Supply chain risk control processing method and equipment

Abstract

The present application provides a supply chain risk control processing method and device, which obtains the commodity evaluation information of the first node of the supply chain, classifies the commodity evaluation information based on a preset random forest algorithm, and determines the classified commodity evaluation short text as the commodity risk control word corresponding to the first node. The keywords and commodity risk control words in the commodity evaluation information are input into the risk control text model to determine the supply relationship queue between the first node and several second nodes. According to the supply chain information, the historical transaction data corresponding to the supply relationship queue with a risk control value greater than a first preset threshold is input into the preset convolutional neural network model in turn to determine the risk control index of the second node corresponding to the supply relationship queue. Based on the risk control value, risk control index, historical transaction data and the preset expert prediction model, the risk control risk level of the corresponding second node is determined to generate prompt information based on the risk control risk level, and the prompt information is sent to the corresponding supervision terminal.

Description

Technical Field

[0001] The present application relates to the field of supply chain technology, and in particular to a supply chain risk control processing method and equipment. Background Art

[0002] A supply chain refers to the network of upstream and downstream businesses involved in the production and distribution process that delivers products or services to end users. The supply chain can be visually depicted as a lush tree, with each business and user forming part of the tree. While the relationships between businesses in a supply chain can foster mutual support and survival, they can also lead to a "survival of the fittest" situation, causing a stable supply chain to become unbalanced.

[0003] Currently, the main drivers of supply chain instability include internal factors, external competition, and environmental factors. For businesses, once a supply chain becomes unbalanced, supply relationship risks may arise, potentially leading to financial and reputational losses, or even bankruptcy. Supply chain risks often arise from seemingly insignificant factors. If businesses can promptly identify risks, implement mitigation measures, and manage supply chain risks, they will improve supply chain stability. Currently, supply chain risk management primarily relies on manual labor, relying on the experience of relevant personnel and resulting in wasted manpower. Therefore, a technical solution is needed that can conserve human resources and promptly identify and manage supply chain risks. Summary of the Invention

[0004] The embodiments of the present application provide a supply chain risk control processing method and device, which are used to help enterprises to timely identify risks, provide avoidance measures, perform risk control on the supply chain, and improve the stability of the supply chain.

[0005] In one aspect, an embodiment of the present application provides a supply chain risk control processing method, the method comprising:

[0006] Obtain product evaluation information for a first node in the supply chain. The product evaluation information includes at least evaluation information directly related to the first node and evaluation information indirectly related to the first node. Classify the product evaluation information based on a preset random forest algorithm, and determine the classified product evaluation short text as the product risk control term corresponding to the first node. The product evaluation short text consists of several words that characterize the product. Input the keywords and product risk control terms within the product evaluation information into a pre-trained risk control text model to determine a supply relationship queue between the first node and several second nodes. Each supply relationship queue includes at least one second node and its corresponding risk control value. Based on the supply chain information, input historical transaction data corresponding to supply relationship queues with risk control values ​​greater than a first preset threshold into a preset convolutional neural network model to determine the risk control indicators of the second nodes corresponding to the supply relationship queues. Based on the risk control value, risk control indicator, historical transaction data, and a preset expert prediction model, determine the risk control risk level of the corresponding second node, generate a prompt message based on the risk control risk level, and send the prompt message to the corresponding supervisory terminal.

[0007] In an embodiment of the present application, a number of commodity names are determined based on supply chain information. The supply chain information includes at least: the name of the supply commodity, the flow direction of the supply commodity, the price of the supply commodity, and the logistics information of the supply commodity. Each commodity name is vectorized, and the vectorized commodity name is input into a preset evaluation text library to determine the commodity evaluation text corresponding to the commodity name vector and the evaluation words corresponding to the commodity evaluation text. The evaluation words represent the evaluation intention of the commodity evaluation text to the commodity. In the case that there is no corresponding commodity evaluation text for the commodity name vector, the commodity evaluation text corresponding to the commodity name vector is determined, and the commodity evaluation text is input into the classifier of the random forest algorithm to determine the corresponding evaluation words. The evaluation words and the corresponding commodity evaluation text are added to the classifier extension set. In the case that the number of elements in the classifier extension set is equal to the number of elements in the preset training set of the classifier, the classifier extension set is input into the classifier to update the classifier until the number of training times of the preset training set reaches the second preset threshold.

