Financial risk early warning method and device based on multi-model fusion, and electronic equipment

Abstract

The invention discloses a financial risk early warning method and device based on multi-model fusion and electronic equipment, and relates to the field of financial science and technology or other related fields, and the method comprises the steps: obtaining financial statement data of a target enterprise, screening financial indexes based on a statistical method, and forming a target index set; based on the target index set, calculating the deviation degree of the target enterprise relative to the industry level index through a first calculation model to obtain a first score value; calculating the financial health degree of the target enterprise through a second calculation model and a preset financial health standard grade to obtain a second score value; judging whether the target enterprise hits a preset potential risk group rule set or not through a third calculation model to obtain a group recognition result; and generating a risk early warning signal based on the first score value, the second score value and the group identification result according to a preset signal mapping rule. Through the method and the device, the technical problem of low comprehensive efficiency of financial risk identification caused by defects existing in a financial risk early warning mechanism in related technologies is solved.

Description

Technical Field

[0001] This invention relates to the field of financial technology or other related fields, and more specifically, to a financial risk early warning method, device, and electronic device based on multi-model fusion. Background Technology

[0002] As financial markets become increasingly complex, the financial risks faced by non-retail enterprises are becoming more multidimensional and volatile. Traditional financial early warning systems often focus on monitoring single or static financial indicators, resulting in low overall effectiveness in the current economic environment. On the one hand, the failure to fully consider industry characteristics and the dynamic changes in corporate financial indicators significantly reduces the timeliness and accuracy of early warning signals. On the other hand, high-risk industries present both opportunities and threats, but existing technologies often employ a "one-size-fits-all" approach, making it difficult to accurately identify potential high-value customers, leading to rigid credit strategies and wasted resources.

[0003] Of particular concern is that different industries (such as typical sub-sectors of manufacturing (such as machinery, chemicals, building materials, etc.) and micro and small enterprises) have significant differences in financial risk characteristics. However, existing early warning systems generally ignore these differences and adopt uniform standards, which undoubtedly exacerbates the false alarm rate of early warning signals and makes financial institutions lack sufficient flexibility and foresight when dealing with complex market environments.

[0004] In summary, the financial risk early warning mechanisms in related technologies have flaws, resulting in low overall efficiency in financial risk identification. Currently, no effective solution has been proposed to address these issues. Summary of the Invention

[0005] The main objective of this application is to provide a financial risk early warning method, device, and electronic device based on multi-model fusion, so as to at least solve the technical problem that the financial risk early warning mechanism in the related technology has defects, resulting in low overall efficiency of financial risk identification.

[0006] To achieve the above objectives, according to one aspect of this application, a financial risk early warning method based on multi-model fusion is provided. The method includes: acquiring financial statement data of a target enterprise, and selecting financial indicators from the financial statement data using statistical methods to form a target indicator set; calculating the deviation of the target enterprise from industry-level indicators using a first calculation model based on the target indicator set to obtain a first score value; calculating the financial health of the target enterprise using a second calculation model and a preset financial health standard level based on the target indicator set to obtain a second score value; determining whether the target enterprise hits a preset potential risk group rule set using a third calculation model based on the target indicator set to obtain a group identification result; and generating a risk early warning signal based on the first score value, the second score value, and the group identification result according to a preset signal mapping rule.

[0007] Furthermore, the steps for obtaining the financial statement data of the target company include: retrieving the annual report text of the target company from a designated enterprise information disclosure platform; parsing the annual report text to obtain a parsing result, wherein the parsing result includes at least the financial data items in the balance sheet, income statement, and cash flow statement; and standardizing and formatting the financial data items in the parsing result to obtain the financial statement data.

[0008] Further, the step of selecting financial indicators from the financial statement data based on statistical methods to construct a target indicator set includes: identifying all companies within the sub-sector to which the target company belongs as peer companies, and obtaining the financial statement data of the peer companies to obtain peer financial data; determining the peer financial distribution based on the peer financial data, and determining the quantile ratio according to the peer financial distribution, wherein the quantile ratio includes at least a first quantile ratio and a second quantile ratio, and the first quantile ratio is less than the second quantile ratio; using the first quantile ratio and the second quantile ratio as boundaries, selecting financial data items in the financial statement data of the target company that are outside the boundaries; and constructing the target indicator set based on all the selected financial data items.

[0009] Further, the step of calculating the deviation of the target enterprise from the industry-level indicators using the first calculation model to obtain the first score value includes: for each financial indicator in the target indicator set, calculating a quantile point based on the peer financial distribution and the quantile ratio, wherein the quantile point includes at least a first numerical point and a second numerical point; calculating the median based on the quantile point, and calculating the absolute deviation rate of each financial indicator from the median, wherein the median represents the industry level of the sub-sector to which it belongs; and weighting and summing the absolute deviation rates corresponding to all the financial indicators according to a first preset weight to obtain the first score value, wherein the first preset weight is obtained by training the first calculation model based on historical financial data.

[0010] Further, the step of calculating the financial health of the target enterprise and obtaining a second score value through the second calculation model and preset financial health standard levels includes: determining the evaluation level corresponding to each financial indicator based on the specified enterprise performance evaluation standard value, wherein the evaluation level includes: excellent value, good value, average value, low value, and poor value; interpolating the actual value of each financial indicator of the target enterprise with the standard value of the corresponding evaluation level to obtain the level score of each financial indicator; and performing a weighted average of the level scores of each financial indicator according to a second preset weight to obtain the second score value, wherein the second preset weight is obtained by training the second calculation model based on historical financial data.

[0011] Further, the step of determining whether the target enterprise matches the preset potential risk group rule set through the third calculation model to obtain the group identification result includes: obtaining the preset potential risk group rule set, wherein the preset potential risk group rule set records N financial risk feature patterns, where N is a preset value; analyzing the target indicator set of the target enterprise to obtain the corresponding financial indicator distribution pattern; matching the financial indicator distribution pattern with each financial risk feature pattern in the preset potential risk group rule set and calculating the matching degree; setting the group identification result to a hit state when any matching degree is higher than or equal to a preset matching degree threshold; and setting the group identification result to a miss state when all matching degrees are lower than the preset matching degree threshold.

[0012] Further, the step of generating a risk warning signal based on the first score value, the second score value, and the group identification result according to a preset signal mapping rule includes: setting the first score value exceeding a first preset threshold as a first judgment condition; setting the second score value below a second preset threshold as a second judgment condition; and setting the group identification result as a hit state as a third judgment condition; generating a first type of risk warning signal when the third judgment condition is met and either the first or second judgment condition is met; generating a second type of risk warning signal when only the third judgment condition is met, or when only the first and second judgment conditions are met simultaneously; generating a third type of risk warning signal when only the first judgment condition is met, or when only the second judgment condition is met; and generating a fourth type of risk warning signal when none of the judgment conditions are met.

[0013] To achieve the above objectives, according to another aspect of this application, a financial risk early warning device based on multi-model fusion is also provided. The device includes: an acquisition unit, configured to acquire financial statement data of a target enterprise and, based on statistical methods, filter financial indicators from the financial statement data to form a target indicator set; a first calculation unit, configured to, based on the target indicator set, calculate the degree of deviation of the target enterprise relative to industry-level indicators using a first calculation model to obtain a first score value; a second calculation unit, configured to, based on the target indicator set, calculate the financial health of the target enterprise using a second calculation model and a preset financial health standard level to obtain a second score value; a judgment unit, configured to, based on the target indicator set, determine whether the target enterprise hits a preset potential risk group rule set using a third calculation model to obtain a group identification result; and a generation unit, configured to, according to a preset signal mapping rule, generate a risk early warning signal based on the first score value, the second score value, and the group identification result.

[0014] Furthermore, the acquisition unit includes: a crawling module, used to crawl the annual report text of the target enterprise from a designated enterprise information disclosure platform; a parsing module, used to parse the annual report text to obtain a parsing result, wherein the parsing result includes at least: financial data items from the balance sheet, income statement, and cash flow statement; and a processing module, used to standardize and format the financial data items in the parsing result to obtain the financial statement data.

[0015] Furthermore, the acquisition unit further includes: a first acquisition module, configured to determine all companies within the sub-sector of the target company as peer companies, and acquire the financial statement data of the peer companies to obtain peer financial data; a first determination module, configured to determine the peer financial distribution based on the peer financial data, and determine the quantile ratio according to the peer financial distribution, wherein the quantile ratio includes at least a first quantile ratio and a second quantile ratio, and the first quantile ratio is less than the second quantile ratio; a filtering module, configured to filter financial data items outside the boundaries of the target company's financial statement data, using the first quantile ratio and the second quantile ratio as boundaries; and a construction module, configured to construct the target indicator set based on all the filtered financial data items.

