A financial risk assessment and prediction method and system for the automotive industry chain

Abstract

The present invention relates to the field of knowledge graph technology, and in particular, to a method and system for evaluating and predicting financial risks in the automotive industry chain. The method comprises obtaining a first indicator of the central enterprise at each moment; calculating the financial risk of the central enterprise at each moment according to the first indicator and weight; obtaining a second indicator of the upstream enterprise at each moment; based on the graph convolutional neural network and the fully connected layer, according to the second indicator; using a time series analysis model, according to the financial risk of the central enterprise at each moment, and the financial risk of the upstream enterprise at each moment, predicting the financial risk of each enterprise at a future moment. The present invention can accurately evaluate the financial risk of an enterprise based on a knowledge graph and an intelligent model.

Description

Technical Field

[0001] The present invention relates to the field of knowledge graph technology, and in particular to a method and system for evaluating and predicting financial risks in an automotive industry chain. Background Art

[0002] The automobile industry has a large economic scale and a wide range of impact. The upstream and downstream cooperation of the automobile industry chain is strong, the effect of industrial and financial integration is obvious, and the demand for funds is large. It is an important source of large-scale growth in bank corporate loans. The financial risks of automobile companies are important assessment objects.

[0003] At present, financial institutions evaluate the investment and financing services of enterprises in the automotive industry chain, but they face the following problems: the on-site inspection method is inefficient and lacks multi-source data to verify and supplement the enterprise situation, and the evaluation and prediction capabilities for the automotive industry are insufficient. At present, there is a lack of investment and financing analysis products for the automotive industry on the market, and financial institutions lack professional knowledge. The evaluation and risk prediction of emerging enterprises are prone to deviations, resulting in funds taking on greater financial risks or not flowing to enterprises with real development prospects.

[0004] In view of this, the present invention is proposed. Summary of the invention

[0005] The purpose of the present invention is to provide a method and system for evaluating and predicting the financial risks of the automotive industry chain, so as to accurately assess the financial risks of enterprises based on knowledge graphs and intelligent models.

[0006] In order to achieve the above object, the present invention adopts the following technical solutions:

[0007] In a first aspect, the present invention provides a method for evaluating and predicting financial risks in an automobile industry chain, comprising:

[0008] According to the financial and supply relationships between enterprises in the automotive industry chain, a first knowledge graph is constructed, wherein entities in the first knowledge graph represent enterprises, and edges represent financial and supply relationships;

[0009] According to the structural characteristics of the first knowledge graph, multiple core enterprises and upstream enterprises of each core enterprise are selected from the first knowledge graph;

[0010] Obtaining the first indicator of the central enterprise at each moment; calculating the financial risk of the central enterprise at each moment according to the first indicator and the weight; the first indicator includes external environment, enterprise quality and public opinion information;

[0011] Obtaining a second indicator of the upstream enterprise at each moment; based on the graph convolutional neural network and the fully connected layer, calculating the financial risk of the upstream enterprise at each moment according to the second indicator; the second indicator includes the external environment, enterprise quality, public opinion information, the impact factor of the upstream enterprise affected by the connected central enterprise, the operation level of the industrial chain where the upstream enterprise is located, and the financial risks of all central enterprises connected to the upstream enterprise;

[0012] Using a time series analysis model, based on the financial risk of the central enterprise at each moment and the financial risk of the upstream enterprise at each moment, the financial risk of each enterprise at a future moment is predicted;

[0013] Wherein, based on the graph convolutional neural network and the fully connected layer, according to the second indicator, calculating the financial risk of the upstream enterprise at each moment includes:

[0014] Determine the financial risk at each historical moment based on the repayment overdue situation of upstream enterprises at each historical moment provided by financial institutions;

[0015] quantifying the second indicator;

[0016] Training the graph convolutional neural network and the fully connected layer according to the second indicator and financial risk at each historical moment;

[0017] The second indicator at the current moment is input into the graph convolutional neural network and the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment.

[0018] In a second aspect, the present invention provides a financial risk assessment and prediction system for the automotive industry chain, including: a data layer, a technical layer, and a service layer;

[0019] The data layer is used to store the first indicator and the second indicator at each moment;

[0020] The technical layer is used to construct a first knowledge graph based on the financial and supply relationships between enterprises in the automotive industry chain, wherein entities in the first knowledge graph represent enterprises and edges represent financial and supply relationships; based on the structural characteristics of the first knowledge graph, multiple central enterprises and upstream enterprises of each central enterprise are selected from the first knowledge graph;

[0021] The service layer is used to obtain the first indicator of the central enterprise at each moment; calculate the financial risk of the central enterprise at each moment according to the first indicator and the weight; the first indicator includes the external environment, enterprise quality and public opinion information; obtain the second indicator of the upstream enterprise at each moment; based on the graph convolutional neural network and the fully connected layer, calculate the financial risk of the upstream enterprise at each moment according to the second indicator; the second indicator includes the external environment, enterprise quality, public opinion information, the impact factor of the upstream enterprise affected by the connected central enterprise, the operation level of the industrial chain where the upstream enterprise is located, and the financial risks of all central enterprises connected to the upstream enterprise; use the time series analysis model to predict the financial risk of each enterprise at a future moment according to the financial risk of the central enterprise at each moment and the financial risk of the upstream enterprise at each moment;

[0022] Wherein, based on the graph convolutional neural network and the fully connected layer, according to the second indicator, the financial risk of the upstream enterprise at each moment is calculated, including: determining the financial risk at each historical moment according to the overdue repayment situation of the upstream enterprise at each historical moment provided by the financial institution; quantifying the second indicator; training the graph convolutional neural network and the fully connected layer according to the second indicator and the financial risk at each historical moment; inputting the second indicator at the current moment into the graph convolutional neural network and the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment.

