Intelligent financial risk early warning method and system based on management decision

Abstract

The application relates to the technical field of financial risk early warning, and specifically discloses an intelligent financial risk early warning method and system based on management decision, which performs fund chain vulnerability evaluation and analysis on enterprise financial operation data to obtain fund chain vulnerability evaluation data; performs third-party agency credit evaluation data introduction processing based on the fund chain vulnerability evaluation data to obtain third-party credit evaluation data; performs enterprise fund chain vulnerability dynamic simulation according to the third-party credit evaluation data to obtain fund chain vulnerability evolution track data; performs fund chain fracture risk conduction effect quantification based on the fund chain vulnerability evolution track data to obtain a fund chain fracture risk conduction effect index; and performs financial risk precursor factor identification according to the fund chain fracture risk conduction effect index to obtain financial risk precursor factor learning data. The application not only significantly improves the accuracy and real-time performance of fund chain vulnerability identification, but also provides necessary technical support for constructing an intelligent enterprise financial risk control system.

Description

Technical Field

[0001] This invention belongs to the field of financial risk early warning technology, and relates to an intelligent financial risk early warning method and system based on management decision-making. Background Technology

[0002] In corporate financial management, the robustness of the cash flow is a core indicator for maintaining a company's continuous operation and risk resistance. Cash flow vulnerability assessment aims to identify a company's resilience to fluctuations in cash flow, limited financing channels, or external market shocks. However, in reality, most companies' financial assessment systems have significant limitations. First, traditional models generally rely on static financial statement data, such as the debt-to-equity ratio or current ratio, ignoring the dynamic characteristics of cash inflows and outflows over time, making it difficult to promptly capture potential liquidity crisis signals. Second, companies often over-rely on short-term loans or supply chain credit financing, lacking real-time monitoring and sensitivity analysis of the cash turnover cycle. When the external financing environment tightens, the risk of cash flow disruption is severely underestimated. More critically, existing financial monitoring systems generally fail to establish a real-time feedback mechanism between "cash health status" and "operating activity intensity." When sales growth leads to a surge in accounts receivable without a corresponding improvement in cash flow, the system may still misjudge the company's operating condition as good, masking potential liquidity risks. The lack of dynamic modeling of cash flow resilience and risk transmission paths means that financial decision-makers often only discover the severity of the cash shortage after a crisis erupts, missing the best opportunity for risk prevention and control. Summary of the Invention

[0003] In view of the problems existing in the prior art, the present invention provides an intelligent financial risk early warning method and system based on management decision-making to solve the above-mentioned technical problems.

[0004] To achieve the above and other objectives, the technical solution adopted by the present invention is as follows:

[0005] The first aspect of this invention provides an intelligent financial risk early warning method based on management decision-making, the method comprising:

[0006] Step S1: Collect enterprise financial operation data through a multi-source data acquisition system to obtain enterprise financial operation data; conduct a cash flow vulnerability assessment and analysis on the enterprise financial operation data to obtain cash flow vulnerability assessment data;

[0007] Step S2: Based on the vulnerability assessment data of the capital chain, import and process the credit assessment data of the third-party institution to obtain the third-party credit assessment data; perform dynamic simulation of the vulnerability of the enterprise's capital chain based on the third-party credit assessment data to obtain the evolution trajectory data of the capital chain vulnerability.

[0008] Step S3: Quantify the risk transmission effect of capital chain rupture based on the data of capital chain vulnerability evolution trajectory to obtain the capital chain rupture risk transmission effect index; identify the financial risk antecedent factors based on the capital chain rupture risk transmission effect index to obtain financial risk antecedent factor learning data.

[0009] Step S4: Based on the learning data of the financial risk antecedent factors, perform multi-level early warning threshold analysis and generate early warning signals to obtain the financial risk early warning level, and send the financial risk early warning level to the management decision terminal to execute the enterprise financial risk early warning.

[0010] A second aspect of the present invention provides an intelligent financial risk early warning system based on management decision-making, the system comprising:

[0011] The cash flow vulnerability assessment data acquisition module collects enterprise financial and operational data through a multi-source data acquisition system to obtain enterprise financial and operational data; it then performs a cash flow vulnerability assessment analysis on the enterprise financial and operational data to obtain cash flow vulnerability assessment data.

[0012] Vulnerability Evolution Trajectory Data Acquisition Module: Based on the vulnerability assessment data of the capital chain, third-party credit assessment data is imported and processed to obtain third-party credit assessment data; based on the third-party credit assessment data, dynamic simulation of the enterprise's capital chain vulnerability is performed to obtain the capital chain vulnerability evolution trajectory data;

[0013] Risk Antecedent Factor Acquisition Module: Based on the evolution trajectory data of capital chain vulnerability, the risk transmission effect of capital chain rupture is quantified to obtain the capital chain rupture risk transmission effect index; based on the capital chain rupture risk transmission effect index, financial risk antecedent factors are identified to obtain financial risk antecedent factor learning data;

[0014] Financial risk early warning terminal: Based on the learning data of the aforementioned financial risk pre-factors, it performs multi-level early warning threshold analysis and generates early warning signals to obtain the financial risk early warning level, and sends the financial risk early warning level to the management decision terminal to execute the enterprise financial risk early warning.