[0016] Further, the first calculation unit includes: a first calculation module, used to calculate a quantile for each financial indicator in the target indicator set based on the peer financial distribution and the quantile ratio, wherein the quantile includes at least a first numerical point and a second numerical point; a second calculation module, used to calculate the median based on the quantile and calculate the absolute deviation rate of each financial indicator from the median, wherein the median represents the industry level of the sub-sector to which it belongs; and a weighted summation module, used to perform a weighted summation of the absolute deviation rates corresponding to all the financial indicators according to a first preset weight to obtain the first score value, wherein the first preset weight is obtained by training the first calculation model based on historical financial data.

[0017] Further, the second calculation unit includes: a second determining module, used to determine the evaluation level corresponding to each of the financial indicators based on the specified enterprise performance evaluation standard value, wherein the evaluation level includes: excellent value, good value, average value, low value, and poor value; a third calculation module, used to interpolate the actual value of each of the financial indicators of the target enterprise with the standard value of the corresponding evaluation level to obtain the level score of each of the financial indicators; and a weighted average module, used to perform a weighted average of the level scores of each of the financial indicators based on a second preset weight to obtain the second score value, wherein the second preset weight is obtained by training the second calculation model based on historical financial data.

[0018] Further, the judgment unit includes: a second acquisition module, used to acquire a preset potential risk group rule set, wherein the preset potential risk group rule set records N financial risk feature patterns, and N is a preset value; an analysis module, used to analyze the target indicator set of the target enterprise to obtain the corresponding financial indicator distribution pattern; a matching module, used to match the financial indicator distribution pattern with each financial risk feature pattern in the preset potential risk group rule set, and calculate the matching degree; a first setting module, used to set the group identification result to a hit state when any of the matching degrees is higher than or equal to a preset matching degree threshold; and a second setting module, used to set the group identification result to a miss state when all the matching degrees are lower than the preset matching degree threshold.

[0019] Further, the generation unit includes: a third setting module, configured to set the first score value exceeding a first preset threshold as a first judgment condition, set the second score value below a second preset threshold as a second judgment condition, and set the group identification result as a hit state as a third judgment condition; a first generation module, configured to generate a first type of risk warning signal when the third judgment condition is met and either the first judgment condition or the second judgment condition is met; a second generation module, configured to generate a second type of risk warning signal when only the third judgment condition is met, or when only the first judgment condition and the second judgment condition are met simultaneously; a third generation module, configured to generate a third type of risk warning signal when only the first judgment condition is met, or when only the second judgment condition is met; and a fourth generation module, configured to generate a fourth type of risk warning signal when all judgment conditions are not met.

[0020] To achieve the above objectives, according to another aspect of this application, a computer-readable storage medium is also provided, the computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the financial risk early warning method based on multi-model fusion as described above.

[0021] To achieve the above objectives, according to another aspect of this application, an electronic device is also provided, including one or more processors and a memory, the memory being used to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors cause the one or more processors to implement the financial risk early warning method based on multi-model fusion as described above.

[0022] To achieve the above objectives, according to another aspect of this application, a computer program product is also provided, including computer instructions, wherein when the computer instructions are executed by a processor, they implement the steps of the financial risk early warning method based on multi-model fusion described in any one of the above claims.

[0023] This invention proposes a financial risk early warning method based on multi-model fusion. First, the financial statement data of the target enterprise is obtained, and financial indicators are screened from the financial statement data using statistical methods to form a target indicator set. Then, based on the target indicator set, the deviation of the target enterprise from the industry level indicators is calculated by a first calculation model to obtain a first score value. Next, based on the target indicator set, the financial health of the target enterprise is calculated by a second calculation model and a preset financial health standard level to obtain a second score value. Then, based on the target indicator set, a third calculation model is used to determine whether the target enterprise hits a preset potential risk group rule set to obtain a group identification result. Finally, a risk early warning signal is generated based on the first score value, the second score value, and the group identification result according to a preset signal mapping rule.

[0024] This invention employs a multi-model fusion approach, utilizing statistical screening, dynamic deviation calculation, quantitative assessment of financial health, and group risk characteristic matching to achieve comprehensive and accurate identification of corporate financial risks. This results in improved early warning accuracy, shortened lead time, and effective differentiation between risks and opportunities. Consequently, it addresses the technical problem of deficiencies in financial risk early warning mechanisms in related technologies, leading to low overall efficiency in financial risk identification.

[0025] Specifically, this invention first uses statistical methods to screen financial indicators from the financial statement data of target companies, constructing a highly focused set of target indicators. Then, it uses a first calculation model to calculate the deviation of the target company's financial indicators from the industry level, obtaining a first score value to intuitively reflect the dynamic trend of the company's financial situation. Subsequently, a second calculation model, combined with preset financial health standard levels, is used to deeply assess the financial health of the target company, and the resulting second score value is used to characterize the company's financial robustness, further enhancing the comprehensiveness and depth of risk warning. Finally, a third calculation model is introduced to examine the risk characteristics of the target company from a group perspective, obtaining group identification results, greatly broadening the risk warning perspective and improving the ability to identify complex risks.

[0026] Based on the outputs of the first three models, this invention automatically generates risk warning signals according to preset signal mapping rules. It transforms the first score, second score, and group identification results into easily understandable and responsive signal levels, which are then applied to financial institutions in risk decision-making. The entire technical solution constructs an intelligent, dynamic, and multi-layered enterprise financial risk warning system. By integrating risk identification mechanisms from different dimensions, it significantly improves the accuracy, timeliness, and flexibility of the warning system, effectively overcoming the limitations of existing warning technologies. Attached Figure Description

[0027] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0028] Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing a financial risk early warning method based on multi-model fusion is shown.

[0029] Figure 2 This is a flowchart of an optional financial risk early warning method based on multi-model fusion according to an embodiment of the present invention;

[0030] Figure 3 This is a schematic diagram of an optional financial risk early warning device based on multi-model fusion according to an embodiment of the present invention;

[0031] Figure 4 This is a structural block diagram of an electronic device that performs a financial risk early warning method based on multi-model fusion according to an embodiment of the present invention. Detailed Implementation

[0032] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0033] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0034] It should be noted that the financial risk early warning method and device based on multi-model fusion in this application can be used in the field of fintech for credit risk assessment of enterprises, and can also be used in any field other than fintech for credit risk assessment of enterprises. This application does not limit the application field of the financial risk early warning method and device based on multi-model fusion.

[0035] It should be noted that all relevant information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, and displayed data) involved in this application are information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, processing, transmission, provision, disclosure, use, and handling of such data comply with the laws, regulations, and standards of the relevant regions, necessary confidentiality measures have been taken, and they do not violate public order and good morals. Corresponding access points are provided for users to choose to authorize or refuse access. For example, this system has interfaces with relevant users or organizations. Before obtaining relevant information, a request to obtain the information needs to be sent to the aforementioned user or organization through the interface, and the relevant information is obtained only after receiving consent from the aforementioned user or organization.

[0036] The information collection (e.g., user voice, video, and text collection) and analysis operations involved in this application have provided users with corresponding operation entry points during execution, allowing users to choose to agree to or reject the automated decision results; if the user chooses to reject, the process will proceed to the expert decision-making process.

[0037] The following embodiments of the present invention can be applied to various systems, applications, or devices that require early warning of corporate financial risks and credit decision support, enabling a dynamic financial risk early warning mechanism based on multi-model fusion. The present invention uses big data technology and machine learning algorithms to perform deep learning and analysis on financial data, and then constructs a dynamic triple early warning model, which can better capture subtle changes in corporate financial indicators, promptly detect potential financial risk signals, and identify high-quality customers in high-risk industries.

[0038] The present invention will now be described in detail with reference to various embodiments.

[0039] Example 1

[0040] According to an embodiment of the present invention, an embodiment of a financial risk early warning method based on multi-model fusion is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0041] The financial risk early warning method based on multi-model fusion provided in Embodiment 1 of the present invention can be executed on a mobile terminal, computer terminal or similar computing device. Figure 1 A hardware structure block diagram of a computer terminal (or mobile device) for implementing a multi-model fusion-based financial risk early warning method is shown. Figure 1 As shown, the computer terminal 10 (or mobile device) may include one or more processors 102 (shown as 102a, 102b, ..., 102n in the figure) 102 (processor 102 may include, but is not limited to, a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of a BUS bus), a network interface, a power supply, and / or a camera. Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the aforementioned electronic device. For example, computer terminal 10 may also include... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0042] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits are generally referred to herein as "data processing circuits". These data processing circuits may be embodied, in whole or in part, in software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuits may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the computer terminal 10 (or mobile device). As involved in the embodiments of this application, the data processing circuits serve as a processor control mechanism (e.g., selection of a variable resistor termination path connected to an interface).