[0008] In an embodiment of the present application, product evaluation information is input into a random forest algorithm to determine a number of evaluation words corresponding to the product evaluation information. The word vector of each evaluation word is determined by a preset Word2vec model. The word vector weight of the word vector is determined based on the word paragraph corresponding to the evaluation word. Based on the word vector and the word vector weight, the product evaluation short text corresponding to the evaluation word is determined, and the product evaluation short text is used as the product risk control word corresponding to the first node. The product evaluation short text is pre-set in the short text database.

[0009] In an embodiment of the present application, keywords in the evaluation information corresponding to product risk control terms are input into the risk control text model to determine at least one second node corresponding to the evaluation information. Keywords are the terms in the evaluation information after removing stop words. Based on the at least one second node and the order of commodity transactions in the supply chain, a supply ray is established between the first node and the at least one second node to determine a supply relationship queue. The supply ray points from the commodity supply node to the commodity purchase node.

[0010] In an embodiment of the present application, based on the transaction commodity information of the first node and the second node, the commodity event name corresponding to the transaction commodity information is determined from the preset event comparison table. Among them, the commodity event name includes at least commodity quality problems, regional epidemics, and corporate financial breaks. According to the preset portrait model, several predicted events of the commodity risk control words and the probability of occurrence of each predicted event are determined. Among them, the predicted event is a preset malicious event that affects the stability of the supply chain. Determine the cosine similarity between each predicted event and each commodity event name, and the predicted event that meets the preset conditions is a pending event, and determine the risk control value corresponding to the pending event. Among them, the preset conditions are that the cosine similarity is greater than the first preset value and the probability of occurrence of the predicted event is greater than the second preset value. According to the risk control value corresponding to the pending event and the supply ray between the first node and at least one second node, determine the supply relationship queue between the first node and several second nodes.

[0011] In an embodiment of the present application, the historical transaction data corresponding to the supply relationship queue whose risk control value is greater than the first preset threshold is determined according to the corresponding commodity transaction sequence of the supply chain. The historical transaction data includes the commodity transaction time, transaction quantity, commodity type, and commodity production process. The historical transaction data is input into the convolutional neural network model to determine a number of risk control words for each transaction commodity and the inverse document frequency of each risk control word among the several risk control words. Based on the inverse document frequency and the risk value of each risk control word itself, two risk control word sequences of the supply relationship queue are generated. The two risk control word sequences are de-duplicated and merged, and the risk control word sequence after de-duplication and merging is used as the risk control indicator of the supply relationship queue.

[0012] In this embodiment of the present application, historical transaction data and risk control values ​​are input into an expert prediction model to determine the transaction risk weight for each historical transaction in the historical transaction data. The risk control value is then updated based on the transaction risk weight. Based on the updated risk control value and risk control indicator, the expert prediction model is used to assign an indicator to the risk control indicator to determine a corresponding risk index value. The risk index value is greater than 0 and not greater than 1. Based on the risk control value and risk index value, the risk control risk level of each second node is determined.

[0013] In an embodiment of the present application, a supply relationship diagram of supply chain information is established through network mapping software, and the risk control risk level is marked in the relationship connection line between the first node and each second node. According to each risk control risk level, the second node in the supply relationship diagram whose risk control risk level is greater than the third preset value is determined to be the node to be processed. Based on the historical transaction data of the first node and the node to be processed, the corresponding alternative node and subsequent transaction method of the node to be processed are determined to generate risk control processing information. Among them, the commodity transaction quantity, commodity type and commodity production process of the alternative node match the historical transaction data of the node to be processed. The risk control processing information includes risk control processing targets and risk control processing methods.

[0014] In an embodiment of the present application, when a node to be processed is determined, a prompt message is generated and the prompt message is sent to the supervisory terminal corresponding to the node to be processed. When receiving supplementary information data fed back by the supervisory terminal, the supplementary information data is sent to the source terminal corresponding to the product evaluation information to determine whether the supplementary information data is valid. If so, the updated risk control risk level of the second node is determined based on the preset risk control value, risk control indicators, historical transaction data and expert prediction model, and whether the second node is a node to be processed is determined based on the updated risk control risk level, until the number of acquisitions of the supplementary information data is greater than the fourth preset value, risk control processing information is generated, and a prompt message is generated and sent to the supervisory terminals corresponding to the first node and the node to be processed respectively. Otherwise, the corresponding alternative nodes and subsequent transaction methods of the node to be processed are determined to generate risk control processing information, and a prompt message is generated and sent to the supervisory terminals corresponding to the first node and the node to be processed respectively.

