Risk and hidden danger management system and method based on big data

Abstract

The invention provides a big data-based risk and hidden danger management system and method, relates to the field of intelligent risk management, and adopts a big data-based analysis technology to analyze a financial statement of a to-be-assessed supplier to obtain financial health features of the supplier. Meanwhile, historical delivery time and historical delivery quality information of the to-be-evaluated supplier are analyzed to obtain supplier delivery condition semantic emphasis features, and the two features are considered to judge whether the to-be-evaluated supplier has a supply interruption risk. Therefore, whether the supplier has the supply interruption risk or not can be judged more accurately, and powerful support can be provided for supply chain risk hidden danger management decisions of enterprises.

Description

Technical Field

[0001] The present application relates to the field of intelligent risk management, and more specifically, to a risk hidden danger management system and method based on big data. Background Art

[0002] Supplier risk management refers to a series of management activities that enterprises carry out to identify, evaluate, monitor and respond to various risks that may arise from suppliers in order to ensure the stability and efficiency of the supply chain. In traditional supplier risk management, the use of data is relatively limited, usually simply checking a few key figures in financial statements or counting some basic data in delivery records, such as the number of deliveries and the number of delayed deliveries. The potential value of these data has not been fully explored, which leads to an in-depth understanding of the supplier's true operating conditions, which may lead to errors in the decision-making of supplier risk management.

[0003] Therefore, an optimized risk management solution is needed. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present application provides a risk hidden danger management system and method based on big data.

[0005] According to one aspect of the present application, a risk management system based on big data is proposed, which includes:

[0006] The data collection module for suppliers to be evaluated is used to obtain the financial statement data, historical delivery time and historical delivery quality information of suppliers to be evaluated;

[0007] A financial data analysis module for suppliers to be evaluated, used to analyze the financial statement data of the suppliers to be evaluated to obtain a characteristic vector of the financial health of the suppliers;

[0008] A delivery data analysis module for suppliers to be evaluated, used for analyzing the historical delivery time and historical delivery quality information of the suppliers to be evaluated to obtain a semantically emphasized feature vector of the suppliers' historical delivery conditions;

[0009] The financial delivery feature fusion module of the supplier to be evaluated is used to perform target domain scale modulation based on the intrinsic decomposition space on the feature vector of the supplier's financial health and the semantic emphasis feature vector of the supplier's historical delivery to obtain a supplier evaluation fusion feature vector;

[0010] The supplier risk identification result generation module is used to obtain the risk identification result of the supplier to be evaluated based on the supplier evaluation fusion feature vector.

[0011] In the above-mentioned risk management system based on big data, the financial data analysis module of the supplier to be evaluated includes: a financial data preprocessing unit of the supplier to be evaluated, which is used to preprocess the financial statement data of the supplier to be evaluated to obtain the preprocessed supplier financial statement content; a financial data feature extraction unit of the supplier to be evaluated, which is used to segment the preprocessed supplier financial statement content and then pass it through a financial statement content understander to obtain a feature vector of the supplier's financial health status.

[0012] In the above-mentioned risk management system based on big data, the delivery data analysis module of the supplier to be evaluated includes: a supplier historical delivery situation semantic feature extraction unit, which is used to extract semantic features of the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain a supplier historical delivery situation semantic feature matrix; a supplier historical delivery situation semantic feature emphasis unit, which is used to emphasize the semantic features of the supplier historical delivery situation semantic feature matrix to obtain the supplier historical delivery situation semantic emphasis feature vector.

[0013] In the above-mentioned risk management system based on big data, the supplier's historical delivery situation semantic feature extraction unit includes: a historical delivery situation semantic feature generation subunit, which is used to segment the historical delivery time and historical delivery quality information of the supplier to be evaluated and obtain multiple supplier historical delivery situation semantic feature vectors through a historical delivery situation semantic capture; a historical delivery situation semantic feature arrangement subunit, which is used to arrange the multiple supplier historical delivery situation semantic feature vectors in two dimensions to obtain the supplier's historical delivery situation semantic feature matrix.

[0014] In the above-mentioned risk management system based on big data, the supplier's historical delivery situation semantic feature emphasis unit includes: a historical delivery situation semantic emphasis feature generation subunit, used to pass the supplier's historical delivery situation semantic feature matrix through a historical delivery situation semantic feature emphasizer to obtain the supplier's historical delivery situation semantic emphasis feature matrix; a historical delivery situation semantic emphasis feature dimensionality reduction subunit, used to expand the supplier's historical delivery situation semantic emphasis feature matrix to obtain the supplier's historical delivery situation semantic emphasis feature vector.

[0015] In the above-mentioned risk management system based on big data, the financial statement content understander is a bidirectional long short-term memory neural network model, the historical delivery situation semantic capturer is a converter-based historical delivery situation semantic capturer, and the historical delivery situation semantic feature emphasizer is a convolutional neural network model using a spatial attention mechanism.

[0016] In the above-mentioned risk management system based on big data, the financial delivery feature fusion module of the supplier to be evaluated is used to: map the supplier financial health feature vector and the supplier historical delivery semantic emphasis feature vector to the intrinsic decomposition space to obtain the supplier financial health intrinsic feature vector and the supplier historical delivery semantic emphasis intrinsic feature vector; perform feature strength adjustment on the supplier financial health intrinsic feature vector and the supplier historical delivery semantic emphasis intrinsic feature vector based on the target domain mapping to obtain the supplier financial health modulated intrinsic feature vector and the supplier historical delivery semantic emphasis modulated intrinsic feature vector; construct a supplier evaluation fine-grained interactive semantic matrix between the supplier financial health modulated intrinsic feature vector and the supplier historical delivery semantic emphasis modulated intrinsic feature vector; and fuse the supplier financial health modulated intrinsic feature vector and the supplier historical delivery semantic emphasis modulated intrinsic feature vector based on the supplier evaluation fine-grained interactive semantic matrix to obtain the supplier evaluation fused feature vector.