[0043] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the financial risk early warning method based on multi-model fusion in this embodiment of the application. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned financial risk early warning method based on multi-model fusion. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to the computer terminal 10 via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0044] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the computer terminal 10. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0045] The display may be, for example, a touchscreen liquid crystal display (LCD) that allows the user to interact with the user interface of the computer terminal 10 (or mobile device).

[0046] Under the above operating environment, the present invention provides, as follows: Figure 2The financial risk early warning method shown is based on multi-model fusion. The main body of this method is the financial system or the financial risk early warning system of the lower-level non-retail enterprises. It combines big data analysis and machine learning technology and is used in the credit risk management scenario of financial institutions. In particular, it is used to solve the problem of early and accurate identification of financial risks in high-risk industries. It constructs a three-fold early warning model for industry segmentation, which includes financial feature calculation, abnormal change capture and dynamic indicator screening, and group risk identification steps of cluster analysis model. The goal is to improve the accuracy of early warning, shorten the early warning period, and effectively distinguish between risky and opportunity customers.

[0047] The embodiments of the present invention will now be described in detail with reference to each specific step.

[0048] Figure 2 This is a flowchart of an optional financial risk early warning method based on multi-model fusion according to an embodiment of the present invention, such as... Figure 2 As shown, the method includes the following steps:

[0049] Step S201: Obtain the financial statement data of the target company, and use statistical methods to filter financial indicators from the financial statement data to form a target indicator set.

[0050] In this embodiment, the target enterprise refers to a company or organization that needs to undergo a financial health assessment. These enterprises often belong to non-retail industries, have complex financial structures and business models, and require more sophisticated risk management methods. The financial statement data covers the annual financial reports provided by the target enterprise, including but not limited to: balance sheet, income statement, cash flow statement, etc., which are the core documents for assessing the enterprise's financial condition.

[0051] Financial indicators are key values ​​extracted from financial statement data, used to quantitatively analyze a company's financial health, such as indicators of asset quality, debt risk, profitability, and business growth. In this invention, statistical methods are used to screen a large number of financial indicators to select effective indicators highly correlated with industry characteristics, forming a target indicator set. This process not only considers changes in indicator values ​​but also incorporates industry standards and historical company data to ensure that the selected indicators accurately reflect the company's true financial situation.

[0052] It should be noted that this embodiment is based on the financial risk warning needs of the target enterprise. The financial statement data is obtained by the target enterprise actively providing financial statement data, or by capturing financial statement data within the scope of the data access permitted by the target enterprise. If there is no such need and data authorization, the above-mentioned financial risk warning method based on multi-model fusion will not be implemented.

[0053] Furthermore, the statistical methods in this embodiment specifically refer to techniques used for data screening and preprocessing, including: outlier detection, which involves identifying outliers in the data and eliminating inaccurate indicators that may be caused by data entry errors or special events; correlation analysis, which uses statistical methods, such as Pearson correlation coefficient or Spearman rank correlation, to analyze the relationship between various financial indicators and corporate risk, and select indicators that are highly correlated with risk; industry comparison, which involves comparing the financial indicators of the target company with those of other companies in the same industry to identify the special characteristics of the target company's financial performance and select differentiated indicators; and quantile analysis, which involves calculating the quantiles of financial indicators to determine the distribution range of the indicators, thereby identifying those values ​​that are in abnormal positions, which may indicate a deterioration in the company's financial situation.

[0054] As an example, suppose we analyze the financial situation of a manufacturing company. According to this embodiment, indicators such as the accounts receivable-to-current-assets ratio and current asset turnover ratio can be extracted from the company's annual report. Through industry comparison and percentile analysis, it is found that the company's accounts receivable-to-current-assets ratio is significantly higher than the industry 95th percentile, while its current asset turnover ratio is lower than the industry 5th percentile, indicating that the company faces a high risk of bad debts and liquidity problems. Therefore, these two indicators are selected into the target indicator set for subsequent risk assessment.

[0055] Optionally, in the financial risk early warning method based on multi-model fusion provided in this embodiment of the invention, the step of obtaining the financial statement data of the target enterprise includes: capturing the annual report text of the target enterprise from a designated enterprise information disclosure platform; parsing the annual report text to obtain the parsing result, wherein the parsing result includes at least the financial data items in the balance sheet, income statement and cash flow statement; and standardizing and formatting the financial data items in the parsing result to obtain the financial statement data.

[0056] It should be noted that the designated corporate information disclosure platforms refer to platforms that publicly release corporate annual financial information, such as the official websites of stock exchanges, professional financial information databases, and business registration and information disclosure systems. These platforms contain a large number of publicly available corporate annual reports and are the primary source of corporate financial data.

[0057] To further clarify, annual reports record comprehensive corporate information, including but not limited to: business overview, financial data, and management analysis. The process of parsing an annual report involves extracting standardized financial data items, such as accounts receivable from the balance sheet, total revenue from the income statement, and cash inflows from operating activities from the cash flow statement. The parsing results should clearly display the values ​​of these key data items.

[0058] Specifically, the balance sheet records a company's assets, liabilities, and owner's equity at a specific point in time, and is used to demonstrate the company's financial condition; the income statement records a company's revenue, costs, and profits over a certain period, and is used to reflect the company's profitability; and the cash flow statement records the inflow and outflow of cash and cash equivalents over a certain period, and is used to demonstrate the company's cash flow status.

[0059] The difference between financial data items and financial indicators lies in their numerical values. Financial data items are the values ​​directly reflected in financial statements, such as the amount of accounts receivable. Financial indicators, on the other hand, are values ​​with specific financial meanings derived from the calculation and analysis of financial data items, such as accounts receivable turnover. Financial indicators are derived from financial data items and are used to gain a deeper understanding of a company's financial health.

[0060] Regarding data processing, standardization refers to transforming financial data from different sources and in different formats into a unified numerical standard. For example, this involves processing all companies' financial data according to time, currency unit, and accounting standards to ensure data comparability and consistency. Formatting, on the other hand, involves adjusting the data format to suit the needs of subsequent data processing and model training. For example, converting data into tabular form or performing data cleaning to remove outliers and missing values.

[0061] For example, standardization processes require converting data from different accounting standards to International Financial Reporting Standards (IFRS) or Generally Accepted Accounting Principles (GAAP) to achieve global comparability. Formatting processes include converting all values ​​to the same number of decimal places or converting data into a format acceptable to specific machine learning models.

[0062] In addition to the steps mentioned above, supplementary data processing operations may include: data cleaning, i.e., removing obviously erroneous or inconsistent financial data; outlier detection, ensuring the quality of data used in the early warning model; and data validation, confirming the accuracy and completeness of the data by comparing multiple data sources. These supplementary operations can further improve data quality, thereby enhancing the accuracy and reliability of the early warning model.

[0063] Optionally, in the financial risk early warning method based on multi-model fusion provided in this embodiment of the invention, the step of selecting financial indicators from financial statement data based on statistical methods to form a target indicator set includes: determining all companies within the sub-sector to which the target company belongs as peer companies, and obtaining the financial statement data of the peer companies to obtain peer financial data; determining the peer financial distribution based on the peer financial data, and determining the quantile ratio based on the peer financial distribution, wherein the quantile ratio includes at least the first quantile ratio and the second quantile ratio, and the first quantile ratio is less than the second quantile ratio; using the first quantile ratio and the second quantile ratio as boundaries, selecting financial data items in the target company's financial statement data that are outside the boundaries; and constructing a target indicator set based on all selected financial data items.

[0064] It should be noted that the target company's sub-sector refers to the specific business area or market segment in which the company operates, such as sub-sectors like machinery, chemicals, and building materials within the manufacturing industry. Sub-sectors typically share similar business models, market environments, and financial characteristics. "Competitor companies" refers to other companies within the same sub-sector as the target company. The financial data of these competitors can be used as a benchmark, representing standard behavioral patterns within the industry, to identify abnormal changes in the target company's financial condition.

[0065] The financial statement data of peer companies in this embodiment of the invention are mainly obtained through legal and compliant information disclosure platforms, such as stock exchange websites, annual reports officially released by companies, and business data reports provided by relevant departments.