[0023] Compared with the prior art, the present invention has the following beneficial effects:

[0024] (1) The present invention divides enterprises into core enterprises and upstream enterprises based on the financial and supply relationships between enterprises, thereby adopting different methods to evaluate financial risks according to the characteristics of the enterprises, thereby obtaining financial risks that are more in line with the characteristics of the enterprises.

[0025] (2) The present invention adopts a variety of first indicators and second indicators to calculate the financial risks of enterprises from multiple dimensions.

[0026] (3) The present invention adopts a time series analysis model to predict the financial risk of each enterprise at a future moment based on the financial risk of the central enterprise at each moment and the financial risk of the upstream enterprise at each moment, thereby predicting the financial risk based on the changing characteristics of the financial risk over time.

[0027] (4) The present invention determines the financial risk at each historical moment based on the overdue repayment situation of the upstream enterprise at each historical moment provided by the financial institution; quantifies the second indicator; trains the graph convolutional neural network and the fully connected layer based on the second indicator and the financial risk at each historical moment; inputs the second indicator at the current moment into the graph convolutional neural network and the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment; maps the second indicator: external environment, enterprise quality, public opinion information, the influence factor of the upstream enterprise affected by the connected central enterprise, and the operation level of the industrial chain where the upstream enterprise is located into a graph structure, extracts features through the graph convolutional neural network and connects through the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment. Because the relationship between enterprises is a knowledge graph structure, the features therein can be well learned through the graph convolutional neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0029] Figure 1 It is a flow chart of a method for evaluating and predicting financial risks in the automotive industry chain provided by an embodiment of the present invention;

[0030] Figure 2 is a schematic diagram of the structure of the first knowledge graph provided by an embodiment of the present invention;

[0031] Figure 3 is a schematic diagram of the structure of the second knowledge graph provided by an embodiment of the present invention;

[0032] Figure 4 It is a structural schematic diagram of a financial risk evaluation and prediction system for an automobile industry chain provided by the present invention. DETAILED DESCRIPTION

[0033] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0034] Example 1

[0035] This embodiment is applicable to the evaluation of financial risks of enterprises in the automotive industry chain at historical and current times, as well as the prediction of financial risks at future times. Historical moments, current moments and future moments can be in monthly, quarterly or annual units. Financial risk refers to the risk of overdue repayments, bankruptcy, etc. of an enterprise, which can be expressed as a probability value between 0 and 100%, which reflects the health of the enterprise's operations.

[0036] See also Figure 1 , the method provided in this embodiment includes the following steps.

[0037] S110. Construct the first knowledge graph based on the financial and supply relationships between enterprises in the automotive industry chain.

[0038] Financial and supply relations include: equity, investment and financing, and supply relations between enterprises (i.e. the types and quantities of goods sold). Figure 2 ,The entities in the first knowledge graph represent enterprises, and the edges represent financial and ,supply relationships.

[0039] S120. According to the structural characteristics of the first knowledge graph, select multiple core enterprises and upstream enterprises of each core enterprise from the first knowledge graph.

[0040] Specifically, the enterprise scale, transaction density, centrality, in-degree, out-degree and location characteristics of each node of the first knowledge graph relationship network are calculated to determine the central enterprise and the upstream enterprises of the central enterprise. For example, the centrality of each entity node is calculated, and the central enterprise with a centrality greater than a threshold is selected as the central enterprise, and the upstream enterprise that supplies goods to the central enterprise is selected. In practice, there are multiple central enterprises in the first knowledge graph, and each central enterprise also has multiple upstream enterprises.

[0041] S130, obtaining a first indicator of the central enterprise at each moment; and calculating the financial risk of the central enterprise at each moment according to the first indicator and the weight.

[0042] The first indicators include external environment, enterprise quality and public opinion information. Specifically, the external environment includes economic data of the region where the central enterprise is located, the concentration of the automobile industry and the intensity of industry competition; enterprise quality includes enterprise information, market operation data and development trends; public opinion information includes industry public opinion information and enterprise risk public opinion information. The first indicator at different times may be different. The first indicator at each moment here is the first indicator at each historical moment and the current moment.

[0043] After collecting the first indicator at each moment in advance, the first indicator at each moment is stored in a database for future use.

[0044] S130 specifically includes the following steps:

[0045] S131. Categorize the industries of core enterprises.