[0017] In the above-mentioned risk management system based on big data, the supplier risk identification result generation module is used to: pass the supplier evaluation fusion feature vector through the supply risk classifier to obtain the risk identification result, and the risk identification result is used to indicate whether the supplier to be evaluated has a supply interruption risk.

[0018] According to another aspect of the present application, a risk management method based on big data is proposed, which includes:

[0019] Obtain financial statement data, historical delivery time and historical delivery quality information of the supplier to be evaluated;

[0020] Performing data analysis on the financial statement data of the supplier to be evaluated to obtain a characteristic vector of the supplier's financial health status;

[0021] Performing data analysis on the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain a semantically emphasized feature vector of the supplier's historical delivery situation;

[0022] Performing target domain scale modulation based on the intrinsic decomposition space on the supplier's financial health feature vector and the supplier's historical delivery semantic emphasis feature vector to obtain a supplier evaluation fusion feature vector;

[0023] Based on the supplier evaluation fusion feature vector, a risk identification result of the supplier to be evaluated is obtained.

[0024] In the above-mentioned risk management method based on big data, the risk identification result of the supplier to be evaluated is obtained based on the supplier evaluation fusion feature vector, including: passing the supplier evaluation fusion feature vector through a supply risk classifier to obtain the risk identification result, and the risk identification result is used to indicate whether the supplier to be evaluated has a supply interruption risk.

[0025] This application has significant technical effects due to the adoption of the above technical solutions:

[0026] The risk management system and method based on big data provided by this application adopts big data-based analysis technology, analyzes the financial statements of the supplier to be evaluated to obtain the supplier's financial health characteristics, and analyzes the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain the semantic emphasis characteristics of the supplier's delivery situation, and considers these two characteristics to determine whether the supplier to be evaluated has a supply interruption risk. In this way, it can be more accurately determined whether the supplier has a supply interruption risk, thereby providing strong support for the enterprise's supply chain risk management decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0028] Figure 1 This is a system block diagram of a risk management system based on big data according to an embodiment of the present application.

[0029] Figure 2 It is a block diagram of a financial data analysis module for suppliers to be evaluated in a risk management system based on big data according to an embodiment of the present application.

[0030] Figure 3 It is a block diagram of a delivery data analysis module of a supplier to be evaluated in a risk management system based on big data according to an embodiment of the present application.

[0031] Figure 4 It is a block diagram of a semantic feature extraction unit for supplier historical delivery status in a risk management system based on big data according to an embodiment of the present application.

[0032] Figure 5 It is a block diagram of a semantic feature emphasis unit for supplier historical delivery status in a risk management system based on big data according to an embodiment of the present application.

[0033] Figure 6 The present invention is a flowchart of a risk management method based on big data according to an embodiment of the present application. DETAILED DESCRIPTION

[0034] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0035] As mentioned in the above background technology, supplier risk management, as a key link in the enterprise supply chain management, aims to ensure the stable and efficient operation of the entire supply chain system. This management activity involves comprehensive and in-depth identification, accurate assessment, continuous monitoring and effective response to various risks that suppliers may encounter or cause. It is not only related to the smooth daily operation of the enterprise, but also directly affects the enterprise's cost control, product quality, market competitiveness and even the realization of long-term development strategy.

[0036] However, in the traditional supplier risk management model, the collection, analysis and use of data are often relatively extensive and limited. Enterprises usually rely only on some basic financial data, such as a few limited numbers such as net profit, operating income, debt-to-asset ratio in financial statements, and basic data in delivery records, such as the total number of deliveries, the number of delayed deliveries, and order completion rate, to understand the supplier's operating conditions. The potential value of these data has not been fully explored, which will lead to an inability to gain an in-depth understanding of the supplier's true operating conditions. This limitation may cause errors in the supplier's risk management decisions. Therefore, a risk management solution based on big data is expected.

[0037] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also demonstrated a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks provides new solutions and solutions for risk management.

[0038] Figure 1 : is a system block diagram of a risk hidden danger management system based on big data according to an embodiment of the present application. Figure 1As shown, in the risk hidden danger management system 100 based on big data, it includes: a data collection module 110 related to the supplier to be evaluated, which is used to obtain the financial statement data, historical delivery time and historical delivery quality information of the supplier to be evaluated; a financial data analysis module 120 for the supplier to be evaluated, which is used to perform data analysis on the financial statement data of the supplier to be evaluated to obtain the supplier's financial health feature vector; a delivery data analysis module 130 for the supplier to be evaluated, which is used to perform data analysis on the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain the supplier's historical delivery semantic emphasis feature vector; a financial delivery feature fusion module 140 for the supplier to be evaluated, which is used to perform target domain scale modulation based on the intrinsic decomposition space on the supplier's financial health feature vector and the supplier's historical delivery semantic emphasis feature vector to obtain a supplier evaluation fusion feature vector; a supplier risk identification result generation module 150, which is used to obtain the risk identification result of the supplier to be evaluated based on the supplier evaluation fusion feature vector.