[0066] The first step, based on the acquired peer financial data, is to determine the peer financial distribution. This process involves statistical analysis of the data to identify the distribution of various financial indicators within the entire industry. Statistical graphs such as histograms, kernel density estimation, or box-and-whisker plots can be used to visually represent the distribution characteristics.

[0067] Then, based on the distribution, the quantile ratio is determined. For example, the first quantile ratio could be 5%, meaning that only 5% of the peer companies have financial data below this threshold; while the second quantile ratio could be set at 95%, indicating that only 5% of the peer companies have financial data above this threshold. This setting is used to capture extreme financial performance at both ends of the industry, and then analyze whether the target company is in these extreme abnormal states.

[0068] Furthermore, screening financial data items outside the quantile threshold in the target company's financial statements helps identify financial indicators that appear abnormal compared to the industry average. This screening helps focus on indicators that foreshadow a deterioration or improvement in the company's financial health, allowing for the use of more predictive financial indicators in subsequent model building. For example, if a target company's accounts receivable turnover ratio is significantly lower than the first quantile of its peers, it indicates potential problems with debt collection and requires close monitoring.

[0069] In this embodiment of the invention, the target indicator set serves as the input for constructing the early warning model. It includes all financial data items outside the percentile limits in the financial statement data. By comparing the performance of the target company with that of its peers on the selected financial indicators, it is determined whether the target company faces a high level of financial risk.

[0070] Step S202: Based on the target indicator set, calculate the degree of deviation of the target enterprise from the industry level indicators through the first calculation model to obtain the first score value.

[0071] It should be noted that the deviation of a target company from industry-level indicators refers to the difference between the target company's financial indicators (such as accounts receivable turnover, current ratio, etc.) and the average or typical values ​​of its sub-sector. This is used to identify abnormal financial performance that may indicate risks.

[0072] For example, if a target company's accounts receivable turnover ratio is significantly lower than the industry average, it indicates that the company faces a higher risk of bad debts and needs to conduct more rigorous reviews of its customers' creditworthiness. Conversely, if a target company's current ratio is significantly higher than the industry average, it may mean that the company has good short-term solvency, but it may also indicate that its capital utilization is inefficient and that too many current assets are not being used effectively.

[0073] The main functions of assessing this deviation in this embodiment of the invention are twofold: risk warning, which can promptly identify financial anomalies of enterprises and warn of potential credit risks, thereby enabling early intervention to mitigate or avoid losses; and credit strategy optimization, which helps financial institutions to better understand the risk characteristics of different industries, adjust credit strategies, and accurately allocate credit resources, including identifying high-quality customers in high-risk industries and avoiding a one-size-fits-all approach to credit.

[0074] Optionally, in the financial risk early warning method based on multi-model fusion provided in this embodiment of the invention, the step of calculating the deviation of the target enterprise from the industry level indicators through the first calculation model to obtain the first score value includes: for each financial indicator in the target indicator set, calculating the quantile point based on the peer financial distribution and quantile ratio, wherein the quantile point includes at least a first value point and a second value point; calculating the median based on the quantile point, and calculating the absolute deviation rate of each financial indicator from the median, wherein the median represents the industry level of the sub-sector to which it belongs; and weighting and summing the absolute deviation rates corresponding to all financial indicators according to the first preset weight to obtain the first score value, wherein the first preset weight is obtained by training the first calculation model based on historical financial data.

[0075] The aforementioned technical objective is to provide early warnings of financial risks for non-retail enterprises through multi-model fusion, improving the accuracy and timeliness of risk warnings, while distinguishing between risks and opportunities, thus providing financial institutions with a more scientific basis for credit decisions. Within this framework, the goal of the first calculation model is to quantify the degree of deviation of the target enterprise from industry-level indicators, obtaining a first score value as a preliminary quantitative indicator for risk warning.

[0076] The first calculation model is a statistical or machine learning model used to calculate the degree of deviation of a target company's financial indicators from the industry average. This model can be improved based on traditional statistical analysis models (such as quantile analysis and standard deviation analysis) to adapt to more complex industry characteristic analysis. By introducing machine learning techniques (such as support vector regression and random forest) to learn the nonlinear relationships in historical financial data, the impact of indicator deviation on risk can be assessed more accurately.

[0077] For each financial indicator in the target indicator set, such as accounts receivable turnover ratio and current ratio, the model first calculates the first and second value points based on the financial distribution of peers and using previously determined quantile ratios (such as 5% and 95%). These two value points represent extremely low and extremely high financial performance in the industry, respectively, and are used to define the abnormal range of the indicator.

[0078] The process of calculating the median based on quantiles involves statistically analyzing the values ​​of all companies in the same industry for a specific financial indicator and finding the value in the middle as the median. That is, half of the companies are below this value, and the other half are above it. This value represents the industry average for its specific sub-sector and serves as a benchmark for measuring the degree of deviation of the target company's indicator.

[0079] Calculating the absolute deviation rate of each financial indicator from the median involves comparing the target company's financial indicator value with the industry average (median), calculating the absolute value of the difference between the target company's indicator value and the median, and then dividing it by the median. The resulting ratio (such as 10%, 20%, etc.) is used to reflect the degree of deviation between the target company's indicator and the industry average.

[0080] Furthermore, the first preset weight is used to characterize the importance of each financial indicator in the target indicator set for the overall risk assessment. The weight is determined based on the analysis of the contribution of each indicator to risk warning during the model training process and is derived from historical financial data. This weight is used to reflect the differences in the risk warning value of different financial indicators in specific industries and risk environments, thereby ensuring that the model can be more comprehensive and accurate in assessing corporate risks.

[0081] The weighted summation of the absolute deviation rates corresponding to all financial indicators involves multiplying the absolute deviation rate of each financial indicator of the target company from the median by the corresponding first preset weight, and then summing all the products to obtain the first score value, which is used to comprehensively reflect the degree of risk deviation of the target company on multiple financial indicators.

[0082] Step S203: Based on the target indicator set, the financial health of the target enterprise is calculated using the second calculation model and preset financial health standard levels to obtain the second score.

[0083] It's important to clarify that the preset financial health standard tiers are a series of gradations established based on industry financial standards and historical data to assess a company's financial health. Each tier represents a different level of financial health. For example, a standard tier system might include five levels from "healthiest" to "least healthy," with each level corresponding to a specific set of financial indicator thresholds. When a target company's financial indicators reach or exceed the threshold of a certain tier, the company is classified into the corresponding health tier.

[0084] For example, in this embodiment of the invention, the financial health standard levels can be set as follows: Level 1 (healthiest), all key financial indicators such as current ratio and net profit margin are at the best level in the industry; Level 2, financial indicators are slightly below the best level but still within the healthy range; Level 3, financial indicators begin to show small-scale anomalies but are still within a controllable range; Level 4, financial indicators deviate significantly from the industry average and there is a certain financial risk; Level 5 (unhealthiest), financial indicators deteriorate severely and the company faces high risks.

[0085] The financial health of the target company is a comprehensive evaluation indicator. Based on various financial indicators in the target indicator set, a second score is obtained through quantitative evaluation using a second calculation model. In simple terms, it scores the strength of the company's financial situation; a higher score indicates a better financial health, and vice versa.

[0086] Assessing financial health serves several purposes. It not only identifies potential financial risks within a company but also quantifies their severity, providing financial institutions with more nuanced risk assessments to facilitate credit decisions and risk management. Furthermore, it helps identify potential clients in high-risk industries—companies operating in high-risk sectors but possessing sound financial health and promising growth prospects.

[0087] Optionally, in the financial risk early warning method based on multi-model fusion provided in this embodiment of the invention, the step of calculating the financial health of the target enterprise and obtaining a second score value through a second calculation model and preset financial health standard levels includes: determining the evaluation level corresponding to each financial indicator based on the specified enterprise performance evaluation standard value, wherein the evaluation level includes: excellent value, good value, average value, low value, and poor value; interpolating the actual value of each financial indicator of the target enterprise with the standard value of the corresponding evaluation level to obtain the level score of each financial indicator; and weighting the level scores of each financial indicator according to a second preset weight to obtain the second score value, wherein the second preset weight is obtained by training the second calculation model based on historical financial data.

[0088] It should be noted that the specified corporate performance evaluation standard value refers to the expected or standard range of financial indicators determined by experts or through data analysis based on industry characteristics and historical data. It is usually based on the industry average level, and different corporate performance levels such as excellent, good, average, low, and poor. For example, for the current ratio, the industry average might be 1.5, while an excellent value could be set above 2.0, and a poor value below 1.0.