[0015] As described above, the intelligent financial risk early warning method and system based on management decision-making provided by the present invention has at least the following beneficial effects:

[0016] This invention, by introducing a multi-source data acquisition and dynamic evaluation mechanism, enables multi-dimensional, full-process monitoring and early warning of corporate cash flow vulnerability. First, comprehensive collection and modeling analysis of corporate financial operational data avoids the limitations of relying solely on static reports, allowing for the dynamic capture of abnormal cash flow signals. Second, the introduction of third-party credit assessment data into the cash flow vulnerability simulation effectively enhances the external validity and objectivity of the assessment model, making cash flow risk prediction more closely aligned with the real financial environment. Third, quantifying the transmission effect of cash flow disruption risk not only reveals the diffusion path of risk across different business segments but also extracts pre-emptive financial risk factors, providing high-value input for training risk characteristics in machine learning models. Finally, a multi-level early warning threshold dynamic analysis and early warning signal generation mechanism transforms potential risks into actionable management decision information, enabling proactive intervention before a crisis erupts. Overall, this system deeply integrates internal and external corporate financial data, credit data, and dynamic modeling methods, significantly improving the accuracy and real-time performance of cash flow vulnerability identification and providing necessary technical support for building an intelligent and forward-looking corporate financial risk control system. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a schematic diagram showing the connections between the steps of the method of the present invention.

[0019] Figure 2 This is a schematic diagram showing the connections of the various modules in the system of the present invention. Detailed Implementation

[0020] The following description, in conjunction with the implementation of this invention, is merely an example and illustration of the concept of this invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the inventive concept or exceed the scope defined in these claims, all of which should fall within the protection scope of this invention. Example

[0021] Please see Figure 1 As shown, an intelligent financial risk early warning method based on management decision-making includes:

[0022] Step S1: Collect enterprise financial operation data through a multi-source data acquisition system to obtain enterprise financial operation data; conduct a cash flow vulnerability assessment and analysis on the enterprise financial operation data to obtain cash flow vulnerability assessment data;

[0023] Step S1 includes the following steps:

[0024] Step S11: Collect enterprise financial operation data through a multi-source data acquisition system to obtain enterprise financial operation data. The multi-source data acquisition system includes a financial statement data acquisition module, a bank transaction data acquisition module, a tax declaration data acquisition module, and an operating data acquisition module.

[0025] Step S12: Perform data cleaning and standardization on the enterprise's financial operation data to obtain standardized financial operation data;

[0026] Step S13: Conduct a cash flow vulnerability assessment and analysis on the standardized financial operating data to obtain preliminary cash flow vulnerability assessment data;

[0027] Step S14: Extract multi-dimensional vulnerability indicators from the preliminary cash flow vulnerability assessment data to obtain the cash flow vulnerability assessment data.

[0028] In this embodiment of the invention, when collecting enterprise financial operation data through a multi-source data acquisition system, an integrated data acquisition platform is first constructed, comprising a financial statement data acquisition module, a bank transaction data acquisition module, a tax declaration data acquisition module, and an operational data acquisition module. The financial statement data acquisition module automatically extracts key accounting item data from the balance sheet, profit and loss statement, and cash flow statement through an API interface with the enterprise's financial software system, and records their time-series changes. The bank transaction data acquisition module obtains daily balance changes, fund inflow and outflow frequency, and amount distribution characteristics of all of the enterprise's bank accounts through a secure data channel established with the banking system. The tax declaration data acquisition module connects to the tax declaration platform to collect the declared amount, actual payment amount, and late payment fee records for various taxes. The operational data acquisition module extracts business data such as purchase orders, sales contracts, inventory turnover, and accounts receivable aging from the enterprise's ERP system. The heterogeneous data collected by each module, after unified timestamp alignment and account entity identification, forms the initial enterprise financial operation dataset.

[0029] When cleaning and standardizing corporate financial operation data, the process begins with rule-based data validation to identify and remove outliers that clearly exceed reasonable limits, including but not limited to negative asset items and turnover rates exceeding theoretical maximum values. Next, for fields with missing values, multiple imputation methods are used to fill in the missing values ​​based on historical data patterns of the same company and data characteristics of similar companies in the same industry. Then, data standardization converts financial indicators of different dimensions into comparable dimensionless values. Specifically, minimum-maximum normalization maps each indicator value to the range of zero to one. Simultaneously, logarithmic transformation is performed on non-normally distributed indicator data to eliminate the impact of data skewness on subsequent analysis. Finally, the cleaned and transformed data undergoes consistency checks to ensure that the values ​​of the same indicator from different sources are consistent, thus forming high-quality standardized financial operation data.