[0039] In the embodiment of the present application, the data acquisition module 110 related to the supplier to be evaluated is used to obtain the financial statement data, historical delivery time and historical delivery quality information of the supplier to be evaluated. It should be understood that the financial statement data of the supplier to be evaluated includes the supplier's asset information, liability information, owner's equity information, income statement related information and cash flow statement related information. The historical delivery time information of the supplier to be evaluated includes historical delivery on-time rate information, delivery lead time information, delayed delivery reason record information, etc., and the historical delivery quality information includes product qualification rate information, quality problem type record information and quality improvement measure record information. Specifically, information such as assets, liabilities and cash flow in the financial statements can reflect whether the supplier's capital chain is stable. A stable capital chain is the basis for ensuring that suppliers can produce normally and deliver on time. If the supplier's capital chain is broken, for example, it cannot repay due debts or does not have enough funds to purchase raw materials, then the risk of supply interruption will be greatly increased. By analyzing the cost and operating income data in the profit statement, the supplier's cost control ability and profitability can be understood. If the supplier's costs are too high, the profits are meager or even loss-making, it may sacrifice product quality or postpone delivery time in order to reduce costs, thereby increasing the risk of supply interruption. Historical delivery time information is directly related to the reliability of supplier supply. Low on-time delivery rate and unstable delivery lead time will expose enterprises to the risk of production interruption. For example, in the just-in-time (JIT) production mode, the production plan of an enterprise is closely dependent on the on-time delivery of suppliers. Once suppliers frequently delay delivery, the entire production process of the enterprise will be seriously affected. Historical delivery quality information is an important basis for evaluating the quality of suppliers' products. Low product qualification rate or frequent quality problems will not only affect the production efficiency of the enterprise, but also lead to a series of problems such as product recalls and customer complaints, increasing the operating costs and reputation loss of the enterprise. Moreover, quality problems may be the external manifestation of deep-seated problems such as chaotic production management and imperfect quality control system of suppliers. If these problems are not solved, supply risks will exist for a long time. In general, obtaining these data can comprehensively evaluate suppliers from two major aspects: financial and operational. Financial status is the basic guarantee for the supplier's continued operation, while delivery time and quality are the direct reflection of the supplier's fulfillment of supply obligations. The combination of these data can more accurately determine whether the supplier has the risk of supply interruption, and thus provide strong support for the supply chain risk management of enterprises.

[0040] In the embodiment of the present application, the to-be-evaluated supplier financial data analysis module 120 is used to perform data analysis on the financial statement data of the to-be-evaluated supplier to obtain a characteristic vector of the supplier's financial health. Specifically, Figure 2 FIG. 1 is a block diagram of a financial data analysis module for suppliers to be evaluated in a risk management system based on big data according to an embodiment of the present application. Figure 2As shown, the supplier financial data analysis module 120 to be evaluated includes: a supplier financial data preprocessing unit 121 to be evaluated, used for preprocessing the financial statement data of the supplier to be evaluated to obtain the preprocessed supplier financial statement content; a supplier financial data feature extraction unit 122 to be evaluated, used for segmenting the preprocessed supplier financial statement content and then passing it through a financial statement content understander to obtain a feature vector of the supplier's financial health status.

[0041] In an embodiment of the present application, the financial data preprocessing unit 121 of the supplier to be evaluated is used to preprocess the financial statement data of the supplier to be evaluated to obtain the content of the supplier's financial statement after preprocessing. Accordingly, considering that the financial statement data structure is relatively complex, it contains a large number of tables, subjects and data items. For example, there are many accounting subjects in the balance sheet, income statement and cash flow statement, and there are complex cross-references between these subjects. Directly processing such a complex data structure is very difficult for subsequent analysis tools and models, and it is necessary to convert it into a form that is easier to understand and process. In addition, the original financial statement data may have various quality problems, such as the data may contain errors, omissions or inconsistencies, which will cause the original data to be unable to be directly used for effective analysis. Therefore, in order to improve the quality of the supplier's financial data and simplify the structural information of the data expression, it is necessary to preprocess the financial statement data of the supplier to be evaluated to obtain the content of the supplier's financial statement after preprocessing. After preprocessing steps such as error checking and correction, data cleaning, etc., the accuracy of the relevant data in the supplier's financial statement is improved, and the form of expression of the preprocessed financial statement content is more reasonable, which can be directly used to evaluate the financial health of the supplier, and enhance the availability of the data.

[0042] In an embodiment of the present application, the financial data feature extraction unit 122 of the supplier to be evaluated is used to segment the content of the pre-processed supplier financial statement and obtain the feature vector of the supplier's financial health through a financial statement content understander. It should be understood that the pre-processed financial statement content is in text form. For a computer, text is a continuous sequence of characters, and it is difficult to directly understand the semantic information therein. In order to facilitate the computer to better understand the semantics of financial data, it is first necessary to segment the content of the pre-processed supplier financial statement. The segmented vocabulary can more accurately represent the various concepts and indicators in the financial statements. For example, words such as "current ratio" and "accounts receivable turnover rate" have clear meanings in financial analysis. Through word segmentation, these important financial concepts can be separated, providing a basis for the subsequent extraction of financial health feature vectors. If word segmentation is not performed, these valuable information will be submerged in the continuous text and difficult to be effectively mined and utilized. Then, considering that only the keywords after word segmentation are only superficial vocabulary information, there are complex semantic relationships between these words. In order to deeply explore the relationship between these keywords and understand their true meaning in the context of financial statements, it is necessary to use a financial statement content understander to semantically encode the multiple supplier financial statement content keywords obtained by word segmentation. In this application, the financial statement content understander specifically refers to a bidirectional long short-term memory neural network model. The bidirectional long short-term memory neural network (Bi-LSTM) consists of a forward LSTM and a backward LSTM. The forward LSTM processes the sequence of financial statement keywords after word segmentation in the order from the beginning to the end of the text to capture positive semantic information; the backward LSTM processes from the end of the text to the beginning to capture reverse semantic information. Due to the bidirectional structure, Bi-LSTM can capture the contextual semantic information of financial statement keywords, which is conducive to understanding the complex semantic relationships in financial statements. For example, when analyzing a sentence like “Although the accounts receivable turnover rate has decreased, the company expects future revenue to increase by strengthening sales channel management,” the two-way structure can consider both the unfavorable factor of “decreased accounts receivable turnover rate” and the favorable factor of “expected future revenue to increase,” thereby more accurately assessing the supplier’s financial health.