[0046] This embodiment uses the core enterprises belonging to the same sub-industry as a group to calculate financial risks. Specifically, first, according to the industry relationship of the enterprise industrial chain, a second knowledge graph is constructed; see Figure 3 , the entities in the second knowledge graph represent subdivided industries, and the edges represent the subordinate relationship between subdivided industries. Then obtain the description text of the central enterprise; the description text includes the text description of the central enterprise's business scope, product description, etc. Split the description text into multiple segments, such as splitting by punctuation marks, or splitting by sentence components. Perform semantic similarity analysis on each segment and the entity in the second knowledge graph to obtain at least one entity similar to each segment. Specifically, each segment and each entity in the second knowledge graph are mapped to the vector space (VSM), and the angle between the vector of each segment and each entity is calculated to determine at least one entity similar to each segment, that is, at least one subdivided industry. This is only a rough industry matching through the second knowledge graph. It is also necessary to input each segment and at least one entity similar to each segment into the classifier to obtain the subdivided industry of the central enterprise. The classifier can be a naive Bayes classifier, a support vector machine, or a neural network. A segment and at least one entity similar to it can be spliced ​​and input into the classifier for semantic classification to obtain the classification probability of each subdivided industry and determine the subdivided industry with the highest classification probability.

[0047] S132. For the core enterprises in the same industry, the analytic hierarchy process is used to determine the weight of each first indicator at each moment.

[0048] The analytic hierarchy process, referred to as AHP, refers to a decision-making method that decomposes elements that are always related to decision-making into levels such as goals, criteria, and plans, and conducts qualitative and quantitative analysis on this basis. In this embodiment, the quasi-layer is each first indicator. At each historical moment and the current moment, experts evaluate the importance of the first indicators in pairs, and then determine the weight of each first indicator. It should be noted that the importance of different first indicators may be different at different moments, so it is necessary to use the analytic hierarchy process to determine the weight of the first indicator at each moment.

[0049] S133: quantify the first indicator.

[0050] The content of the first indicator is introduced above, but if the first indicator is involved in the calculation, it needs to be quantified. Specifically, the economic data of the region where the enterprise is located can be expressed by the gross domestic product (GDP) of the region. The concentration of the automobile industry is the concentration of automobile supply chain enterprises in the region where the enterprise is located, which can be expressed by the total number of automobile supply chain enterprises in the region / the total number of automobile supply chain enterprises in the country. The intensity of industry competition can be calculated by the total number and turnover of automobile supply chain enterprises in the region where the enterprise is located, and the industry competition intensity is set to a number from 1 to 9. The intensity of industry competition is determined according to the size of the total number and turnover. For example, if the total number of automobile supply chain enterprises in the region where an enterprise is located is 100 and the turnover is 10 billion, it is set to 9. If the total number of automobile supply chain enterprises in the region where an enterprise is located is 10 and the turnover is 500 million, it is set to 1. Enterprise information includes quantitative information such as the year of establishment of the enterprise and the number of employees. Market operation data include market share and operating efficiency. For example, there are a total of H car companies, obtain the number of vehicles sold by each car company, and calculate the proportion of each car company's sales of vehicles to the total sales of all car companies to obtain the market share. The number of vehicles sold by each car company is divided by the number of vehicles in stock to obtain the operating efficiency. The development trend is the future development trend of the company in terms of company size, which can be expressed by the historical profit growth rate or historical revenue growth rate. The industry public opinion information is quantified as the impact of news and consultations within the industry on the automotive industry chain. The impact quantification values ​​of -10 to 10 are pre-set, as well as the keywords corresponding to different impact quantification values: development, positive, construction, relocation, demolition, risk, etc. The news and consultations within the industry are matched with the keywords to obtain the impact corresponding to the matched keywords. Similarly, the enterprise risk public opinion information is quantified as the impact of the enterprise's news and consultations on the enterprise. The impact quantification values ​​of -10 to 10 are pre-set, as well as the keywords corresponding to different impact quantification values: inquiry, cooperation, suspension, investment, etc. The enterprise's news and consultations are matched with the keywords to obtain the impact corresponding to the matched keywords.

[0051] S134. Use the weight at each moment to perform weighted summation on the first indicator to obtain the financial risk of the central enterprise at each moment.

[0052] Preferably, considering that the first indicator may deviate from the actual situation, the financial risk at each moment is verified. Specifically, after S134, the financial indicator analysis of the central enterprise at each moment is obtained and the financial risk is calculated. The financial indicator analysis includes the profitability, operating capacity, debt repayment capacity, and development capacity of the enterprise (these financial indicators are calculated based on the actual data of the financial institution and weighted). The present invention only needs to obtain the calculated financial risk from the financial institution; each moment here includes each historical moment and the current moment. Calculate the consistency coefficient of the financial risk obtained in S134 and the financial risk obtained by the financial institution. For example, the size of the consistency coefficient is determined according to the size of the difference between the two. The larger the difference between the two, the smaller the consistency coefficient. When the two are equal, the consistency coefficient reaches the maximum value. Optionally, the reciprocal of the difference is used as the consistency coefficient. Verify the financial risk of the central enterprise at each moment according to the consistency coefficient. If the consistency coefficient is greater than or equal to the set value, it is considered that the financial risk obtained in S134 is verified and can be applied to subsequent steps; if the consistency coefficient is less than the set value, it means that the financial risk obtained in S134 is not verified, and the weight of the first indicator needs to be updated and re-verified.

[0053] S140. Obtain the second indicator of the upstream enterprise at each moment; calculate the financial risk of the upstream enterprise at each moment according to the second indicator based on the graph convolutional neural network and the fully connected layer.

[0054] Each moment here includes every historical moment and the current moment. The second indicator includes the external environment, corporate quality, public opinion information, the impact factor of upstream enterprises affected by the connected central enterprises, the operation level of the industrial chain where the upstream enterprises are located, and the financial risks of all central enterprises connected to the upstream enterprises. Among them, the specific contents of the external environment, corporate quality, and public opinion information refer to the description of the first indicator, which will not be repeated here.