[0030] When conducting a vulnerability assessment of the cash flow to standardized financial operating data, the following steps are taken: First, an assessment system based on cash flow is constructed. This involves calculating the ratio of net operating cash flow to current liabilities to assess the safety margin of short-term solvency; and analyzing the matching degree between cash outflows from investing activities and cash inflows from financing activities to determine the stability of funding support for investment expansion. Second, from a debt structure perspective, the rationality of the company's capital structure is assessed by calculating the ratio of interest-bearing debt to equity capital; and the potential risks of the debt maturity structure are identified by measuring the proportion of short-term debt in total debt. Third, from a profit protection perspective, the company's ability to pay debt interest is assessed by calculating the coverage ratio of EBITDA to interest expense; and the quality of profitability is determined by analyzing the difference between operating profit and net cash flow from operating activities. Finally, a weighted comprehensive evaluation method is used, assigning different weights to the indicators of each dimension according to their impact on cash flow vulnerability. Preliminary cash flow vulnerability assessment data is obtained through linear weighted calculation. The determination of weight coefficients uses the analytic hierarchy process (AHP), with expert scoring used to construct a judgment matrix and calculate the relative importance of each indicator.

[0031] When extracting multi-dimensional vulnerability indicators from preliminary cash flow vulnerability assessment data, the process begins with extracting dynamic characteristics of cash flow vulnerability from a time perspective. This includes calculating the moving average of vulnerability assessment values ​​to observe long-term trends, calculating its standard deviation to measure volatility, and separating seasonal variations using time series decomposition. Secondly, from a structural perspective, the contribution of each assessment indicator to overall vulnerability is extracted. By calculating the correlation coefficient between each indicator value and the overall assessment value, key drivers affecting cash flow security are identified. Factor analysis is used to extract the main common factors influencing cash flow vulnerability, and the variance explained by each common factor is calculated. Thirdly, from a comparative perspective, the company's relative position within the industry is extracted. By calculating the deviation of the company's cash flow vulnerability assessment value from the industry average, the company's relative risk status is assessed. Quantile calculations determine the company's percentile rank in the industry risk distribution. Finally, the feature indicators extracted from each dimension are integrated to form a comprehensive cash flow vulnerability assessment dataset reflecting the characteristics of corporate cash flow vulnerability.

[0032] Step S2: Based on the vulnerability assessment data of the capital chain, import and process the credit assessment data of the third-party institution to obtain the third-party credit assessment data; perform dynamic simulation of the vulnerability of the enterprise's capital chain based on the third-party credit assessment data to obtain the evolution trajectory data of the capital chain vulnerability.

[0033] Step S2 includes the following steps:

[0034] Step S21: Obtain credit assessment data from third-party institutions, wherein the credit assessment data from third-party institutions includes bank credit rating data, supplier credit assessment data, and customer credit rating data;

[0035] Step S22: Based on the vulnerability assessment data of the capital chain, conduct a credibility analysis of the third-party credit assessment data to obtain the credibility weight of the third-party credit assessment data;

[0036] Step S23: Perform weighted fusion processing on the third-party credit assessment data according to the credibility weight of the third-party credit assessment data to obtain weighted third-party credit assessment data;

[0037] Step S24: Based on the vulnerability assessment data of the capital chain and the weighted third-party credit assessment data, conduct a dynamic simulation of the vulnerability of the enterprise's capital chain to obtain the trajectory data of the vulnerability of the capital chain.

[0038] In this embodiment of the invention, during the stage of acquiring third-party credit assessment data, data interfaces are established with bank credit management systems, supplier assessment platforms, and customer credit databases to systematically collect external credit evaluation information of enterprises. Bank credit rating data includes the enterprise's credit rating, credit line utilization rate, loan delinquency records, and guarantee circle risk contagion index. Cross-bank correlation analysis of the enterprise's credit records across various financial institutions forms a unified bank credit view. Supplier credit assessment data is obtained by extracting indicators such as payment due date compliance, order fulfillment rate, and frequency of quality disputes from the supplier management system, and using a sliding time window statistical method to calculate the comprehensive supplier evaluation score for the past twelve months. Customer credit rating data comes from the enterprise's CRM system, and is obtained by analyzing parameters such as customer payment timeliness, order stability, and bad debt history, combined with... The business climate indicators of the client's industry are weighted and adjusted to form a credit risk distribution map of the client group. After completing the collection of multi-source third-party data, when conducting credibility analysis of third-party credit assessment data based on the vulnerability assessment data of the capital chain, the Pearson correlation analysis method is first used to calculate the correlation strength between each third-party indicator and the core indicators of capital chain vulnerability, and effective credit indicators with a significance higher than the set threshold are screened out. Then, by calculating the information overlap of different data sources under the same credit dimension, the influence of duplicate counting is identified and eliminated. Next, the coefficient of variation method is used to evaluate the dispersion of each credit indicator, and indicators with excessive volatility are smoothed. Finally, the analytic hierarchy process is used to construct a credit data credibility evaluation model, calculate the relative weight of each credit data source by constructing a judgment matrix, and obtain the credibility weight of third-party credit assessment data after normalization of the weights.