[0043] In the embodiment of the present application, the delivery data analysis module 130 of the supplier to be evaluated is used to perform data analysis on the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain a semantically emphasized feature vector of the supplier's historical delivery situation. Specifically, Figure 3 FIG. 1 is a block diagram of a delivery data analysis module of a supplier to be evaluated in a risk management system based on big data according to an embodiment of the present application. Figure 3As shown, the supplier delivery data analysis module 130 to be evaluated includes: a supplier historical delivery situation semantic feature extraction unit 131, which is used to extract semantic features of the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain a supplier historical delivery situation semantic feature matrix; a supplier historical delivery situation semantic feature emphasis unit 132, which is used to emphasize the semantic features of the supplier historical delivery situation semantic feature matrix to obtain the supplier historical delivery situation semantic emphasis feature vector.

[0044] In the embodiment of the present application, the supplier historical delivery semantic feature extraction unit 131 is used to extract semantic features from the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain the supplier historical delivery semantic feature matrix. Specifically, Figure 4 FIG. 1 is a block diagram of a semantic feature extraction unit for supplier historical delivery status in a risk management system based on big data according to an embodiment of the present application. Figure 4 As shown, the supplier historical delivery situation semantic feature extraction unit 131 includes: a historical delivery situation semantic feature generation subunit 1311, which is used to segment the historical delivery time and historical delivery quality information of the supplier to be evaluated and obtain multiple supplier historical delivery situation semantic feature vectors through a historical delivery situation semantic capturer; a historical delivery situation semantic feature arrangement subunit 1312, which is used to arrange the multiple supplier historical delivery situation semantic feature vectors in two dimensions to obtain the supplier historical delivery situation semantic feature matrix.

[0045] In an embodiment of the present application, the historical delivery situation semantic feature generation subunit 1311 is used to segment the historical delivery time and historical delivery quality information of the supplier to be evaluated and obtain multiple supplier historical delivery situation semantic feature vectors through the historical delivery situation semantic capturer. It should be understood that the historical delivery time and quality information are usually a continuous time series data, containing a large number of different types of details. For example, the delivery time information may cover the delivery date of different orders, the delivery lead time of different products, and various delayed delivery situations; the delivery quality information may include the qualified rate of multiple products, the frequency and time of different quality problems, etc. If these information are not segmented, they will be too messy and difficult to perform effective semantic analysis. Therefore, it is first necessary to segment the historical delivery time and historical delivery quality information of the supplier to be evaluated. Through segmentation, the local features in each data segment can be highlighted to avoid the features of the overall data being averaged or masked. For example, for the supplier to be evaluated, its overall on-time delivery rate may look good, but through segmentation, it is found that the on-time delivery rate is low in a specific market area or for orders from certain large customers, which is a local feature that needs to be focused on. Next, a historical delivery situation semantic capturer is needed to perform semantic encoding processing on the segmented historical delivery time and historical delivery quality information. The historical delivery situation semantic capturer in this application specifically refers to a historical delivery situation semantic capturer based on a converter. Those of ordinary skill in the art should understand that the converter structure can calculate the correlation between different positions in the input data through multiple parallel attention heads. In the semantic capture of historical delivery situations, each attention head can focus on different aspects of the input segmented data. For example, one attention head may focus on the relationship between delivery time and product category, and another attention head may focus on the relationship between quality problem type and time period. The multi-head attention mechanism can capture semantic information in the data from multiple angles at the same time, avoiding the limitation that traditional models can only analyze from a single angle. Through this data processing method, the semantic characteristics of the delivery time and quality information and the correlation characteristics between the two information can be accurately captured.

[0046] In an embodiment of the present application, the historical delivery situation semantic feature arrangement subunit 1312 is used to arrange the multiple suppliers' historical delivery situation semantic feature vectors in two dimensions to obtain the supplier's historical delivery situation semantic feature matrix. It should be understood that each of the multiple suppliers' historical delivery situation semantic feature vectors may focus on a specific dimension of the delivery situation, such as delivery punctuality, delivery quality stability, quality problem type distribution, etc. Although each vector contains valuable information when it exists alone, it is relatively scattered. In order to further integrate and utilize this information, it is necessary to arrange the multiple suppliers' historical delivery situation semantic feature vectors in two dimensions to obtain the supplier's historical delivery situation semantic feature matrix in the technical solution of the present application. The two-dimensionally arranged matrix can better support the analysis of the interaction between different features, thereby more deeply exploring the hidden information in the supplier's historical delivery situation, so as to provide more powerful support for the comprehensive assessment of supplier supply risks.

[0047] In the embodiment of the present application, the supplier historical delivery situation semantic feature emphasis unit 132 is used to perform semantic feature emphasis on the supplier historical delivery situation semantic feature matrix to obtain the supplier historical delivery situation semantic emphasis feature vector. Specifically, Figure 5 FIG. 1 is a block diagram of a semantic feature emphasis unit for supplier historical delivery status in a risk management system based on big data according to an embodiment of the present application. Figure 5 As shown, the supplier's historical delivery situation semantic feature emphasis unit 132 includes: a historical delivery situation semantic emphasis feature generation subunit 1321, used for passing the supplier's historical delivery situation semantic feature matrix through a historical delivery situation semantic feature emphasizer to obtain the supplier's historical delivery situation semantic emphasis feature matrix; a historical delivery situation semantic emphasis feature dimension reduction subunit 1322, used for expanding the supplier's historical delivery situation semantic emphasis feature matrix to obtain the supplier's historical delivery situation semantic emphasis feature vector.