[0055] The central enterprise connected to the upstream enterprise is determined according to the connection relationship of the entities in the first knowledge graph. The influence factor of the upstream enterprise affected by the connected central enterprise is calculated in the following way: if the upstream enterprise is connected to multiple central enterprises, the weight corresponding to each central enterprise at each moment is determined according to the closeness of the financial and supply relationship between the upstream enterprise and each central enterprise at each moment; the financial risk of the central enterprise at each moment is weighted and summed to obtain the influence factor of the upstream enterprise affected by the connected central enterprise. The closeness of the supply relationship is calculated in the following way: the financial transaction amount or supply volume between the upstream enterprise and each central enterprise at each moment is obtained; according to the ratio of the financial transaction amount or supply volume of each central enterprise, the closeness of the financial and supply relationship between the upstream enterprise and each central enterprise at each moment is determined. Assume that in a certain year, upstream enterprise A connects central enterprise B and central enterprise C. The financial transaction amount (such as investment amount, loan amount, etc.) between upstream enterprise A and central enterprise B is 10 million, and the financial transaction amount between upstream enterprise A and central enterprise C is 20 million. Then the ratio of financial transaction amount is 1:2, the financial and supply relationship closeness between upstream enterprise A and central enterprise B is 1, and the financial and supply relationship closeness between upstream enterprise A and central enterprise C is 2. Similarly, upstream enterprise A supplies 1,000 pieces to central enterprise B, and upstream enterprise A supplies 2,000 pieces to central enterprise C. Then the supply ratio is 1:2, the financial and supply relationship closeness between upstream enterprise A and central enterprise B is 1 / 3, and the financial and supply relationship closeness between upstream enterprise A and central enterprise C is 2 / 3. Define the financial risk of central enterprise B as 30%, and the financial risk of central enterprise C as 40%, then the impact factor of upstream enterprise A affected by the connected central enterprise is 30%*1 / 3+40%*2 / 3≈0.37.

[0056] If an upstream enterprise is connected to a central enterprise, the financial risk of the central enterprise at each moment is used as the impact factor of the upstream enterprise being affected by the connected central enterprise. Assuming that in a certain year, upstream enterprise A is only connected to central enterprise D, and the financial risk of central enterprise D is 50%, then the impact factor of upstream enterprise A being affected by the connected central enterprise is 0.5.

[0057] The operation level of the industrial chain in which the upstream enterprises are located includes: the stability and concentration of the financial and supply relations of the upstream enterprises in the industrial chain. Among them, stability refers to the dynamic nature of the financial transaction amount and product transaction relationship between enterprises in the time dimension. For example, if the upstream enterprise S has been supplying central enterprise A, the stability is high; for example, if the upstream enterprise S supplied central enterprise A the year before last, supplied another enterprise last year, and changed to another enterprise this year, the stability of upstream enterprise S is low. Concentration refers to the degree of concentration of transactions between enterprises. If an upstream enterprise is the sole supplier of a certain component to a central enterprise, the concentration of the upstream enterprise is high, and the concentration decreases as the proportion of upstream enterprises supplying each enterprise decreases.

[0058] This step focuses on using graph convolutional neural networks and fully connected layers to calculate the financial risks of upstream companies. Before using graph convolutional neural networks and fully connected layers to calculate the financial risks of upstream companies, the graph convolutional neural networks and fully connected layers need to be trained in advance. Specifically, the financial risk at each historical moment is determined based on the overdue repayment situation of upstream companies at each historical moment provided by financial institutions. The financial risk can be determined specifically based on the amount and / or number of overdue repayments. If the amount of overdue repayments is large and the number of overdue repayments is large, the financial risk is high. Then, the second indicator is quantified. The quantitative processing of the external environment, corporate quality, and public opinion information can be found in the description of the first indicator, which will not be repeated here.

[0059] Specifically, the quantitative method of the stability of the financial and supply relationship of upstream enterprises in the industrial chain includes: calculating the year-on-year change rate W1 of the number of enterprises with financial or supply relationships, the average cooperation years W2 of enterprises with financial or supply relationships, and the average year-on-year change rate W3 of the products or amounts supplied to each enterprise for each upstream enterprise, and obtaining three data representations of stability: W1, W2, and W3. Optionally, the year-on-year change rate is the year-on-year change rate of two adjacent moments (monthly or annual). The quantitative method of the concentration of financial and supply relationships of upstream enterprises in the industrial chain includes: sorting multiple enterprises with which each upstream enterprise has financial or supply relationships in order of supply products or amounts from high to low. Filter the top 20% of enterprises to calculate the total amount or total product volume W4, calculate the total amount or total product volume W5 of each upstream enterprise's transactions for all enterprises, and use W4 / W5 as the concentration of each upstream enterprise.

