycles, settlement modes, component qualities and the like of the suppliers, the entities of the manufacturers comprise production capacities, product qualities, development capacities, partnerships and the like of the manufacturers, the entities of the distributors comprise distribution networks, market coverage, inventory management capacities, logistics capacities and the like, the entities of the retailers comprise sales capacities, the purchase capacities, the entity of the consumers comprise the consumer and the purchasing capacities of the retailers, the logistics service providers comprise the consumer's and the logistics service areas, the consumer's satisfaction and the weather conditions comprise the consumer's satisfaction and the logistics conditions. Of course, the entity type and entity in the automobile supply chain can be set according to the actual situation.
In the embodiment, the automobile supply chain is assumed to be in a pyramid structure, each entity type is a level of the automobile supply chain, and one or more relationships are included between every two entities, wherein the relationships comprise a supply relationship, a manufacturing relationship, a distribution relationship, a retail relationship, a logistics relationship and a purchasing relationship, for example, a financial condition of a supplier and production capacity of the manufacturer have a supply relationship, a logistics capability of the distributor and the supply chain of the retailer have a retail relationship and a logistics relationship, and the entities serve as nodes, and the relationship between the entities serves as edges to construct a knowledge graph of the automobile supply chain.
And S2, calculating risks generated by node risk changes on the whole automobile supply chain.
In one embodiment, a compute nodeRisk to the supply chain of a motor vehicle:
;
In the formula,Representing nodesTo the nodeIs used for the side-weight of (1),Representing nodesAt the time ofThe amount of initial risk change at the time of the event,Representation and nodeA set of connected nodes that are connected to each other,Representing nodesAt the level of (a) the (c) is,Representing nodesAt the time ofInitial risk variable quantity, side weight and nodeSum nodeThe entropy weight and the initial index value of the relation are positively correlated;
NodeTo the nodeSide weight of (2)The method comprises the following steps:
;
In the formula,Representing nodesAnd nodeIs the first of (2)The entropy weight of the individual relationship,Representing nodesAnd nodeFirst, theThe initial index value of the individual relationship is,Representing nodesAnd nodeTotal number of relationships between.
The method also comprises the step of calculating the node by adopting an entropy weight methodAnd nodeIs the first of (2)Entropy weight of personal relationshipThe method specifically comprises the following steps:
;
In the formula,Representing nodesAnd nodeBetween the firstInformation entropy of the personal relationship.
As shown in fig. 2, a schematic diagram of an automobile supply chain knowledge graph according to an embodiment of the present invention.
Exemplary As shown in FIG. 2, assume that three nodes are included in the automotive supply chain, node A, node B, and node C, with node A at a first level, node B at a second level, node C at a third level, and node A is acquired at timeThe initial risk variation at the time is 0.05, and the node B is at the timeThe initial risk variation is 0.02, and node C is at timeIf the manufacturing relationship and the distribution relationship exist between the alpha side of the node A and the alpha side of the node B, the manufacturing relationship and the distribution relationship exist between the beta side of the node A and the beta side of the node C, the entropy weight of the manufacturing relationship between the node A and the node B is 0.4, the entropy weight of the manufacturing relationship between the node A and the node C is 0.3, the entropy weight of the distribution relationship between the node A and the node B is 0.6, and the entropy weight of the distribution relationship between the node A and the node C is 0.7;
The method comprises the steps of obtaining an initial index value of a manufacturing relationship between a node A and a node B to be 0.5, obtaining an initial index value of a distribution relationship between the node A and the node B to be 0.7, obtaining an initial index value of a distribution relationship between the node A and the node C to be 0.6, substituting the initial index value of the distribution relationship between the node A and the node C to be 0.8 into the side weight calculation formula to obtain an edge weight value of 0.62 for the node A and the node B, substituting the side weight value of 0.74 for the node B and the node C into the risk formula of the automobile supply chain to obtain the total risk of the supply chain caused by the node A to be 0.0285.
In one embodiment, the initial index value is a degree of association of the relationship between the nodes, and the degree of association is obtained by quantifying the relationship between the nodes by a quantitative evaluation method or a risk factor analysis method, for example, the initial index value obtained by quantifying the manufacturing relationship between the node a and the node B by the quantitative evaluation method is 0.5, and the initial index value obtained by quantifying the distribution relationship between the node a and the node B by the quantitative evaluation method is 0.7.