[0048] In an embodiment of the present application, the historical delivery situation semantic emphasis feature generation subunit 1321 is used to pass the supplier's historical delivery situation semantic feature matrix through a historical delivery situation semantic feature emphasizer to obtain a supplier's historical delivery situation semantic emphasis feature matrix. It should be understood that although a feature matrix containing rich semantic information has been obtained through the previous steps, the importance of different features in assessing the risk of supplier supply disruption may be different. Some features may be closely related to the risk of supply disruption, while others may be only secondary factors. In order to emphasize these key, highly risk-related features and weaken those relatively unimportant information so that subsequent risk assessments can focus on the most valuable features, the technical solution of the present application needs to pass the supplier's historical delivery situation semantic feature matrix through a historical delivery situation semantic feature emphasizer to obtain a supplier's historical delivery situation semantic emphasis feature matrix. The historical delivery situation semantic feature emphasizer in the present application specifically refers to a convolutional neural network model using a spatial attention mechanism. Specifically, the core of the convolutional neural network model is the convolution layer. In this model, the convolution layer slides on the semantic feature matrix through the convolution kernel and performs convolution operations on the local area in the matrix; the spatial attention mechanism is the key part of this model for emphasizing features. It calculates the attention weight of each position on the feature map output by the convolution layer to determine the importance of different positions in the feature map; the pooling layer is usually used to downsample the feature map to reduce the data dimension while retaining the main feature information; the fully connected layer can output the final semantic emphasis feature matrix of the supplier's historical delivery situation based on the previously emphasized features. The elements in this matrix have been weighted by the spatial attention mechanism and processed by other neural network layers, and are more focused on reflecting the key information related to the risk of supply interruption. By performing this step of processing, the spatial relationship between features can be effectively captured, that is, the spatial interaction and correlation of different delivery situation features, which helps to understand the overall pattern of the supplier's historical delivery situation.

[0049] In an embodiment of the present application, the historical delivery situation semantic emphasis feature dimensionality reduction subunit 1322 is used to expand the supplier's historical delivery situation semantic emphasis feature matrix to obtain the supplier's historical delivery situation semantic emphasis feature vector. It should be understood that compared with a two-dimensional matrix, a vector is more convenient for performing various calculation operations. In order to facilitate calculations while retaining the original feature information, it is necessary to expand the supplier's historical delivery situation semantic emphasis feature matrix. The supplier's historical delivery situation semantic emphasis feature vector obtained after expansion is a reflection of the supplier's delivery situation characteristics as a whole.

[0050] In the embodiment of the present application, the financial delivery feature fusion module 140 of the supplier to be evaluated is used to modulate the target domain scale of the supplier's financial health feature vector and the supplier's historical delivery semantic emphasis feature vector based on the intrinsic decomposition space to obtain the supplier evaluation fusion feature vector. It should be understood that the supplier's financial health and historical delivery are two key dimensions for evaluating its supply capacity and supply quality. Financial health reflects the supplier's economic strength, financial stability and long-term operational sustainability. For example, a good financial situation means that the supplier has sufficient funds for raw material procurement, production equipment maintenance and employee salary payment, which is the basis for ensuring stable supply. The historical delivery situation directly reflects the supplier's supply performance in the actual operation process, including whether the delivery is on time and whether the delivery quality is qualified. It is not comprehensive to rely solely on information from a single dimension for evaluation. Therefore, in order to comprehensively evaluate the risk of supplier supply interruption, the supplier's financial health feature vector and the supplier's historical delivery semantic emphasis feature vector need to be integrated in the technical solution of the present application.

[0051] In particular, considering that the supplier's financial health feature vector mainly comes from financial statement data, these data are quantitative and structured, reflecting the company's financial status and operating efficiency; while the historical delivery semantic emphasis feature vector comes from the supplier's historical delivery time and quality information, these data are usually qualitative and unstructured, and need to be extracted through semantic analysis; the different data sources and natures result in the processed supplier's financial health feature vector and the supplier's historical delivery semantic emphasis feature vector being distributed in different feature spaces. Since these two feature vectors come from different feature spaces, direct fusion may not be able to fully capture the complex relationship between them, which may lead to inaccurate final supplier risk assessment results, and thus affect the company's risk management decisions. Therefore, in the technical solution of the present application, it is necessary to perform target domain scale modulation based on the intrinsic decomposition space on the supplier's financial health feature vector and the supplier's historical delivery semantic emphasis feature vector to obtain the supplier evaluation fusion feature vector.

[0052] Specifically, in an embodiment of the present application, the supplier financial health feature vector and the supplier historical delivery semantic emphasis feature vector are modulated on a target domain scale based on an intrinsic decomposition space to obtain a supplier evaluation fusion feature vector, including: mapping the supplier financial health feature vector and the supplier historical delivery semantic emphasis feature vector to an intrinsic decomposition space to obtain a supplier financial health intrinsic feature vector and a supplier historical delivery semantic emphasis intrinsic feature vector; performing feature strength adjustment based on target domain mapping on the supplier financial health intrinsic feature vector and the supplier historical delivery semantic emphasis intrinsic feature vector to obtain a modulated supplier financial health intrinsic feature vector and a modulated supplier historical delivery semantic emphasis intrinsic feature vector; constructing a supplier evaluation fine-grained interaction semantic matrix between the supplier financial health modulated intrinsic feature vector and the supplier historical delivery semantic emphasis modulated intrinsic feature vector; and fusing the supplier financial health modulated intrinsic feature vector and the supplier historical delivery semantic emphasis modulated intrinsic feature vector based on the supplier evaluation fine-grained interaction semantic matrix to obtain the supplier evaluation fusion feature vector.