[0015] On the other hand, an embodiment of the present application provides a supply chain risk control processing device, the device comprising:

[0016] At least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0017] Obtain product evaluation information for a first node in the supply chain. This product evaluation information includes at least evaluation information directly related to the first node and evaluation information indirectly related to the first node. Classify the product evaluation information based on a preset random forest algorithm, and determine the classified product evaluation short text as the product risk control term corresponding to the first node. The product evaluation short text consists of several words that characterize the product. Input the keywords and product risk control terms within the product evaluation information into a pre-trained risk control text model to determine a supply relationship queue between the first node and several second nodes. Each supply relationship queue includes at least one second node and its corresponding risk control value. Based on the supply chain information, input historical transaction data corresponding to each supply relationship queue with a risk control value greater than a first preset threshold into a preset convolutional neural network model to determine the risk control indicator for the second node corresponding to the supply relationship queue. Determine the risk control level of the corresponding second node based on the risk control value, risk control indicator, historical transaction data, and a preset expert prediction model. Generate a prompt based on the risk control risk level and send the prompt to the corresponding monitoring terminal.

[0018] Through the above scheme, based on the product evaluation information of the first node in the supply chain, several second nodes that have a supply relationship with the first node are determined, and by processing the product evaluation information, the predicted impact of each second node on the supply chain (risk control value) is obtained. And through historical transaction data and expert prediction models, the risk control risk level of each second node is obtained, thereby prompting the corresponding regulatory terminal, and also prompting the risk control processing target and risk control processing method. The above scheme can help enterprises to timely discover risks in the supply chain and provide enterprises with measures to avoid risks, so as to perform risk control on the supply chain and improve the stability of the supply chain. In addition, by presetting the random forest algorithm, the product evaluation information is classified, and it can be classified with an efficient classifier to ensure classification efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 A flow chart of a supply chain risk control processing method in an embodiment of the present application;

[0021] Figure 2 This is a structural diagram of a supply chain risk control processing device in an embodiment of the present application. DETAILED DESCRIPTION

[0022] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0023] The embodiments of the present application provide a supply chain risk control processing method and device to help enterprises detect risks in a timely manner, provide avoidance measures, perform risk control on the supply chain, and improve the stability of the supply chain.

[0024] The following describes in detail various embodiments of the present application with reference to the accompanying drawings.

[0025] The present application embodiment provides a supply chain risk control processing method, such as Figure 1 As shown, the method may include steps S101-S105:

[0026] S101, the server obtains the product evaluation information of the first node of the supply chain.

[0027] The product evaluation information includes at least: evaluation information that is directly related to the first node and evaluation information that is indirectly related to the first node.

[0028] The first node of a supply chain refers to an enterprise terminal within the supply chain network structure that needs or desires to conduct risk control monitoring of current supply relationships. Enterprise terminals can be enterprise mobile phones, computers, and other devices, and this application does not specifically limit this. Product evaluation information refers to evaluation information related to the first node. For example, if the first node in the supply chain has a direct product transaction relationship with Enterprise A, evaluation information for the first node or for Enterprise A is considered product evaluation information. This evaluation information includes, but is not limited to, news, interview information, user evaluations of products, entrepreneur evaluations of enterprises, and product logistics evaluations.

[0029] For another example, the first node in the supply chain has an indirect commodity trading relationship with Company B, and Company B has a direct commodity trading relationship with Company A. The evaluation information of Company B also belongs to the commodity evaluation information of the first node.

[0030] It should be noted that the server, as the executor of the supply chain risk control processing method, is only an example. The executor is not limited to the server, and this application does not make any specific restrictions on this.

[0031] S102: The server classifies the product evaluation information based on a preset random forest algorithm, and determines the classified product evaluation short text as the product risk control word corresponding to the first node.

[0032] The short text of product review consists of several words that represent the product situation.

[0033] In an embodiment of the present application, before the server classifies the product review information based on a preset random forest algorithm and determines that the classified product review short text is the product risk control word corresponding to the first node, the method further includes:

[0034] First, the server determines several product names based on the supply chain information.

[0035] The supply chain information includes at least: the name of the supplied goods, the flow direction of the supplied goods, the price of the supplied goods, and the logistics information of the supplied goods.