[0060] Then, according to the second indicator and financial risk at each historical moment, the graph convolutional neural network and the fully connected layer are trained. For an upstream enterprise, since the second indicator of the upstream enterprise includes not only the information of the upstream enterprise itself, but also the financial risk of the central enterprise connected to it, the financial risk of the central enterprise here can be used as the value of the central enterprise entity, and the remaining second indicators (including the external environment, enterprise quality, public opinion information, and the level of operation of the industrial chain in which the upstream enterprise is located) are quantified as the value of the upstream enterprise entity, and the aforementioned influencing factors are used as the value of the edge connecting the upstream enterprise and the central enterprise to form a knowledge graph, and the feature vector of the knowledge graph is extracted through the graph convolutional neural network, and the feature vector is input into the fully connected layer to output the financial risk of the upstream enterprise. By iterating the parameters of the graph convolutional neural network and the fully connected layer, the gap between the output financial risk and the financial risk determined according to the overdue repayment situation is minimized. Finally, the trained graph convolutional neural network and the fully connected layer are obtained. Finally, the second indicator at the current moment is input into the graph convolutional neural network and the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment. After obtaining the trained graph convolutional neural network and fully connected layer, we no longer need to use the overdue repayment information provided by financial institutions. We can directly use the graph convolutional neural network and fully connected layer to determine the financial risks at the current moment. As time goes by, when the current moment is called a historical moment, the financial risks of upstream companies at each historical moment and the current moment will be obtained.

[0061] S150. Using a time series analysis model, based on the financial risk of the central enterprise at each moment and the financial risk of the upstream enterprise at each moment, predict the financial risk of each enterprise at a future moment.

[0062] Optionally, the time series analysis model is a long short-term memory model LSTM. The training process of LSTM is as follows: collect multiple historical moments t of an enterprise (core enterprise or upstream enterprise) n ,t n-1 ,t n-2 ,…,t3,t2,t1, these are all confirmed financial risks. n ,t n-1 ,t n-2 ,…,t3,t2,t1: Select training samples from the financial risks at historical moment t m ,t m-1 ,t m-2 ,…t p ,t p-1 Financial risks, labeled as historical moments t p-2financial risk; where m, n, p are natural numbers, and n≥m>p. That is, the training samples are known financial risks at multiple consecutive historical moments, and the labels are the financial risks at the moment after the last moment in the training samples. The LSTM is trained using the training samples and labels.

[0063] After the training is completed, the financial risks of a central enterprise or an upstream enterprise at multiple historical moments and the financial risks at the current moment are input into the time series analysis model to obtain the financial risks of the enterprise at future moments.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] (1) The present invention divides enterprises into core enterprises and upstream enterprises based on the financial and supply relationships between enterprises, thereby adopting different methods to evaluate financial risks according to the characteristics of the enterprises, thereby obtaining financial risks that are more in line with the characteristics of the enterprises.

[0066] (2) The present invention adopts a variety of first indicators and second indicators to calculate the financial risks of enterprises from multiple dimensions.

[0067] (3) The present invention adopts a time series analysis model to predict the financial risk of each enterprise at a future moment based on the financial risk of the central enterprise at each moment and the financial risk of the upstream enterprise at each moment, thereby predicting the financial risk based on the changing characteristics of the financial risk over time.

[0068] (4) The present invention determines the financial risk at each historical moment based on the overdue repayment situation of the upstream enterprise at each historical moment provided by the financial institution; quantifies the second indicator; trains the graph convolutional neural network and the fully connected layer based on the second indicator and the financial risk at each historical moment; inputs the second indicator at the current moment into the graph convolutional neural network and the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment; maps the second indicator: external environment, enterprise quality, public opinion information, the influence factor of the upstream enterprise affected by the connected central enterprise, and the operation level of the industrial chain where the upstream enterprise is located into a graph structure, extracts features through the graph convolutional neural network and connects through the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment. Because the relationship between enterprises is a knowledge graph structure, the features therein can be well learned through the graph convolutional neural network.

[0069] Example 2

[0070] This embodiment provides a financial risk assessment and prediction system for the automotive industry chain. Figure 4 , including: data layer, technology layer and service layer;

[0071] The data layer is used to store the first indicator and the second indicator at each moment;

[0072] The technical layer is used to construct a first knowledge graph based on the financial and supply relationships between enterprises in the automotive industry chain, wherein entities in the first knowledge graph represent enterprises and edges represent financial and supply relationships; based on the structural characteristics of the first knowledge graph, multiple central enterprises and upstream enterprises of each central enterprise are selected from the first knowledge graph;

[0073] The service layer is used to obtain the first indicator of the central enterprise at each moment; calculate the financial risk of the central enterprise at each moment according to the first indicator and the weight; the first indicator includes the external environment, enterprise quality and public opinion information; obtain the second indicator of the upstream enterprise at each moment; based on the graph convolutional neural network and the fully connected layer, calculate the financial risk of the upstream enterprise at each moment according to the second indicator; the second indicator includes the external environment, enterprise quality, public opinion information, the impact factor of the upstream enterprise affected by the connected central enterprise, and the operation level of the industrial chain of the upstream enterprise; use the time series analysis model to predict the financial risk of each enterprise at a future moment according to the financial risk of the central enterprise at each moment and the financial risk of the upstream enterprise at each moment;

[0074] Wherein, based on the graph convolutional neural network and the fully connected layer, according to the second indicator, the financial risk of the upstream enterprise at each moment is calculated, including: determining the financial risk at each historical moment according to the overdue repayment situation of the upstream enterprise at each historical moment provided by the financial institution; quantifying the second indicator; training the graph convolutional neural network and the fully connected layer according to the second indicator and the financial risk at each historical moment; inputting the second indicator at the current moment into the graph convolutional neural network and the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment.