[0053] More specifically, in an embodiment of the present application, a supplier evaluation fine-grained interaction semantic matrix is ​​constructed between the supplier's financial health modulated intrinsic feature vector and the supplier's historical delivery situation semantically emphasized modulated intrinsic feature vector, including: multiplying the transposed vectors of the supplier's financial health modulated intrinsic feature vector and the supplier's historical delivery situation semantically emphasized modulated intrinsic feature vector and then dividing the result by the scale of the supplier's historical delivery situation semantically emphasized modulated intrinsic feature vector to obtain a supplier evaluation fine-grained interaction matrix; performing a convolution coding operation based on a convolution kernel of 3×3 on the supplier evaluation fine-grained interaction matrix to obtain the supplier evaluation fine-grained interaction semantic matrix.

[0054] More specifically, in an embodiment of the present application, the supplier financial health status modulated eigenfeature vector and the supplier historical delivery status semantically emphasized modulated eigenfeature vector are fused based on the supplier evaluation fine-grained interactive semantic matrix to obtain a supplier evaluation fused semantic feature vector, including: multiplying the supplier evaluation fine-grained interactive semantic matrix by the supplier financial health status modulated eigenfeature vector to obtain the supplier financial health status feature vector after information field mapping; multiplying the supplier evaluation fine-grained interactive semantic matrix by the supplier historical delivery status semantically emphasized modulated eigenfeature vector to obtain the supplier historical delivery status semantically emphasized feature vector after information field mapping; weightedly fusing the supplier financial health status feature vector after information field mapping and the supplier historical delivery status semantically emphasized feature vector after information field mapping to obtain the supplier evaluation fused semantic feature vector.

[0055] In an embodiment of the present application, the target domain scale modulation based on the intrinsic decomposition space is performed on the supplier financial health feature vector and the supplier historical delivery semantic emphasis feature vector to obtain the supplier evaluation fusion feature vector, including: processing the supplier financial health feature vector and the supplier historical delivery semantic emphasis feature vector according to the following formula to obtain the supplier evaluation fusion feature vector; wherein the formula is:

[0056]

[0057]

[0058]

[0059]

[0060]

[0061]

[0062]

[0063]

[0064] in, Represents the characteristic vector of supplier financial health, The semantically emphasized feature vector represents the supplier's historical delivery status. represents principal component analysis, A sequence of eigendecomposition vectors representing the financial health of suppliers, represents the diagonal matrix of supplier financial health, and Represents the first and second positions of the diagonal matrix of the supplier's financial health status. The eigenvalues ​​at the positions, represents the transpose of a vector, represents the vector concatenation function, , , The first, second and third sequences of the eigendecomposition vectors of the supplier’s financial health status are represented by feature vectors, The intrinsic feature vector representing the financial health of the supplier, The semantically emphasizing sequence of intrinsic decomposition vectors representing the supplier's historical delivery status, represents the diagonal matrix of supplier financial health, and The first and second positions of the diagonal matrix are respectively the historical delivery status of the supplier. The eigenvalues ​​at the positions, , The first, second and third vectors of the sequence of semantically emphasized eigendecomposition vectors represent the historical delivery status of suppliers. feature vectors, The semantics of the historical delivery status of the supplier emphasizes the intrinsic feature vector, The first eigenvector representing the supplier's financial health eigenvalues, The semantics of the intrinsic feature vector representing the supplier's historical delivery status is emphasized. eigenvalues, represents the vector two norm, represents the natural exponential function, represents the modulated intrinsic eigenvector of the supplier’s financial health, The semantically emphasized modulated intrinsic feature vector represents the supplier's historical delivery situation. represents matrix multiplication, represents the scale of the vector, represents the fine-grained interaction matrix for supplier evaluation, Express Perform convolutional coding operation based on a convolution kernel of 3×3. represents the fine-grained interaction semantic matrix of supplier evaluation, Represents the characteristic vector of the supplier's financial health after information field mapping, The semantically emphasized feature vector represents the supplier's historical delivery situation after information field mapping. and represents the weighted hyperparameter, Represents the supplier evaluation fusion feature vector.

[0065] That is, in the technical solution of the present application, the target domain scale modulation based on the intrinsic decomposition space is performed on the characteristic vector of the supplier's financial health and the characteristic vector of the semantic emphasis of the supplier's historical delivery. The process first maps the characteristic vector of the supplier's financial health and the characteristic vector of the semantic emphasis of the supplier's historical delivery to the intrinsic decomposition space to obtain the intrinsic characteristic vector of the supplier's financial health and the characteristic vector of the semantic emphasis of the supplier's historical delivery. A person of ordinary skill in the art should know that the process of mapping to the intrinsic decomposition space is a combination of dimensionality reduction and feature extraction. In a specific example, the eigenvalue decomposition technique of principal component analysis (PCA) is used to find the intrinsic decomposition space, and the intrinsic decomposition space mapping is performed. The mapped intrinsic characteristic vector of the supplier's financial health and the characteristic vector of the semantic emphasis of the supplier's historical delivery not only represent the projection of the original characteristic vector in the new coordinate system, but also reflect the main direction of change of the data, which helps to avoid the "dimensionality disaster", that is, the problem that the sample density drops sharply as the number of features increases.