[0036] Specifically, based on the sales data of enterprises in the supply chain, the commodity names of commodities circulating between various nodes of the supply chain are obtained, such as: wine, crops, snacks, etc.

[0037] The server then vectorizes each product name and enters the vectorized product name into a pre-set review text library to determine the product review text corresponding to the product name vector and the corresponding evaluation terms. The evaluation terms represent the evaluation intent of the product review text.

[0038] Word2vec can be used for vectorization processing to vectorize the product names and obtain the vector corresponding to each product name.

[0039] In the case that there is no corresponding product evaluation text for the product name vector, the server then determines the product evaluation text corresponding to the product name vector and inputs the product evaluation text into the classifier of the random forest algorithm to determine the corresponding evaluation words.

[0040] The server can perform the above operations over the internet. Based on the product vector, the server searches the internet and obtains the product evaluation text corresponding to the vector. For example, for product X, the evaluation text may be: poor quality, expensive, pattern does not match the sample pattern, etc. The classifier will classify the evaluation words into: quality problem, price problem, appearance paint problem, etc.

[0041] Next, the server adds the evaluation words and the corresponding product evaluation text to the classifier extension set.

[0042] The evaluation words and product evaluation texts obtained by the classifier are added to the classifier extension set. The maximum number of elements in the extension set is equal to the number of elements in the training set. For example, if a set in the training set consists of 10 words, then the maximum number of elements in the extension set is 10, that is, there can be 10 evaluation words.

[0043] When the number of elements in the classifier extension set is equal to the number of elements in the preset training set of the classifier, the classifier extension set is input into the classifier to update the classifier until the number of training times of the preset training set reaches a second preset threshold.

[0044] The second preset threshold is set according to actual use and is not specifically limited in this application.

[0045] In the embodiment of the present application, the product review information is classified based on a preset random forest algorithm, and the classified product review short text is determined as the product risk control word corresponding to the first node, specifically including:

[0046] First, the server inputs the product review information into the random forest algorithm to determine several evaluation terms corresponding to the product review information. Next, the server determines the word vector of each evaluation term by presetting the Word2vec model. Subsequently, the server determines the word vector weight of the word vector based on the word paragraph corresponding to the evaluation term. For example, if the word paragraph is x, the vector weight is 1 / x. Then, based on the word vector and the word vector weight, the server determines the product review short text corresponding to the evaluation term, and uses the product review short text as the product risk control term corresponding to the first node. Among them, the product review short text is pre-set in the short text database.

[0047] Specifically, the server performs vector multiplication on the word vector and the vector weight to obtain a new vector. Through the new vector, the server matches the product evaluation short text corresponding to the new vector from the short text database to obtain the product risk control word.

[0048] S103, the server inputs the keywords and product risk control words in the product evaluation information into a pre-trained risk control text model to determine the supply relationship queue between the first node and several second nodes.

[0049] The supply relationship queue includes at least one second node information and its corresponding risk control value. The risk control value represents the degree of impact of the product risk control word on the supply chain forecast.

[0050] In the embodiment of the present application, keywords and product risk control words in product evaluation information are input into a pre-trained risk control text model to determine the supply relationship queue between a first node and a plurality of second nodes, specifically including:

[0051] First, the server inputs the keywords in the evaluation information corresponding to the product risk control words into the risk control text model to determine at least one second node corresponding to the evaluation information.

[0052] Keywords are words in the review information after stop words are removed. The risk control text model is a pre-trained neural network model that can identify the second and first nodes involved in the review information. For example, if the review information contains a company name or a description of a problem with a related product, the risk control text model can process the review information to obtain a second node corresponding to the company name, or a company name that may cause a problem, thereby obtaining a second node. The risk control text model sample includes at least the names of all companies in the aforementioned supply chain and product-related information, such as product processing techniques and product selling companies.

[0053] Next, the server establishes a supply ray between the first node and the at least one second node based on the at least one second node and the commodity transaction sequence of the supply chain to determine a supply relationship queue, wherein the supply ray points from the commodity supply node to the commodity purchase node.

[0054] In the embodiment of the present application, keywords and product risk control words in product evaluation information are input into a pre-trained risk control text model to determine the supply relationship queue between a first node and a plurality of second nodes, specifically including:

[0055] First, the server determines the product event name corresponding to the transaction product information from the preset event comparison table based on the transaction product information of the first and second nodes. The product event name includes at least product quality issues, regional epidemics, and company financial disruptions.