[0089] The specific steps for determining the evaluation level include: statistically analyzing the historical data of each financial indicator to identify the boundary values ​​of different performance levels; using these boundary values ​​as reference points for financial health scoring; and classifying all companies in the industry into five levels—excellent, good, average, low, and poor—based on the actual values ​​of their financial indicators.

[0090] Furthermore, interpolation is a mathematical method used to estimate unknown values ​​between known numerical points. In this context, interpolation is primarily used to map the actual values ​​of a target company's financial indicators to rating levels. For example, if the actual value of an indicator falls between a good value and the average value, interpolation can be used to give a score between the good value and the average value, thereby quantifying the indicator's performance.

[0091] For example, in practice, suppose that in the current ratio tier scoring, a good value corresponds to a score of 80, and an average value corresponds to a score of 60. If the target company's actual current ratio is 1.6, which is between the good value of 1.5 and the average value of 1.2, then an interpolated score can be calculated based on its relative distance from the two tier boundary values ​​to ensure the continuity and objectivity of the scoring.

[0092] Furthermore, the second pre-defined weight is used to characterize the contribution of each financial indicator in the target indicator set to the overall financial health assessment. The weights are determined based on the model's training process, that is, by learning from historical financial data to identify which financial indicators are more important for judging a company's financial health and assigning them higher weights. The weighted averaging operation ensures that those financial indicators more important for the overall health assessment have a larger proportion in the final score, thereby improving the accuracy and effectiveness of the assessment.

[0093] Step S204: Based on the target indicator set, the third calculation model is used to determine whether the target enterprise matches the preset potential risk group rule set, and the group identification result is obtained.

[0094] It should be noted that the pre-defined potential risk group rule set is a data structure containing multiple rule sets, predefined based on historical data and industry characteristics. It is used to identify specific types of risk characteristics, particularly targeting groups of companies that may exhibit financial instability. Each rule in the rule set is a condition set for a set of financial indicators. If a target company's performance on these indicators meets the rule conditions, it is considered to have "hit" that rule and is thus classified into the potential risk group.

[0095] For example, one rule could be: "Accounts receivable turnover days are more than twice the industry average, and net operating cash flow has been negative for two consecutive years." If a target company's financial data meets the above conditions, it is considered to have met this rule and is a member of the potential risk group.

[0096] Determining whether a target company falls within a pre-defined set of rules for potential risk groups means checking whether its financial indicators meet the conditions of any one or more rules within the set. This is equivalent to conducting a risk screening of the target company, identifying any specific patterns or group phenomena that foreshadow financial risk. For financial institutions, the results of group identification not only supplement the limitations of individual company risk assessment but also help companies discover early signs of systemic risk, providing a basis for the formulation of risk management strategies.

[0097] Assessing group identification results helps financial institutions evaluate industry risk from a macro perspective, focusing not only on the risk status of individual companies but also on insights into group risk trends within the industry. For example, if multiple companies simultaneously meet pre-defined potential risk criteria, it may indicate widespread financial pressure or adverse economic cycles across the industry, helping financial institutions adjust their credit strategies, such as tightening credit standards or identifying specific companies with stronger risk resistance.

[0098] Optionally, in the financial risk early warning method based on multi-model fusion provided in this embodiment of the invention, the step of determining whether the target enterprise hits the preset potential risk group rule set through a third calculation model to obtain the group identification result includes: obtaining the preset potential risk group rule set, wherein the preset potential risk group rule set records N financial risk feature patterns, and N is a preset value; analyzing the target indicator set of the target enterprise to obtain the corresponding financial indicator distribution pattern; matching the financial indicator distribution pattern with each financial risk feature pattern in the preset potential risk group rule set and calculating the matching degree; setting the group identification result to a hit state when any matching degree is higher than or equal to a preset matching degree threshold; and setting the group identification result to a miss state when all matching degrees are lower than the preset matching degree threshold.

[0099] In this embodiment, the preset potential risk group rule set includes multiple predefined financial risk characteristic patterns. Each pattern is constructed based on the common characteristics of high-risk enterprise groups shown in historical data, such as high debt ratio, low cash flow, and abnormal profit growth patterns.

[0100] The target company's financial indicators need to undergo preprocessing and feature extraction to form a distribution pattern reflecting its financial condition, including the values, trends, and distribution of each financial indicator. Matching is achieved by comparing the target company's financial indicator distribution pattern with each risk pattern in the preset rule set to check for similar combinations of financial characteristics. It should be noted that this matching is not limited to direct numerical comparison but can also include the trends and distribution patterns of the indicators.

[0101] Furthermore, the calculation of the matching degree is essentially a quantification of the similarity between the target company and each risk pattern. This can be achieved based on distance calculations between indicators (such as Euclidean distance, Manhattan distance), correlation coefficients, or more complex confidence scores from machine learning classifiers. A high matching degree indicates that the financial characteristics of the target company are highly consistent with the specific risk pattern, suggesting a higher probability of risk.

[0102] The setting of the preset matching threshold depends on the dataset used to train the model and the target accuracy of risk identification. It is a parameter used to distinguish between high risk and low risk, and is usually optimized by the model's performance on the training and validation sets to ensure that risk warnings are both accurate and timely.

[0103] Step S205: Based on the preset signal mapping rules, a risk warning signal is generated based on the first score value, the second score value, and the group identification result.

[0104] In this embodiment, the preset signal mapping rule serves as a bridge connecting the model output and the final warning signal, ensuring that the warning signal not only reflects the enterprise's risk level but also guides financial institutions to take appropriate risk management actions. For example, if the target enterprise triggers all three models simultaneously, and both the first and second score values ​​far exceed the industry average, the group identification result is at the highest risk level. In this case, a red warning signal will be issued, indicating that the enterprise faces extremely high risk and requires immediate action, such as freezing loan amounts or activating emergency response plans.

[0105] If the target company only triggers the abnormal change capture model and has a high first score, but the second score and the group identification results are still within a controllable range, a yellow warning signal will be issued, prompting financial institutions to pay attention to the company and strengthen monitoring, but there is no need to take aggressive measures immediately.

[0106] If no early warning model is triggered, or if the triggered early warning model, together with the first score, the second score, and the group identification results, indicates that the company's financial situation is stable and the risk is low, the system may not issue an early warning signal, or may only provide information on the green and healthy status, indicating that the company's current financial risk is low.

[0107] Optionally, in the financial risk early warning method based on multi-model fusion provided in this embodiment of the invention, the step of generating a risk early warning signal based on a first score value, a second score value, and a group identification result according to a preset signal mapping rule includes: setting a first judgment condition for a first score value exceeding a first preset threshold, setting a second judgment condition for a second score value lower than a second preset threshold, and setting a third judgment condition for a group identification result being in a hit state; generating a first type of risk early warning signal when the third judgment condition is met and either the first or second judgment condition is met; generating a second type of risk early warning signal when only the third judgment condition is met, or when only the first and second judgment conditions are met simultaneously; generating a third type of risk early warning signal when only the first judgment condition is met, or when only the second judgment condition is met; and generating a fourth type of risk early warning signal when none of the judgment conditions are met.

[0108] It should be noted that the first preset threshold and the first judgment condition are set based on industry characteristics and risk appetite, and are used to determine whether the first score value output by the abnormal change capture model exceeds the warning threshold. When the first score value exceeds the first preset threshold, it indicates that the target company has significant risks in terms of abnormal changes in financial indicators. That is, compared with other companies in the industry, the target company has experienced unusual deterioration or fluctuation in its financial situation, reminding the system that the company may face impending or recently occurring financial problems, requiring close attention from financial institutions.

[0109] The second preset threshold and the second judgment condition are also set based on industry characteristics and risk appetite, but focus on the second score. When the second score is lower than the second preset threshold, it indicates that the company's financial health is below the industry standard, and there may be long-term financial problems or potential risks. The second judgment condition places more emphasis on the company's long-term health relative to the industry standard.

[0110] The third criterion for determining the group identification result involves the target company's matching status within the pre-defined set of rules for potential risk groups. If the group identification result is a "hit" state, meaning the target company matches at least one rule in the high-risk characteristic pattern, it indicates that the company may belong to a group with higher financial risk, and its risk level needs further assessment.

[0111] In some specific implementation scenarios, assuming the first preset threshold is set at 70 points, if the target company obtains a first score of 75 points through the abnormal change capture model, it indicates that the changes in the company's financial indicators have exceeded the normal range, and there may be short-term or impending financial risks. Assuming the second preset threshold is set at 60 points, if the target company obtains a second score of only 55 points, it indicates that the company's financial health is below the industry average, and its long-term financial stability is worrying. If the target company is identified as a member of a risk group in the preset potential risk group rule set, it means that, based on its financial indicator distribution pattern, the company may have characteristics similar to historical high-risk cases.