[0066] Next, after mapping the feature vector to the intrinsic decomposition space, it is necessary to adjust the feature strength of the supplier's financial health intrinsic feature vector and the supplier's historical delivery semantically emphasized intrinsic feature vector based on the target domain mapping to adjust the source domain feature vector to better match the data distribution of the target domain. This operation is intended to capture and compensate for the systematic deviation between the two domains to reduce the hidden disturbance of the class expression of the local feature distribution of the supplier's financial health intrinsic feature vector and the supplier's historical delivery semantically emphasized intrinsic feature vector to the overall class expression. After modulation, the modulated intrinsic feature vector of the supplier's financial health and the modulated intrinsic feature vector of the supplier's historical delivery semantically emphasized can retain the characteristics of the source domain while being more in line with the actual needs of the target domain, thereby enhancing the model's understanding and application of the target domain data.

[0067] After completing the scale modulation of the feature vector, a supplier evaluation fine-grained interactive semantic matrix is ​​constructed between the intrinsic feature vector after the supplier's financial health status is modulated and the intrinsic feature vector after the supplier's historical delivery status is modulated with semantic emphasis. The supplier evaluation fine-grained interactive semantic matrix is ​​used to describe a more detailed relationship between feature vectors. The introduction of the supplier evaluation fine-grained interactive semantic matrix enables the model to go beyond simple point-to-point comparison and consider more complex patterns and contextual information, thereby enhancing the model's sensitivity to subtle differences and improving the quality of feature representation. Finally, the supplier evaluation fused feature vector is obtained by fusion of the intrinsic feature vector after the supplier's financial health status is modulated and the intrinsic feature vector after the supplier's historical delivery status is modulated with semantic emphasis based on the supplier evaluation fine-grained interactive semantic matrix.

[0068] In an embodiment of the present application, the supplier risk identification result generation module 150 is used to obtain the risk identification result of the supplier to be evaluated based on the supplier evaluation fusion feature vector. Specifically, in an embodiment of the present application, the supplier risk identification result generation module is used to: pass the supplier evaluation fusion feature vector through the supply risk classifier to obtain the risk identification result, and the risk identification result is used to indicate whether the supplier to be evaluated has a supply interruption risk. It should be understood that the supplier evaluation fusion feature vector is relatively complex and abstract. In order to convert complex feature information into a clear judgment result, that is, to tell the enterprise whether the supplier has a supply interruption risk, so as to provide strong support for the enterprise's supplier management decision, it is necessary to input the supplier evaluation fusion feature vector into the supply risk classifier for classification processing. The supply risk classifier is constructed based on a machine learning algorithm. Through a large amount of data training, it can capture the mapping relationship between the supplier evaluation fusion features and the risk identification results. When a new supplier evaluation fusion feature vector is input, a specific output judgment can be made. The risk identification result directly indicates whether the supplier to be evaluated has a supply interruption risk. For existing suppliers, if the risk identification results show that there is a risk of supply disruption, the company can consider adjusting the mode of cooperation, such as reducing order volume, increasing supervision, or finding alternative suppliers; if there is no risk of supply disruption, the company can continue to maintain stable cooperation and even consider expanding the scope of cooperation.

[0069] In summary, the risk management system 100 based on big data according to the embodiment of the present application is explained, which adopts the analysis technology based on big data, analyzes the financial statements of the supplier to be evaluated to obtain the financial health characteristics of the supplier, and analyzes the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain the semantic emphasis characteristics of the supplier's delivery situation, and considers these two characteristics to determine whether the supplier to be evaluated has a supply interruption risk. In this way, it can be more accurately determined whether the supplier has a supply interruption risk, thereby providing strong support for the enterprise's supply chain risk management decision.

[0070] Figure 6 Flow chart of the risk hidden danger management method based on big data according to an embodiment of the present application. Figure 6As shown, in the risk hidden danger management method based on big data, it includes: S110, obtaining the financial statement data, historical delivery time and historical delivery quality information of the supplier to be evaluated; S120, performing data analysis on the financial statement data of the supplier to be evaluated to obtain the supplier's financial health feature vector; S130, performing data analysis on the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain the supplier's historical delivery semantic emphasis feature vector; S140, performing target domain scale modulation based on the intrinsic decomposition space on the supplier's financial health feature vector and the supplier's historical delivery semantic emphasis feature vector to obtain a supplier evaluation fusion feature vector; S150, obtaining the risk identification result of the supplier to be evaluated based on the supplier evaluation fusion feature vector.

[0071] Here, those skilled in the art can understand that the specific operations of each step in the above-mentioned risk management method based on big data have been referred to above. Figures 1 to 5 The description of the risk hidden danger management system based on big data has been introduced in detail, and therefore, its repeated description will be omitted.

[0072] In summary, the risk management method based on big data according to the embodiment of the present application is explained, which adopts the analysis technology based on big data, analyzes the financial statements of the supplier to be evaluated to obtain the financial health characteristics of the supplier, and analyzes the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain the semantic emphasis characteristics of the supplier's delivery situation, and considers these two characteristics to determine whether the supplier to be evaluated has a supply interruption risk. In this way, it is possible to more accurately determine whether the supplier has a supply interruption risk, thereby providing strong support for the enterprise's supply chain risk management decision.