[0056] Next, the server determines several predicted events for the product risk control words and the probability of each predicted event based on the preset profiling model. The predicted events are preset malicious events that affect the stability of the supply chain.

[0057] Malicious incidents include: epidemics, illegal fundraising by partners, tax evasion, etc.

[0058] The server then determines the cosine similarity between each predicted event and each product event name. The server then determines the predicted event that meets the preset conditions as a pending event and determines the risk control value corresponding to the pending event. The preset conditions are that the cosine similarity is greater than a first preset value and the probability of the predicted event occurring is greater than a second preset value.

[0059] Finally, the server determines a supply relationship queue between the first node and the plurality of second nodes according to the risk control value corresponding to the pending event and the supply ray between the first node and at least one second node.

[0060] There are supply relationship queues between the first node and each second node, and the number of the supply relationship queues is the same as the number of the second nodes.

[0061] S104: Based on the supply chain information, the server sequentially inputs the historical transaction data corresponding to the supply relationship queues whose risk control values ​​are greater than the first preset threshold into a preset convolutional neural network model to determine the risk control index of the second node corresponding to the supply relationship queue.

[0062] In an embodiment of the present application, based on supply chain information, historical transaction data corresponding to supply relationship queues whose risk control values ​​are greater than a first preset threshold are sequentially input into a preset convolutional neural network model to determine the risk control indicator of the second node corresponding to the supply relationship queue, specifically including:

[0063] First, the server determines the historical transaction data corresponding to the supply relationship queue with a risk control value greater than a first preset threshold based on the corresponding commodity transaction sequence of the supply chain. The historical transaction data includes commodity transaction time, transaction quantity, commodity type, and commodity production process.

[0064] The server then feeds the historical transaction data into a convolutional neural network model to determine several risk control terms for each traded commodity and the inverse document frequency of each risk control term within the risk control term pool. This convolutional neural network model, trained on a number of transaction samples, converts the transaction samples into text data, performs word segmentation, removes stop words, and then outputs several risk control terms that could affect the transaction and the inverse document frequency of each risk control term within the overall risk control term pool.

[0065] Then, the server generates two risk control word sequences for the supply relationship queue based on the inverse document frequency and the risk value of each risk control word.

[0066] A first risk control word sequence is generated based on the risk control words whose inverse document frequency exceeds a fifth preset threshold, in descending order. A second risk control word sequence is generated based on the risk control words whose intrinsic risk value exceeds a sixth preset threshold, in descending order. The intrinsic risk value of each risk control word is pre-set and obtained by matching it against a pre-set intrinsic risk value comparison table.

[0067] Finally, the server removes duplicates and merges the two risk control word sequences, and uses the risk control word sequences after the removal of duplicates and merger as the risk control indicators of the supply relationship queue.

[0068] Specifically, it is possible to search whether the second risk control word sequence contains the risk control words of the first risk control word sequence. If so, the risk control words are removed from the second risk control word sequence; then the second risk control word sequence is merged to the end of the first risk control word sequence, and the risk control words in the new word sequence are used as risk control indicators.

[0069] S105, the server determines the risk control level of the corresponding second node based on the risk control value, risk control indicators, historical transaction data and a preset expert prediction model, generates prompt information based on the risk control risk level, and sends the prompt information to the corresponding supervision terminal.

[0070] In an embodiment of the present application, the server determines the risk control level of the corresponding second node based on the risk control value, risk control indicators, historical transaction data, and a preset expert prediction model, generates prompt information based on the risk control risk level, and sends the prompt information to the corresponding supervision terminal, specifically including:

[0071] First, the server inputs the historical transaction data and risk control value into the expert prediction model to determine the transaction risk weight of each historical transaction in the historical transaction data, and updates the risk control value according to the transaction risk weight.

[0072] Then, the server uses the updated risk control value and risk control index and the expert prediction model to assign the risk control index to determine the corresponding risk index value. The risk index value is greater than 0 and not greater than 1.

[0073] Then, the server determines the risk control level of each second node based on the risk control value and the risk index value, multiplies the risk control value by the risk index value, and uses the product as the risk control risk level.

[0074] In the embodiment of the present application, after the server determines the risk control level of the corresponding second node, the method further includes:

[0075] First, the server uses network mapping software to create a supply relationship diagram of the supply chain information, and marks the risk control level in the relationship connection between the first node and each second node.