[0112] In this embodiment, the logic for generating the warning signal is as follows: if the third judgment condition is met, and either the first or second judgment condition is met, it indicates that the enterprise is in a high-risk state, exhibiting short-term financial anomalies and a long-term financial health level below industry standards, placing it in the high-risk group. Therefore, a first-type risk warning signal is generated, representing a red warning, the highest risk level. Financial institutions should take immediate action, such as strengthening monitoring or prematurely recalling loans. Other warning signals can also be used, such as emergency email notifications, telephone warnings, or system highlighting.

[0113] If only the third condition is met, or only the first and second conditions are met simultaneously, it means that the company either has financial indicators similar to those of a high-risk group, or faces risks in both short-term financial changes and long-term health, but the degree of risk is relatively low. Therefore, a second type of risk warning signal is generated, representing a yellow warning, i.e., a medium-risk level. Financial institutions need to be vigilant and monitor the company's dynamics, but this is not an emergency. Signal reminders can be provided through system notifications, regular email updates, etc.

[0114] If only the first or second condition is met, it indicates that the company has either experienced abnormal changes in its recent financial indicators or its financial health is below industry standards, but does not simultaneously meet the criteria for identifying a high-risk group, suggesting a relatively singular risk. Therefore, a third type of risk warning signal is generated, representing an orange alert, i.e., a medium-to-high risk level, requiring continuous monitoring by financial institutions and the preparation of corresponding risk management measures. This can be achieved through system-wide alert icons, follow-up teleconferences, and other signal reminders.

[0115] If all the judgment conditions are not met, it means that the company's financial situation is stable, it does not show significant short-term or long-term risks, and it is not identified as part of a high-risk group. Therefore, a fourth type of risk warning signal is generated, representing a green warning, i.e., a low-risk level. Financial institutions can maintain normal monitoring frequency without additional warnings. Signals such as those without special markings within the system or regular financial status reports can be used for alerts.

[0116] Through the above steps S201 to S205, the financial statement data of the target company can be obtained first, and financial indicators can be screened from the financial statement data based on statistical methods to form a target indicator set. Then, based on the target indicator set, the deviation of the target company from the industry level indicators is calculated by the first calculation model to obtain the first score value. Then, based on the target indicator set, the financial health of the target company is calculated by the second calculation model and the preset financial health standard level to obtain the second score value. Then, based on the target indicator set, the third calculation model is used to determine whether the target company hits the preset potential risk group rule set to obtain the group identification result. Finally, according to the preset signal mapping rule, a risk warning signal is generated based on the first score value, the second score value and the group identification result.

[0117] In this embodiment of the invention, a multi-model fusion approach is adopted. Through statistical screening, dynamic deviation calculation, quantitative assessment of financial health, and matching of group risk characteristics, the goal of comprehensively and accurately identifying corporate financial risks is achieved. This results in improving the accuracy of early warning, shortening the lead time, and effectively distinguishing between risks and opportunities. It also solves the technical problem that the financial risk early warning mechanism in related technologies has defects, leading to low overall efficiency in financial risk identification.

[0118] Specifically, this embodiment of the invention first uses statistical methods to screen financial indicators from the financial statement data of the target company, constructing a highly focused set of target indicators. Then, it uses a first calculation model to calculate the degree of deviation of the target company's financial indicators from the industry level, obtaining a first score value to intuitively reflect the dynamic trend of the company's financial situation. Subsequently, a second calculation model, combined with preset financial health standard levels, is used to deeply assess the financial health of the target company, and the resulting second score value is used to characterize the company's financial stability, further enhancing the comprehensiveness and depth of risk warning. Finally, a third calculation model is introduced to examine the risk characteristics of the target company from a group perspective, obtaining group identification results, greatly broadening the risk warning perspective and improving the ability to identify complex risks.

[0119] Based on the outputs of the first three models, this embodiment of the invention automatically generates risk warning signals according to preset signal mapping rules. It transforms the first score, the second score, and the group identification results into easily understandable and responsive signal levels, which are then applied to financial institutions in risk decision-making. The entire technical solution constructs an intelligent, dynamic, and multi-layered enterprise financial risk warning system. By integrating risk identification mechanisms from different dimensions, it significantly improves the accuracy, timeliness, and flexibility of the warning system, effectively overcoming the limitations of existing warning technologies.

[0120] The invention will now be described in conjunction with another alternative embodiment.

[0121] Example 2

[0122] This invention also provides a financial risk early warning device based on multi-model fusion. It should be noted that the financial risk early warning device based on multi-model fusion in this invention includes multiple implementation units, which can be used to execute the financial risk early warning method based on multi-model fusion provided in the first embodiment above. Each implementation unit corresponds to each implementation step in the first embodiment above.

[0123] Figure 3 This is a schematic diagram of an optional financial risk early warning device based on multi-model fusion according to an embodiment of the present invention, such as... Figure 3 As shown, the device may include: an acquisition unit 31, a first calculation unit 32, a second calculation unit 33, a judgment unit 34, and a generation unit 35.

[0124] Among them, the acquisition unit 31 is used to acquire the financial statement data of the target enterprise and filter financial indicators from the financial statement data based on statistical methods to form a target indicator set.

[0125] The first calculation unit 32 is used to calculate the degree of deviation of the target enterprise from the industry level indicators based on the target indicator set and through the first calculation model to obtain the first score value.

[0126] The second calculation unit 33 is used to calculate the financial health of the target enterprise based on the target indicator set, through the second calculation model and the preset financial health standard level, and obtain the second score value.

[0127] Judgment unit 34 is used to determine whether the target enterprise hits the preset potential risk group rule set based on the target indicator set and through the third calculation model, so as to obtain the group identification result.

[0128] The generation unit 35 is used to generate a risk warning signal based on the first score value, the second score value and the group identification result according to the preset signal mapping rules.

[0129] The aforementioned financial risk early warning device based on multi-model fusion can first acquire the financial statement data of the target enterprise through the acquisition unit 31, and then screen financial indicators from the financial statement data based on statistical methods to form a target indicator set. Then, the first calculation unit 32 calculates the deviation of the target enterprise from the industry level indicators based on the target indicator set and the first calculation model to obtain a first score value. Then, the second calculation unit 33 calculates the financial health of the target enterprise based on the target indicator set, the second calculation model and the preset financial health standard level to obtain a second score value. Then, the judgment unit 34 judges whether the target enterprise hits the preset potential risk group rule set based on the target indicator set and the third calculation model to obtain the group identification result. Finally, the generation unit 35 generates a risk early warning signal based on the first score value, the second score value and the group identification result according to the preset signal mapping rules.

[0130] In this embodiment of the invention, a multi-model fusion approach is adopted. Through statistical screening, dynamic deviation calculation, quantitative assessment of financial health, and matching of group risk characteristics, the goal of comprehensively and accurately identifying corporate financial risks is achieved. This results in improving the accuracy of early warning, shortening the lead time, and effectively distinguishing between risks and opportunities. It also solves the technical problem that the financial risk early warning mechanism in related technologies has defects, leading to low overall efficiency in financial risk identification.

[0131] Specifically, this embodiment of the invention first uses statistical methods to screen financial indicators from the financial statement data of the target company, constructing a highly focused set of target indicators. Then, it uses a first calculation model to calculate the degree of deviation of the target company's financial indicators from the industry level, obtaining a first score value to intuitively reflect the dynamic trend of the company's financial situation. Subsequently, a second calculation model, combined with preset financial health standard levels, is used to deeply assess the financial health of the target company, and the resulting second score value is used to characterize the company's financial stability, further enhancing the comprehensiveness and depth of risk warning. Finally, a third calculation model is introduced to examine the risk characteristics of the target company from a group perspective, obtaining group identification results, greatly broadening the risk warning perspective and improving the ability to identify complex risks.

[0132] Based on the outputs of the first three models, this embodiment of the invention automatically generates risk warning signals according to preset signal mapping rules. It transforms the first score, the second score, and the group identification results into easily understandable and responsive signal levels, which are then applied to financial institutions in risk decision-making. The entire technical solution constructs an intelligent, dynamic, and multi-layered enterprise financial risk warning system. By integrating risk identification mechanisms from different dimensions, it significantly improves the accuracy, timeliness, and flexibility of the warning system, effectively overcoming the limitations of existing warning technologies.