[0075] Optionally, the system further includes an application layer; the user can input the enterprise ID on the terminal screen and click the query button. The application layer responds to the user's query operation and displays the financial risk of the enterprise at a future moment on the terminal screen.

[0076] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and this document does not limit this.

[0077] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating and predicting financial risks in the automotive industry chain, characterized in that: include: According to the financial and supply relationships between enterprises in the automotive industry chain, a first knowledge graph is constructed, wherein entities in the first knowledge graph represent enterprises, and edges represent financial and supply relationships; The financial and supply relationships include equity, investment and financing, or supply relationships between enterprises; financial risk refers to the risk of overdue repayment or bankruptcy of an enterprise; According to the structural characteristics of the first knowledge graph, multiple central enterprises and upstream enterprises of each central enterprise are selected from the first knowledge graph; wherein the centrality of each entity node is calculated, and the centrality greater than a threshold is selected as the central enterprise, and the upstream enterprise that supplies goods to the central enterprise is selected; Obtaining the first indicator of the central enterprise at each historical moment and the current moment; calculating the financial risk of the central enterprise at each historical moment and the current moment according to the first indicator and the weight of each first indicator at each historical moment and the current moment; the first indicator includes external environment, enterprise quality and public opinion information; Obtaining a second indicator of the upstream enterprise at each historical moment and the current moment; calculating the financial risk of the upstream enterprise at each historical moment and the current moment based on the graph convolutional neural network and the fully connected layer and the second indicator; the second indicator includes the external environment, enterprise quality, public opinion information, the impact factor of the upstream enterprise affected by the connected central enterprise, the operation level of the industrial chain in which the upstream enterprise is located, and the financial risks of all central enterprises connected to the upstream enterprise; The external environment includes economic data of the region where the enterprise is located, automobile industry concentration and industry competition intensity. The automobile industry concentration refers to the degree of concentration of automobile supply chain enterprises in the region where the enterprise is located, which is expressed by the total number of automobile supply chain enterprises in the region / the total number of automobile supply chain enterprises in the country. The industry competition intensity is calculated by the total number and turnover of automobile supply chain enterprises in the region where the enterprise is located, and the industry competition intensity is determined according to the total number of enterprises and the size of their turnover; the enterprise quality includes enterprise information, market operation data and development trends. The enterprise information includes quantitative information on the years of establishment and the number of employees of the enterprise. The market operation data includes market share and operating efficiency. The development trend is the future development trend of the enterprise in terms of company size, which is expressed by historical profit growth rate or historical revenue growth rate; the public opinion information includes industry public opinion information and enterprise risk public opinion information; the industry public opinion information is quantified as the degree of influence of news and consultation within the industry on the automobile industry chain, which is preset to -10 ~10, and the keywords corresponding to different impact degree quantification values: development, positive, construction, relocation, demolition, risk, match the news and consultation in the industry with the keywords, and thus obtain the impact degree corresponding to the matched keywords; the enterprise risk public opinion information is quantified as the impact degree of the enterprise's news and consultation on the enterprise, and the impact degree quantification value of -10 to 10 is pre-set, as well as the keywords corresponding to different impact degree quantification values: inquiry, cooperation, suspension, investment, match the enterprise's news and consultation with the keywords, and thus obtain the impact degree corresponding to the matched keywords; the industrial chain operation level of the upstream enterprise includes: the stability and concentration of the financial and supply relationship of the upstream enterprise in the industrial chain. Stability is the nature of the dynamic changes in the financial transaction amount and product transaction relationship between enterprises in the time dimension. If an upstream enterprise has been supplying a central enterprise, then the upstream enterprise has high stability; concentration indicates the concentration of transactions between enterprises. If an upstream enterprise is the sole supplier of a central enterprise for a certain component, then the upstream enterprise has high concentration; Using a time series analysis model, based on the financial risks of the central enterprise at each historical moment and the current moment, and the financial risks of the upstream enterprise at each historical moment and the current moment, the financial risks of the central enterprise and the upstream enterprise at the future moment are predicted; Wherein, based on the graph convolutional neural network and the fully connected layer, according to the second indicator, the financial risk of the upstream enterprise at each historical moment and the current moment is calculated, including: Determine the financial risk at each historical moment based on the repayment overdue situation of upstream enterprises at each historical moment provided by financial institutions; quantifying the second indicator; According to the second indicator and financial risk at each historical moment, the graph convolutional neural network and the fully connected layer are trained; the financial risk of the central enterprise is taken as the value of the central enterprise entity, the external environment, enterprise quality, public opinion information, and the industrial chain operation level indicators of the upstream enterprise are quantified as the value of the upstream enterprise entity, and the impact factor is taken as the value of the edge connecting the upstream enterprise and the central enterprise to form a third knowledge graph, and the feature vector of the third knowledge graph is extracted through the graph convolutional neural network, and the feature vector is input into the fully connected layer to output the financial risk of the upstream enterprise; by iterating the parameters of the graph convolutional neural network and the fully connected layer, the gap between the output financial risk of the upstream enterprise and the financial risk of the upstream enterprise determined according to the overdue repayment situation is minimized; finally, the trained graph convolutional neural network and the fully connected layer are obtained; The second indicator at the current moment is input into the graph convolutional neural network and the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment; after obtaining the trained graph convolutional neural network and the fully connected layer, the repayment overdue situation provided by the financial institution is no longer used, and the graph convolutional neural network and the fully connected layer are directly used to determine the financial risk at the current moment. As time goes by, the current moment becomes a historical moment, and the financial risks of the upstream enterprises at each historical moment and the current moment are obtained; The impact factor of the upstream enterprise affected by the connected central enterprise is calculated in the following way: If the upstream enterprise is connected to multiple central enterprises, the weight corresponding to each central enterprise at each moment is determined according to the closeness of the financial and supply relationship between the upstream enterprise and each central enterprise at each moment; the financial risk of the central enterprise at each moment is weighted and summed to obtain the impact factor of the upstream enterprise affected by the connected central enterprise; if the upstream enterprise is connected to one central enterprise, the financial risk of the central enterprise at each moment is used as the impact factor of the upstream enterprise affected by the connected central enterprise; The degree of financial and supply relationship closeness is calculated in the following manner: obtaining the financial transaction amount or supply volume between the upstream enterprise and each central enterprise at each moment; and determining the degree of financial and supply relationship between the upstream enterprise and each central enterprise at each moment based on the ratio of the financial transaction amount or supply volume of each central enterprise.