[0076] Then, the server determines, based on each risk control risk level, the second node in the supply relationship diagram whose risk control risk level is greater than the third preset value as the node to be processed.

[0077] Next, the server determines alternative nodes and subsequent transaction methods for the pending node based on the historical transaction data between the first node and the pending node, generating risk control information. The transaction volume, product type, and production process of the alternative node match the historical transaction data of the pending node. This risk control information includes the risk control objectives and the risk control methods.

[0078] In the embodiment of the present application, the server generates prompt information based on the risk control level and sends the prompt information to the corresponding supervision terminal, specifically including:

[0079] When a node to be processed is determined, the server generates prompt information and sends the prompt information to the supervisory terminal corresponding to the node to be processed. The prompt information can be the text of the evaluation information or the processed evaluation word text.

[0080] In the case of receiving the supplementary information data fed back by the supervision terminal, the server sends the supplementary information data to the source terminal corresponding to the product evaluation information to determine whether the supplementary information data is valid.

[0081] When it is determined that the supplementary information data is valid, the server determines the updated risk control risk level of the second node based on the preset risk control value, risk control indicators, historical transaction data and expert prediction model, and determines whether the second node is a node to be processed based on the updated risk control risk level. After the number of times the supplementary information data is obtained is greater than the fourth preset value, the server generates risk control processing information and prompt information and sends it to the supervision terminals corresponding to the first node and the node to be processed respectively.

[0082] When it is determined that the supplementary information data is invalid, the server determines the corresponding alternative node and subsequent transaction method of the node to be processed to generate risk control processing information, and generates prompt information, which is sent to the supervision terminals corresponding to the first node and the node to be processed respectively.

[0083] In other words, when a supply chain risk is identified, the server can send a prompt to the risk node, prompting it to address the risk. If the risk persists after a preset number of attempts, the server will provide the first node with a risk control solution, such as switching transaction nodes or terminating cooperation. If the risk resolution information provided is invalid, the server will directly determine the sub-control solution.

[0084] Through the above solution, based on the product evaluation information of the first node in the supply chain, several second nodes with a supply relationship with the first node are identified. By processing the product evaluation information, the predicted impact of each second node on the supply chain (risk control value) is obtained. Furthermore, through historical transaction data and expert prediction models, the risk control risk level of each second node is obtained, thereby determining the risk control treatment target and risk control method. The above solution can help enterprises promptly identify risks in the supply chain and provide them with risk avoidance measures, thereby implementing risk control measures for the supply chain and improving supply chain stability.

[0085] In addition, by pre-setting the random forest algorithm, the product evaluation information can be classified and classified using an efficient classifier to ensure classification efficiency and accuracy.

[0086] Figure 2 A schematic diagram of the structure of a supply chain risk control processing device provided in an embodiment of the present application, the device includes:

[0087] At least one processor; and a memory communicatively connected to the at least one processor. The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to:

[0088] Obtain product evaluation information for a first node in the supply chain. This product evaluation information includes at least evaluation information directly related to the first node and evaluation information indirectly related to the first node. Classify the product evaluation information based on a preset random forest algorithm, and determine the classified product evaluation short text as the product risk control term corresponding to the first node. The product evaluation short text consists of several words that characterize the product. Input the keywords and product risk control terms within the product evaluation information into a pre-trained risk control text model to determine a supply relationship queue between the first node and several second nodes. Each supply relationship queue includes at least one second node and its corresponding risk control value. Based on the supply chain information, input historical transaction data corresponding to each supply relationship queue with a risk control value greater than a first preset threshold into a preset convolutional neural network model to determine the risk control indicator for the second node corresponding to the supply relationship queue. Determine the risk control level of the corresponding second node based on the risk control value, risk control indicator, historical transaction data, and a preset expert prediction model. Generate a prompt based on the risk control risk level and send the prompt to the corresponding monitoring terminal.

[0089] The various embodiments in this application are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the device embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0090] The device and method provided in the embodiments of the present application correspond one to one, and therefore, the device also has similar beneficial technical effects as its corresponding method. Since the beneficial technical effects of the method have been described in detail above, the beneficial technical effects of the device will not be repeated here.