And one node comprises a plurality of risk factors, and the sum of the risk level variation amounts of the risk factors is the initial risk variation amount of the node. The first method is to manually increase the initial risk variation, the second method is to obtain the initial risk variation according to the node difference between the current time point and the previous time point, for example, the liability of a certain provider at the previous time point is 0, the liability at the current time point is 0.25, the state of sales from the previous time point to the current time point is changed from 0 to 0.25, the initial risk variation of the node is 0.25, the third method is to construct a neural network prediction model, divide weights according to the influence of each risk factor on the node, predict the probability of the change of the risk factor, multiply the probability of each risk factor with the weights and then add the sum of risk levels, and the second method is used in the embodiment.
And S3, based on risk classification risk levels, inputting the initial risk variation and the corresponding risk levels into the machine model for training to obtain a trained machine model.
Specifically, traversing each node in the supply chain is at timeWhen the risk is generated for the automobile supply chain, the node with the risk smaller than the set threshold value is a low risk, and the node with the risk larger than or equal to the set threshold value is a high risk;
in the present embodiment, the threshold value is set to 0.6, but may be set according to actual situations. And finally, inputting the initial risk variation of each node and the corresponding risk level into a machine model for training to obtain a trained machine model, inputting the initial risk variation of the node to be detected into the trained machine model, and outputting the risk level of the node to be detected. Wherein, the machine model is a random forest model, and the specific training process is as follows:
The method comprises the steps of obtaining a training set, wherein the training set comprises initial risk variation of each node and corresponding risk levels, and labels in the training set are calibrated manually.
The training set is input into a pre-built random forest model for training, the loss between the output predicted value and the label is calculated in the training process, model parameters are adjusted by a gradient descent method to minimize the prediction error, and the parameters of the random forest model are iteratively adjusted until the loss is smaller than a certain value or reaches the set training times, so that the trained random forest model is finally obtained.
And S4, inputting the risk variation quantity of the node to be detected into a trained machine model, and outputting the risk grade of the node to be detected.
In one embodiment, it is assumed that knowledge graph construction of the automotive supply chain has been completed, including entities and their interrelationships of suppliers, manufacturers, distributors, retailers, logistics service providers, and consumers. Each entity type is divided into different levels of the supply chain, e.g., manufacturer is the upstream level of the supplier and retailer is the downstream level of the supplier, the entities of the supplier include financial status, main business and part quality of the supplier, the entities of the manufacturer include production capacity and product quality of the manufacturer, and the financial status of the supplier has a supply relationship with the production capacity of the manufacturer. All of these entities and the relationships between them are represented as nodes and edges, forming a detailed knowledge graph of the automotive supply chain.
Then, the initial risk variation amounts of all nodes in the knowledge graph of the automobile supply chain are collected. And inputting the collected initial risk variation data of all nodes in the automobile supply chain into the trained random forest model, so as to obtain the risk level of the nodes to be detected in the automobile supply chain.
In this embodiment, after determining the risk level of the node to be detected, a reminder may be further performed, for example, a reminder is performed to related management personnel in the form of an alarm or a short message, and adaptive adjustment is performed on each node in time.
According to the scheme, by constructing the knowledge graph-based risk identification method of the automobile supply chain, by utilizing the technologies of an entropy weight method, Z-score standardized preprocessing, a random forest model and the like, complex entity relations and hierarchical structures in the supply chain can be effectively captured, and accurate risk assessment and classification are realized. Particularly, through entropy weight calculation and standardization processing of the multidimensional relation, the identification capability of potential risks is enhanced, and the prediction accuracy and generalization capability of the model are further improved by the random forest model, so that a more reliable risk early warning mechanism is provided for supply chain management.
In the description of the present specification, the meaning of "a plurality", "a number" or "a plurality" is at least two, for example, two, three or more, etc., unless explicitly defined otherwise.
While various embodiments of the present invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. Many modifications, changes, and substitutions will now occur to those skilled in the art without departing from the spirit and scope of the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed in practicing the invention.