2. The method according to claim 1, characterized in that Calculating the financial risk of the central enterprise at each historical moment and the current moment according to the weight of the first indicator and each first indicator at each historical moment and the current moment, including: Categorize the central enterprises by industry; For the core enterprises in the same industry, the analytic hierarchy process is used to determine the weight of each first indicator at each historical moment and the current moment; Quantifying the first indicator; The weight of each first indicator at each historical moment and the current moment is used to perform a weighted summation of the first indicators to obtain the financial risk of the central enterprise at each historical moment and the current moment.

3. The method according to claim 2, characterized in that The industry classification of central enterprises includes: According to the industry relationship of the enterprise industrial chain, a second knowledge graph is constructed; the entities in the second knowledge graph represent subdivided industries, and the edges represent the subordinate relationships between subdivided industries; Obtaining a description text of the central enterprise; Splitting the description text into multiple segments; Perform semantic similarity analysis on each segment and entities in the second knowledge graph to obtain at least one entity similar to each segment; Each segment and at least one entity similar to each segment are input into a classifier to obtain the subdivided industry of the central enterprise.

4. The method according to claim 2, characterized in that: After weighting and summing the first indicators by using the weights of each first indicator at each historical moment and the current moment to obtain the financial risk of the central enterprise at each historical moment and the current moment, it also includes: Obtain the financial risks of the central enterprise at each historical moment and the current moment provided by financial institutions; Calculate the consistency coefficient between the financial risk provided by the financial institution and the financial risk obtained by weighted summation; The financial risk of the central enterprise at each historical moment and the current moment is verified according to the consistency coefficient.

5. The method according to claim 1, characterized in that The time series analysis model is a long short-term memory model LSTM; Before using the time series analysis model to predict the financial risks of the central enterprise and the upstream enterprise at future moments based on the financial risks of the central enterprise at each historical moment and the current moment, and the financial risks of the upstream enterprise at each historical moment and the current moment, the method further includes: Collect multiple historical moments of an enterprise n ,t n-1 ,t n-2 ,…,t3,t2,t1’s financial risk; From multiple historical moments n ,t n-1 ,t n-2 ,…,t3,t2,t1: Select training samples from the financial risks at historical moment t m ,t m-1 ,t m-2 ,…t p ,t p-1 Financial risks, labeled as historical moments t p-2 The training samples are known financial risks at multiple historical moments, the labels are the financial risks at the moment after the last moment in the training samples, m, n, p are natural numbers, and n≥m>p; Using the training samples and labels, training the LSTM; After the training is completed, the financial risks of a central enterprise or an upstream enterprise at multiple historical moments and the financial risks at the current moment are input into the time series analysis model to obtain the financial risks of the enterprise at future moments.