[0091] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0092] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A supply chain risk control processing method, characterized in that: The method comprises: Acquire product evaluation information of a first node in the supply chain; wherein the product evaluation information includes at least: evaluation information directly related to the first node and evaluation information indirectly related to the first node; Based on a preset random forest algorithm, the product evaluation information is classified, and the classified product evaluation short text is determined as the product risk control word corresponding to the first node; the product evaluation short text is composed of a number of words that represent the product situation; Input the keywords in the product review information and the product risk control words into a pre-trained risk control text model to determine a supply relationship queue between the first node and a plurality of second nodes; wherein the supply relationship queue includes at least one second node information and its corresponding risk control value; Based on the supply chain information, historical transaction data corresponding to the supply relationship queues having risk control values ​​greater than a first preset threshold are sequentially input into a preset convolutional neural network model to determine the risk control indicator of the second node corresponding to the supply relationship queue; Based on the risk control value, the risk control indicator, the historical transaction data and the preset expert prediction model, the risk control risk level of the corresponding second node is determined, and prompt information is generated based on the risk control risk level, and the prompt information is sent to the corresponding supervision terminal.

2. The method according to claim 1, characterized in that Before classifying the product review information based on a preset random forest algorithm and determining that the classified product review short text is the product risk control word corresponding to the first node, the method further includes: Determine several commodity names based on the supply chain information; wherein the supply chain information includes at least: the name of the supply commodity, the flow direction of the supply commodity, the price of the supply commodity, and the logistics information of the supply commodity; Each of the product names is vectorized, and the vectorized product names are input into a preset evaluation text library to determine the product evaluation text corresponding to the product name vector and the evaluation words corresponding to the product evaluation text; the evaluation words represent the evaluation intention of the product evaluation text on the product; If there is no corresponding product evaluation text for the product name vector, determining the product evaluation text corresponding to the product name vector, and inputting the product evaluation text into the classifier of the random forest algorithm to determine the corresponding evaluation words; Adding the evaluation words and the corresponding product evaluation text to the classifier extension set; When the number of elements in the classifier extension set is equal to the number of elements in the preset training set of the classifier, the classifier extension set is input into the classifier to update the classifier until the number of training times of the preset training set reaches a second preset threshold.

3. The method according to claim 2, characterized in that Based on a preset random forest algorithm, the product review information is classified, and the classified product review short text is determined as the product risk control word corresponding to the first node, specifically including: Inputting the product evaluation information into the random forest algorithm to determine a number of evaluation words corresponding to the product evaluation information; Determine the word vector of each evaluation word by presetting the Word2vec model; Determining a word vector weight of the word vector according to the word paragraph corresponding to the evaluation word; According to the word vector and the word vector weight, the product evaluation short text corresponding to the evaluation term is determined, so as to use the product evaluation short text as the product risk control word corresponding to the first node; wherein the product evaluation short text is pre-set in a short text database.

4. The method according to claim 1, characterized in that Input the keywords in the product evaluation information and the product risk control words into a pre-trained risk control text model to determine the supply relationship queue between the first node and several second nodes, specifically including: Inputting keywords of the review information corresponding to the product risk control word into the risk control text model to determine at least one second node corresponding to the review information; wherein the keywords are words in the review information after removing stop words; According to the at least one second node and the commodity transaction sequence of the supply chain, a supply ray between the first node and the at least one second node is established to determine the supply relationship queue; wherein the supply ray points from the commodity supply node to the commodity purchase node.

5. The method according to claim 4, characterized in that: Input the keywords in the product evaluation information and the product risk control words into a pre-trained risk control text model to determine the supply relationship queue between the first node and several second nodes, specifically including: Determine, based on the transaction commodity information between the first node and the second node, the commodity event name corresponding to the transaction commodity information from a preset event comparison table; wherein the commodity event name includes at least commodity quality issues, regional epidemics, and company financial disruptions; Determine, based on a preset profiling model, several predicted events for the product risk control term and the probability of occurrence of each predicted event; wherein the predicted event is a preset malicious event that affects the stability of the supply chain; Determine the cosine similarity between each predicted event and each of the product event names, and identify the predicted event that meets a preset condition as a pending event, and determine the risk control value corresponding to the pending event; wherein the preset condition is that the cosine similarity is greater than a first preset value and the probability of occurrence of the predicted event is greater than a second preset value; A supply relationship queue between the first node and the plurality of second nodes is determined according to the risk control value corresponding to the pending event and the supply ray between the first node and the at least one second node.