[0133] Furthermore, the acquisition unit includes: a crawling module, used to crawl the annual report text of the target company from a specified enterprise information disclosure platform; a parsing module, used to parse the annual report text to obtain parsing results, wherein the parsing results include at least: financial data items from the balance sheet, income statement and cash flow statement; and a processing module, used to standardize and format the financial data items in the parsing results to obtain financial statement data.

[0134] Furthermore, the acquisition unit also includes: a first acquisition module, used to determine all companies within the target company's sub-sector as peer companies, and acquire the financial statement data of the peer companies to obtain peer financial data; a first determination module, used to determine the peer financial distribution based on the peer financial data, and determine the quantile ratio based on the peer financial distribution, wherein the quantile ratio includes at least the first quantile ratio and the second quantile ratio, and the first quantile ratio is less than the second quantile ratio; a screening module, used to screen the financial data items in the target company's financial statement data that are outside the boundaries of the first quantile ratio and the second quantile ratio; and a construction module, used to construct a target indicator set based on all the screened financial data items.

[0135] Further, the first calculation unit includes: a first calculation module, used to calculate a quantile for each financial indicator in the target indicator set based on the peer financial distribution and quantile ratio, wherein the quantile includes at least a first numerical point and a second numerical point; a second calculation module, used to calculate the median based on the quantile and calculate the absolute deviation rate of each financial indicator from the median, wherein the median represents the industry level of the sub-sector to which it belongs; and a weighted summation module, used to perform a weighted summation of the absolute deviation rates corresponding to all financial indicators according to a first preset weight to obtain a first score value, wherein the first preset weight is obtained by training a first calculation model based on historical financial data.

[0136] Furthermore, the second calculation unit includes: a second determination module, used to determine the evaluation level corresponding to each financial indicator based on the specified enterprise performance evaluation standard value, wherein the evaluation level includes: excellent value, good value, average value, low value, and poor value; a third calculation module, used to interpolate the actual value of each financial indicator of the target enterprise with the standard value of the corresponding evaluation level to obtain the level score of each financial indicator; and a weighted average module, used to perform a weighted average of the level scores of each financial indicator based on a second preset weight to obtain a second score value, wherein the second preset weight is obtained by training the second calculation model based on historical financial data.

[0137] Furthermore, the judgment unit includes: a second acquisition module, used to acquire a preset potential risk group rule set, wherein the preset potential risk group rule set records N financial risk characteristic patterns, and N is a preset value; an analysis module, used to analyze the target enterprise's target indicator set to obtain the corresponding financial indicator distribution pattern; a matching module, used to match the financial indicator distribution pattern with each financial risk characteristic pattern in the preset potential risk group rule set and calculate the matching degree; a first setting module, used to set the group identification result to a hit state when any matching degree is higher than or equal to a preset matching degree threshold; and a second setting module, used to set the group identification result to a miss state when all matching degrees are lower than the preset matching degree threshold.

[0138] Furthermore, the generation unit includes: a third setting module, used to set a first judgment condition as the first score value exceeding a first preset threshold, a second judgment condition as the second score value being lower than a second preset threshold, and a third judgment condition as the group identification result being a hit state; a first generation module, used to generate a first type of risk warning signal when the third judgment condition is met and either the first or second judgment condition is met; a second generation module, used to generate a second type of risk warning signal when only the third judgment condition is met, or when only the first and second judgment conditions are met simultaneously; a third generation module, used to generate a third type of risk warning signal when only the first judgment condition is met, or when only the second judgment condition is met; and a fourth generation module, used to generate a fourth type of risk warning signal when all judgment conditions are not met.

[0139] It should be noted that the aforementioned acquisition unit 31, first calculation unit 32, second calculation unit 33, judgment unit 34, and generation unit 35 correspond to steps S201 to S205 in Embodiment 1. The instances and application scenarios implemented by the aforementioned units and corresponding steps are the same, but are not limited to the content disclosed in Embodiment 1. It should be noted that the aforementioned modules or units may be hardware or software components stored in a memory (e.g., memory 104) and processed by one or more processors (e.g., processors 102a, 102b, ..., 102n). The aforementioned modules or units may also be part of a device and can run in the computer terminal 10 provided in Embodiment 1.

[0140] The invention will now be described in conjunction with another alternative embodiment.

[0141] Example 3

[0142] The present invention can also provide an electronic device. Figure 4This is a structural block diagram of an electronic device that performs a financial risk early warning method based on multi-model fusion according to an embodiment of the present invention, such as... Figure 4 As shown, the electronic device may include: one or more ( Figure 4 (Only one is shown) Processor 702, memory 704, memory controller, and peripheral interface, wherein the peripheral interface is connected to the radio frequency module, audio module and display.

[0143] The memory can be used to store software programs and modules, such as the program instructions / modules corresponding to the financial risk early warning method and device based on multi-model fusion in this application embodiment. The processor executes various functional applications and data processing by running the software programs and modules stored in the memory, thereby realizing the aforementioned financial risk early warning method based on multi-model fusion. The memory may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory may further include memory remotely located relative to the processor, and these remote memories can be connected to the terminal via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0144] The processor can access information and applications stored in the memory via a transmission device to perform the following steps: acquire the financial statement data of the target company, and select financial indicators from the financial statement data based on statistical methods to form a target indicator set; based on the target indicator set, calculate the degree of deviation of the target company from the industry level indicators using a first calculation model to obtain a first score value; based on the target indicator set, calculate the financial health of the target company using a second calculation model and a preset financial health standard level to obtain a second score value; based on the target indicator set, determine whether the target company hits a preset potential risk group rule set using a third calculation model to obtain a group identification result; and generate a risk warning signal based on the first score value, the second score value, and the group identification result according to a preset signal mapping rule.

[0145] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: retrieve the annual report text of the target company from a designated enterprise information disclosure platform; parse the annual report text to obtain the parsing results, wherein the parsing results include at least the financial data items in the balance sheet, income statement, and cash flow statement; and standardize and format the financial data items in the parsing results to obtain financial statement data.

[0146] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: identifying all companies within the target company's sub-sector as peer companies and obtaining their financial statement data to obtain peer financial data; determining the peer financial distribution based on the peer financial data and determining the quantile ratio according to the peer financial distribution, wherein the quantile ratio includes at least the first quantile ratio and the second quantile ratio, and the first quantile ratio is less than the second quantile ratio; using the first quantile ratio and the second quantile ratio as boundaries, filtering out financial data items outside the boundaries in the target company's financial statement data; and constructing a target indicator set based on all filtered financial data items.

[0147] The processor can also call the information and application stored in the memory through the transmission device to perform the following steps: For each financial indicator in the target indicator set, calculate the quantile point based on the peer financial distribution and quantile ratio, wherein the quantile point includes at least a first value point and a second value point; calculate the median based on the quantile point, and calculate the absolute deviation rate of each financial indicator from the median, wherein the median represents the industry level of the sub-sector to which it belongs; and perform a weighted summation of the absolute deviation rates corresponding to all financial indicators according to a first preset weight to obtain a first score value, wherein the first preset weight is obtained by training a first calculation model based on historical financial data.

[0148] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: determining the evaluation level corresponding to each financial indicator based on the specified enterprise performance evaluation standard value, wherein the evaluation level includes: excellent value, good value, average value, low value, and poor value; interpolating the actual value of each financial indicator of the target enterprise with the standard value of the corresponding evaluation level to obtain the level score of each financial indicator; and performing a weighted average of the level scores of each financial indicator based on a second preset weight to obtain a second score value, wherein the second preset weight is obtained by training a second calculation model based on historical financial data.

[0149] The processor can also access information and applications stored in the memory via a transmission device to perform the following steps: obtaining a preset set of rules for potential risk groups, wherein the preset set of rules for potential risk groups records N financial risk characteristic patterns, where N is a preset value; analyzing the target enterprise's target indicator set to obtain the corresponding financial indicator distribution pattern; matching the financial indicator distribution pattern with each financial risk characteristic pattern in the preset set of rules for potential risk groups and calculating the matching degree; setting the group identification result to a hit state if any matching degree is higher than or equal to a preset matching degree threshold; and setting the group identification result to a miss state if all matching degrees are lower than the preset matching degree threshold.

[0150] The processor can also call the information and application program stored in the memory through the transmission device to perform the following steps: setting a first judgment condition as the first score value exceeding a first preset threshold, setting a second judgment condition as the second score value being lower than a second preset threshold, and setting a group identification result as a hit state as the third judgment condition; generating a first type of risk warning signal when the third judgment condition is met and either the first or second judgment condition is met; generating a second type of risk warning signal when only the third judgment condition is met, or when only the first and second judgment conditions are met simultaneously; generating a third type of risk warning signal when only the first judgment condition is met, or when only the second judgment condition is met; and generating a fourth type of risk warning signal when all judgment conditions are not met.