6. A financial risk assessment and prediction system for the automotive industry chain, characterized in that: include: Data layer, technology layer and service layer; The data layer is used to store the first indicator and the second indicator at each moment; The technical layer is used to construct a first knowledge graph based on the financial and supply relationships between enterprises in the automotive industry chain. The entities in the first knowledge graph represent enterprises, and the edges represent financial and supply relationships. According to the structural characteristics of the first knowledge graph, multiple central enterprises and upstream enterprises of each central enterprise are selected from the first knowledge graph. The financial and supply relationships include equity, investment and financing, or supply relationships between enterprises. The centrality of each entity node is calculated, and the central enterprise with a centrality greater than a threshold is selected as the central enterprise, and the upstream enterprise that supplies goods to the central enterprise is selected. Financial risk refers to the risk of overdue repayment and bankruptcy of an enterprise. The service layer is used to obtain the first indicator of the central enterprise at each historical moment and the current moment; Calculate the financial risk of the central enterprise at each historical moment and the current moment according to the first indicator and the weight of each first indicator at each historical moment and the current moment; the first indicator includes external environment, enterprise quality and public opinion information; Obtaining a second indicator of the upstream enterprise at each historical moment and the current moment; calculating the financial risk of the upstream enterprise at each historical moment and the current moment based on the graph convolutional neural network and the fully connected layer and the second indicator; the second indicator includes the external environment, enterprise quality, public opinion information, the impact factor of the upstream enterprise affected by the connected central enterprise, the operation level of the industrial chain in which the upstream enterprise is located, and the financial risks of all central enterprises connected to the upstream enterprise; The external environment includes economic data of the region where the enterprise is located, automobile industry concentration and industry competition intensity. The automobile industry concentration refers to the degree of concentration of automobile supply chain enterprises in the region where the enterprise is located, which is expressed by the total number of automobile supply chain enterprises in the region / the total number of automobile supply chain enterprises in the country. The industry competition intensity is calculated by the total number and turnover of automobile supply chain enterprises in the region where the enterprise is located, and the industry competition intensity is determined according to the total number of enterprises and the size of their turnover; The enterprise quality includes enterprise information, market operation data and development trends. The enterprise information includes quantitative information on the years of establishment of the enterprise and the number of employees. The market operation data includes market share and operating efficiency. The development trend is the future development trend of the enterprise in terms of company size, which is expressed by historical profit growth rate or historical revenue growth rate; The public opinion information includes industry public opinion information and enterprise risk public opinion information; The industry public opinion information is quantified as the degree of influence of news and consultation within the industry on the automobile industry chain, which is preset to -10 The impact degree quantification values ​​of -10 to 10, and the keywords corresponding to different impact degree quantification values: development, positive, construction, relocation, demolition, risk, match the news and consultation in the industry with the keywords, and thus obtain the impact degree corresponding to the matched keywords; the enterprise risk public opinion information is quantified as the impact degree of the enterprise's news and consultation on the enterprise, pre-set the impact degree quantification values ​​of -10 to 10, and the keywords corresponding to different impact degree quantification values: inquiry, cooperation, suspension, investment, match the enterprise's news and consultation with the keywords, and thus obtain the impact degree corresponding to the matched keywords; The operation level of the industrial chain in which the upstream enterprises are located includes: the stability and concentration of the financial and supply relations of the upstream enterprises in the industrial chain. Stability refers to the dynamic change of the financial transaction amount and product transaction relationship between enterprises in the time dimension. If an upstream enterprise has been supplying goods to a central enterprise, then the upstream enterprise has high stability; concentration refers to the degree of concentration of transactions between enterprises. If an upstream enterprise is the sole supplier of a certain component to a central enterprise, then the upstream enterprise has high concentration. The service layer is used to use a time series analysis model to predict the financial risks of the central enterprise and the upstream enterprise at future moments according to the financial risks of the central enterprise at each historical moment and the current moment, and the financial risks of the upstream enterprise at each historical moment and the current moment; Among them, based on the graph convolutional neural network and the fully connected layer, according to the second indicator, the financial risk of the upstream enterprise at each historical moment and the current moment is calculated, including: determining the financial risk at each historical moment according to the overdue repayment situation of the upstream enterprise at each historical moment provided by the financial institution; quantifying the second indicator; training the graph convolutional neural network and the fully connected layer according to the second indicator and the financial risk at each historical moment; taking the financial risk of the central enterprise as the value of the central enterprise entity, quantifying the external environment, enterprise quality, public opinion information, and the industrial chain operation level indicators of the upstream enterprise as the value of the upstream enterprise entity, taking the influencing factor as the value of the edge connecting the upstream enterprise and the central enterprise, forming a third knowledge graph, and extracting the characteristics of the third knowledge graph through the graph convolutional neural network. vector, input the feature vector into the fully connected layer, and output the financial risk of the upstream enterprise; by iterating the parameters of the graph convolutional neural network and the fully connected layer, minimize the gap between the output financial risk of the upstream enterprise and the financial risk of the upstream enterprise determined according to the overdue repayment situation; finally obtain the trained graph convolutional neural network and the fully connected layer; input the second indicator at the current moment into the graph convolutional neural network and the fully connected layer to obtain the financial risk of the upstream enterprise at the current moment; after obtaining the trained graph convolutional neural network and the fully connected layer, no longer use the overdue repayment situation provided by the financial institution, directly use the graph convolutional neural network and the fully connected layer to determine the financial risk at the current moment, as time goes by, the current moment becomes a historical moment, and the financial risk of the upstream enterprise at each historical moment and the current moment will be obtained; The impact factor of the upstream enterprise affected by the connected central enterprise is calculated in the following way: If the upstream enterprise is connected to multiple central enterprises, the weight corresponding to each central enterprise at each moment is determined according to the closeness of the financial and supply relationship between the upstream enterprise and each central enterprise at each moment; the financial risk of the central enterprise at each moment is weighted and summed to obtain the impact factor of the upstream enterprise affected by the connected central enterprise; if the upstream enterprise is connected to one central enterprise, the financial risk of the central enterprise at each moment is used as the impact factor of the upstream enterprise affected by the connected central enterprise; The degree of financial and supply relationship closeness is calculated in the following manner: obtaining the financial transaction amount or supply volume between the upstream enterprise and each central enterprise at each moment; and determining the degree of financial and supply relationship between the upstream enterprise and each central enterprise at each moment based on the ratio of the financial transaction amount or supply volume of each central enterprise.

7. The system according to claim 6, characterized in that Also includes: Application layer; The application layer is used to display the financial risk of the enterprise at a future time on the terminal screen in response to the user's query operation.

Images