6. The method according to claim 4, characterized in that: According to the supply chain, historical transaction data corresponding to the supply relationship queues whose risk control values ​​are greater than a first preset threshold are sequentially input into a preset convolutional neural network model to determine the risk control indicator of the second node corresponding to the supply relationship queue, specifically including: Determining, based on the commodity transaction sequence corresponding to the supply chain, historical transaction data corresponding to the supply relationship queue whose risk control value is greater than the first preset threshold; the historical transaction data includes commodity transaction time, transaction quantity, commodity type, and commodity production process; Inputting the historical transaction data into the convolutional neural network model to determine a number of risk control terms for each traded commodity and an inverse document frequency of each risk control term among the number of risk control terms; generating two risk control word sequences for the supply relationship queue according to the inverse document frequency and the risk value of each risk control word; The two risk control word sequences are de-duplicated and merged, so that the risk control word sequences after de-duplication and merging are used as the risk control indicators of the supply relationship queue.

7. The method according to claim 1, characterized in that: Determining the corresponding risk control risk level of the second node based on the risk control value, the risk control indicator, the historical transaction data, and a preset expert prediction model specifically includes: Inputting the historical transaction data and the risk control value into the expert prediction model, determining a transaction risk weight for each historical transaction in the historical transaction data, and updating the risk control value according to the transaction risk weight; According to the updated risk control value and the risk control index, assigning an index value to the risk control index through the expert prediction model to determine a risk index value corresponding to the risk control index; the risk index value is greater than 0 and not greater than 1; Determine the risk control risk level of each second node according to the risk control value and the risk indicator value.

8. The method according to claim 1, characterized in that: After determining the corresponding risk control level of the second node, the method further includes: Using network mapping software, a supply relationship diagram of the supply chain information is created, and the risk control risk level is marked on the relationship lines between the first node and each of the second nodes; According to each of the risk control risk levels, determining in the supply relationship diagram, a second node whose risk control risk level is greater than a third preset value as a node to be processed; Based on the historical transaction data of the first node and the node to be processed, the corresponding alternative node and subsequent transaction method of the node to be processed are determined to generate risk control processing information; wherein the commodity transaction quantity, commodity type and commodity production process of the alternative node match the historical transaction data of the node to be processed; the risk control processing information includes risk control processing objectives and risk control processing methods.

9. The method according to claim 8, characterized in that Generate prompt information based on the risk level of the risk control, and send the prompt information to the corresponding supervision terminal, specifically including: When the node to be processed is determined, the prompt information is generated and sent to the supervision terminal corresponding to the node to be processed; When receiving the supplementary information data fed back by the supervision terminal, sending the supplementary information data to the source terminal corresponding to the product evaluation information to determine whether the supplementary information data is valid; If so, determining the updated risk control risk level of the second node based on the preset risk control value, the risk control indicator, the historical transaction data, and the expert prediction model, and determining whether the second node is the node to be processed based on the updated risk control risk level, until the number of times the supplementary information data is obtained exceeds a fourth preset value, generating the risk control processing information and the prompt information, and sending them to the supervision terminals corresponding to the first node and the node to be processed respectively; Otherwise, determine the corresponding alternative node and subsequent transaction method of the node to be processed to generate risk control processing information, generate the prompt information, and send it to the supervision terminals corresponding to the first node and the node to be processed respectively.

10. A supply chain risk control processing device, characterized in that: The device comprises: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Acquire product evaluation information of a first node in the supply chain; wherein the product evaluation information includes at least: evaluation information directly related to the first node and evaluation information indirectly related to the first node; Based on a preset random forest algorithm, the product evaluation information is classified, and the classified product evaluation short text is determined as the product risk control word corresponding to the first node; the product evaluation short text is composed of a number of words that represent the product situation; Input the keywords in the product review information and the product risk control words into a pre-trained risk control text model to determine a supply relationship queue between the first node and a plurality of second nodes; wherein the supply relationship queue includes at least one second node information and its corresponding risk control value; Based on the supply chain information, historical transaction data corresponding to the supply relationship queues having risk control values ​​greater than a first preset threshold are sequentially input into a preset convolutional neural network model to determine the risk control indicator of the second node corresponding to the supply relationship queue; Based on the risk control value, the risk control indicator, the historical transaction data and the preset expert prediction model, the risk control risk level of the corresponding second node is determined, and prompt information is generated based on the risk control risk level, and the prompt information is sent to the corresponding supervision terminal.

Images