Claims

1. A risk management system based on big data, characterized in that: include: The data collection module for suppliers to be evaluated is used to obtain the financial statement data, historical delivery time and historical delivery quality information of suppliers to be evaluated; A financial data analysis module for suppliers to be evaluated, used to analyze the financial statement data of the suppliers to be evaluated to obtain a characteristic vector of the financial health of the suppliers; A delivery data analysis module for suppliers to be evaluated, used for analyzing the historical delivery time and historical delivery quality information of the suppliers to be evaluated to obtain a semantically emphasized feature vector of the suppliers' historical delivery conditions; The financial delivery feature fusion module of the supplier to be evaluated is used to perform target domain scale modulation based on the intrinsic decomposition space on the feature vector of the supplier's financial health and the semantic emphasis feature vector of the supplier's historical delivery to obtain a supplier evaluation fusion feature vector; The supplier risk identification result generation module is used to obtain the risk identification result of the supplier to be evaluated based on the supplier evaluation fusion feature vector.

2. The risk hidden danger management system based on big data according to claim 1 is characterized in that: The financial data analysis module of the supplier to be evaluated includes: A financial data preprocessing unit for suppliers to be evaluated, used for preprocessing the financial statement data of the suppliers to be evaluated to obtain the content of the preprocessed financial statements of the suppliers; The financial data feature extraction unit of the supplier to be evaluated is used to segment the content of the pre-processed supplier financial report and pass it through a financial report content understander to obtain a feature vector of the supplier's financial health status.

3. The risk hidden danger management system based on big data according to claim 2 is characterized in that: The delivery data analysis module of the supplier to be evaluated includes: A supplier historical delivery semantic feature extraction unit, used for extracting semantic features from the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain a supplier historical delivery semantic feature matrix; The supplier's historical delivery situation semantic feature emphasis unit is used to perform semantic feature emphasis on the supplier's historical delivery situation semantic feature matrix to obtain the supplier's historical delivery situation semantic feature emphasis vector.

4. The risk hidden danger management system based on big data according to claim 3 is characterized in that: The supplier's historical delivery situation semantic feature extraction unit includes: A historical delivery situation semantic feature generation subunit is used to process the historical delivery time and historical delivery quality information of the supplier to be evaluated in segments and then pass them through a historical delivery situation semantic capturer to obtain multiple suppliers' historical delivery situation semantic feature vectors; The historical delivery situation semantic feature arrangement subunit is used to arrange the multiple suppliers' historical delivery situation semantic feature vectors in two dimensions to obtain the suppliers' historical delivery situation semantic feature matrix.

5. The risk hidden danger management system based on big data according to claim 4 is characterized in that: The supplier's historical delivery situation semantic feature emphasis unit includes: A historical delivery situation semantic emphasis feature generation subunit, used for passing the supplier's historical delivery situation semantic feature matrix through a historical delivery situation semantic feature emphasizer to obtain a supplier's historical delivery situation semantic emphasis feature matrix; The historical delivery situation semantic emphasis feature dimensionality reduction subunit is used to expand the supplier's historical delivery situation semantic emphasis feature matrix to obtain the supplier's historical delivery situation semantic emphasis feature vector.

6. The risk hidden danger management system based on big data according to claim 5 is characterized in that: The financial statement content understander is a bidirectional long short-term memory neural network model, the historical delivery situation semantic capturer is a converter-based historical delivery situation semantic capturer, and the historical delivery situation semantic feature emphasizer is a convolutional neural network model using a spatial attention mechanism.

7. The risk hidden danger management system based on big data according to claim 6 is characterized in that: The module for integrating the financial delivery characteristics of the supplier to be evaluated is used to: Mapping the supplier financial health feature vector and the supplier historical delivery semantic emphasis feature vector to the intrinsic decomposition space to obtain the supplier financial health intrinsic feature vector and the supplier historical delivery semantic emphasis intrinsic feature vector; Performing feature strength adjustment on the supplier's financial health status intrinsic feature vector and the supplier's historical delivery status semantically emphasized intrinsic feature vector based on target domain mapping to obtain the supplier's financial health status modulated intrinsic feature vector and the supplier's historical delivery status semantically emphasized modulated intrinsic feature vector; Constructing a supplier evaluation fine-grained interaction semantic matrix between the intrinsic feature vector after modulation of the supplier's financial health and the intrinsic feature vector after modulation of the supplier's historical delivery situation with semantic emphasis; The supplier evaluation fused feature vector is obtained by fusing the supplier financial health status modulated intrinsic feature vector and the supplier historical delivery status semantically emphasized modulated intrinsic feature vector based on the supplier evaluation fine-grained interactive semantic matrix.

8. The risk hidden danger management system based on big data according to claim 7 is characterized in that: The supplier risk identification result generating module is used to: pass the supplier evaluation fusion feature vector through the supply risk classifier to obtain the risk identification result, and the risk identification result is used to indicate whether the supplier to be evaluated has a supply interruption risk.

9. A risk management method based on big data, characterized in that: include: Obtain financial statement data, historical delivery time and historical delivery quality information of the supplier to be evaluated; Performing data analysis on the financial statement data of the supplier to be evaluated to obtain a characteristic vector of the supplier's financial health status; Performing data analysis on the historical delivery time and historical delivery quality information of the supplier to be evaluated to obtain a semantically emphasized feature vector of the supplier's historical delivery situation; Performing target domain scale modulation based on the intrinsic decomposition space on the supplier's financial health feature vector and the supplier's historical delivery semantic emphasis feature vector to obtain a supplier evaluation fusion feature vector; Based on the supplier evaluation fusion feature vector, a risk identification result of the supplier to be evaluated is obtained.

10. The risk hidden danger management method based on big data according to claim 9 is characterized in that: Based on the supplier evaluation fusion feature vector, a risk identification result of the supplier to be evaluated is obtained, including: passing the supplier evaluation fusion feature vector through a supply risk classifier to obtain the risk identification result, and the risk identification result is used to indicate whether the supplier to be evaluated has a supply interruption risk.

Images