[0151] This invention provides a financial risk early warning scheme based on multi-model fusion. By employing multi-model fusion, through statistical screening, dynamic deviation calculation, quantitative assessment of financial health, and matching of group risk characteristics, it achieves the goal of comprehensively and accurately identifying corporate financial risks. This improves early warning accuracy, shortens lead time, and effectively distinguishes between risks and opportunities, thereby solving the technical problem of low overall efficiency in financial risk identification due to deficiencies in existing financial risk early warning mechanisms.

[0152] Those skilled in the art will understand that Figure 4 The structure shown is for illustrative purposes only. Electronic devices can also be smartphones, tablets, handheld computers, mobile internet devices (MIDs), PADs, and other terminal devices. Figure 4 This does not limit the structure of the aforementioned electronic device. For example, electronic devices may also include components that are more... Figure 4 The more or fewer components shown (such as network interfaces, display devices, etc.), or having the same Figure 4 The different configurations shown.

[0153] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing the hardware related to the terminal device. The program can be stored in a computer-readable storage medium, which may include: flash drive, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc.

[0154] The invention will now be described in conjunction with another alternative embodiment.

[0155] Example 4

[0156] This invention also provides a computer-readable storage medium. Optionally, in this invention, the computer-readable storage medium can be used to store the program code executed by the financial risk early warning method based on multi-model fusion provided in Embodiment 1.

[0157] Optionally, in this embodiment of the invention, the storage medium may be located in any computer terminal in a group of computer terminals in a computer network, or in any mobile terminal in a group of mobile terminals.

[0158] This invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing the steps of a financial risk early warning method based on multi-model fusion: acquiring the financial statement data of a target enterprise, and filtering financial indicators from the financial statement data based on statistical methods to form a target indicator set; based on the target indicator set, calculating the deviation of the target enterprise from industry-level indicators using a first calculation model to obtain a first score value; based on the target indicator set, calculating the financial health of the target enterprise using a second calculation model and a preset financial health standard level to obtain a second score value; based on the target indicator set, determining whether the target enterprise hits a preset potential risk group rule set using a third calculation model to obtain a group identification result; and generating a risk early warning signal based on the first score value, the second score value, and the group identification result according to a preset signal mapping rule.

[0159] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0160] In the above embodiments of this application, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0161] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0162] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0163] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0164] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0165] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A financial risk early warning method based on multi-model fusion, characterized in that, The method comprises the following steps: obtaining financial statement data of a target enterprise, and screening financial indicators from the financial statement data based on a statistical method to form a target indicator set; calculating, based on the target indicator set, a deviation degree of the target enterprise relative to an industry level indicator by a first calculation model to obtain a first score value; calculating, based on the target indicator set, a financial health degree of the target enterprise by a second calculation model and a preset financial health standard gear to obtain a second score value; judging, based on the target indicator set, whether the target enterprise hits a preset potential risk group rule set by a third calculation model to obtain a group identification result; generating a risk warning signal based on the first score value, the second score value and the group identification result according to a preset signal mapping rule.

2. The financial risk early warning method according to claim 1, characterized in that, The step of obtaining the financial statement data of the target enterprise comprises: grabbing an annual report text of the target enterprise from a designated enterprise information disclosure platform; parsing the annual report text to obtain a parsing result, wherein the parsing result at least contains financial data items in the balance sheet, the profit table and the cash flow table; standardizing and formatting the financial data items in the parsing result to obtain the financial statement data.

3. The financial risk early warning method according to claim 1, characterized in that, The step of screening financial indicators from the financial statement data based on a statistical method to form a target indicator set comprises: determining all enterprises in a subdivision industry to which the target enterprise belongs as peer enterprises, and obtaining financial statement data of the peer enterprises to obtain peer financial data; determining a peer financial distribution based on the peer financial data, and determining quantile proportions according to the peer financial distribution, wherein the quantile proportions at least contain a first quantile proportion and a second quantile proportion, and the first quantile proportion is smaller than the second quantile proportion; screening financial data items in the financial statement data of the target enterprise that are outside the limits with the first quantile proportion and the second quantile proportion as the limits; based on all the screened financial data items, constructing the target indicator set.

4. The financial risk early warning method according to claim 3, characterized in that, The step of calculating, by a first calculation model, a deviation degree of the target enterprise relative to an industry level indicator to obtain a first score value comprises: for each financial indicator in the target indicator set, calculating a quantile point based on the peer financial distribution and the quantile proportions, wherein the quantile point at least contains a first numerical point and a second numerical point; calculating a median based on the quantile point, and calculating an absolute deviation rate of each financial indicator from the median, wherein the median represents the industry level of the subdivision industry; according to a first preset weight, weighting and summing all absolute deviation rates corresponding to the financial indicators to obtain the first score value, wherein the first preset weight is obtained by training the first calculation model based on historical financial data.

5. The financial risk early warning method according to claim 1, characterized in that, The step of calculating, by a second calculation model and a preset financial health standard gear, a financial health degree of the target enterprise to obtain a second score value comprises: According to the specified enterprise performance evaluation standard value, the evaluation range corresponding to each of the financial indicators is determined, wherein the evaluation range includes: excellent value, good value, average value, lower value and poor value; The actual value of each of the financial indicators of the target enterprise is interpolated with the standard value of the corresponding evaluation range to obtain the range score of each of the financial indicators. According to the second preset weight, the range score of each of the financial indicators is weighted and averaged to obtain the second score value, wherein the second preset weight is obtained by training the second calculation model based on historical financial data.

6. The financial risk early warning method according to claim 1, characterized in that, The steps of determining whether the target enterprise hits the preset potential risk group rule set through the third calculation model to obtain the group identification result, comprising: Obtain a preset potential risk group rule set, wherein the preset potential risk group rule set records N financial risk feature patterns, and N is a preset value; Analyze the target indicator set of the target enterprise to obtain a corresponding financial indicator distribution pattern; Match the financial indicator distribution pattern with each of the financial risk feature patterns in the preset potential risk group rule set and calculate the matching degree; In any case where the matching degree is higher than or equal to a preset matching degree threshold, the group identification result is set to a hit state; In the case where all the matching degrees are lower than the preset matching degree threshold, the group identification result is set to a miss state.

7. The financial risk early warning method according to claim 1, characterized in that, According to a preset signal mapping rule, a risk warning signal is generated based on the first score value, the second score value and the group identification result, comprising: Set the first score value exceeding a first preset threshold as a first judgment condition, set the second score value being lower than a second preset threshold as a second judgment condition, and set the group identification result being in a hit state as a third judgment condition; In the case where the third judgment condition is established and any of the first judgment condition and the second judgment condition is established, a first type of risk warning signal is generated; In the case where only the third judgment condition is established, or only the first judgment condition and the second judgment condition are established at the same time, a second type of risk warning signal is generated; In the case where only the first judgment condition is established, or only the second judgment condition is established, a third type of risk warning signal is generated; In the case where all the judgment conditions are not established, a fourth type of risk warning signal is generated.

8. A financial risk early warning device based on multi-model fusion, characterized in that, Comprise: An acquisition unit is configured to acquire financial statement data of a target enterprise, and to filter financial indicators from the financial statement data based on a statistical method to form a target indicator set; A first calculation unit is configured to calculate a deviation degree of the target enterprise relative to an industry level indicator based on the target indicator set through a first calculation model to obtain a first score value; A second calculation unit is configured to calculate a financial health degree of the target enterprise based on the target indicator set, a second calculation model and a preset financial health standard range to obtain a second score value; A judging unit is configured to judge whether the target enterprise hits a preset potential risk group rule set based on the target index set by using a third calculation model, to obtain a group identification result. A generating unit is configured to generate a risk early warning signal according to a preset signal mapping rule based on the first score value, the second score value and the group identification result.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium comprises a stored computer program, wherein the computer program controls a device where the computer readable storage medium is located to perform the financial risk early warning method based on multi-model fusion in any one of claims 1 to 7 when the computer program is running.

10. An electronic device, comprising: The device comprises one or more processors and a memory, and the memory is configured to store one or more programs, wherein the one or more programs enable the one or more processors to implement the financial risk early warning method based on multi-model fusion in any one of claims 1 to 7 when the one or more programs are executed by the one or more processors.

11. A computer program product, characterised in that, The device comprises computer instructions, wherein the computer instructions enable the steps of the financial risk early warning method based on multi-model fusion in any one of claims 1 to 7 when the computer instructions are executed by a processor.

Images