[0039] When performing weighted fusion processing based on the credibility weights of third-party credit assessment data, the process first involves dimensionless processing of different types of credit data, unifying percentage scores, ratings, and quantitative indicators to a standard scoring range of zero to one. Then, a weighted average is calculated for each type of credit score based on its credibility weight. Bank credit data is given higher weight due to its official certification characteristics, while supplier and customer credit data have their weights dynamically adjusted based on their relevance to the company's business. For conflicting credit assessments, a conflict resolution algorithm based on DS evidence theory is used for consistency fusion. By calculating the basic probability allocation functions of different information sources, a final joint credit assessment value is synthesized. After weighted fusion, structured weighted third-party credit assessment data is output. Finally, when performing dynamic simulation of corporate cash flow vulnerability based on cash flow vulnerability assessment data and weighted third-party credit assessment data, a system dynamics-based approach is first constructed. The cash flow simulation model divides a company's cash system into three interacting subsystems: operating cash flow, investing cash flow, and financing cash flow. It then introduces weighted third-party credit assessment data as an external perturbation variable, simulating risk scenarios such as bank credit tightening, shortened supplier payment terms, and delayed customer payments by setting different credit condition scenarios. During the simulation, the Monte Carlo method is used to randomly generate time series of credit events, and the rate of depletion of the company's cash reserves under various credit environments is calculated through multiple iterations. Simultaneously, a critical condition judgment rule for cash flow disruption is established; when the simulated cash balance continuously falls below a safe threshold and financing cannot be obtained through credit channels, it is marked as a cash flow disruption. After tens of thousands of scenario simulations, the probability distribution of cash flow disruption is statistically analyzed, and the cash flow pressure values ​​at different time points are extracted, ultimately forming a cash flow vulnerability evolution trajectory data reflecting the changing trend of the company's cash flow vulnerability over the next twelve months.

[0040] Step S23 includes the following steps:

[0041] Step S231: Based on the vulnerability assessment data of the capital chain, conduct a correlation analysis of the third-party credit assessment data to obtain the correlation coefficient of the credit assessment data;

[0042] Step S232: Based on the correlation coefficient of the credit assessment data, perform correlation screening on the credit assessment data of third-party institutions to obtain high-correlation credit assessment data;

[0043] Step S233: Perform a time series consistency test on the highly correlated credit assessment data to obtain the time series consistency index of the credit assessment data;

[0044] Step S234: Perform dynamic weighted fusion of credit assessment data based on the time series consistency index of credit assessment data to obtain weighted third-party credit assessment data.

[0045] In this embodiment of the invention, when conducting correlation analysis of third-party credit assessment data based on the vulnerability assessment data of the capital chain, a multivariate correlation analysis method is first used to calculate the Pearson correlation coefficient between each sub-indicator in the bank credit rating data and the core assessment indicator of the capital chain vulnerability. The explanatory power of the bank credit data on the enterprise's financial situation is determined by analyzing the degree of linear correlation between the two. At the same time, a grey relational analysis method is used to calculate the similarity in development trend between the supplier credit assessment data sequence and the capital chain vulnerability assessment data sequence. The correlation strength is quantified by comparing the geometric similarity between the data sequences.

[0046] For customer credit rating data, canonical correlation analysis is used to study the overall correlation between multiple sets of customer credit variables and multiple sets of cash flow vulnerability variables. The correlation coefficient is determined by finding the linear combination that best represents the correlation between the two sets of variables. In the calculation process, the time lag effect is also considered. Cross-correlation analysis is used to determine the maximum correlation between credit data and cash flow vulnerability data under different time delays. Finally, a correlation coefficient matrix of credit assessment data that comprehensively reflects the intrinsic connection between various third-party credit data and cash flow vulnerability is obtained.

[0047] When performing correlation screening based on the correlation coefficient of credit assessment data, a correlation threshold is first set, and credit indicators with correlation coefficients below the threshold are considered weakly correlated and removed. Then, principal component analysis is used to reduce the dimensionality of the remaining credit indicators. By calculating the eigenvalues ​​and contribution rates of each credit indicator, the main components with cumulative contribution rates that meet the requirements are selected as core credit features. For credit indicators with high correlation coefficients but multicollinearity, the degree of multicollinearity is calculated using the variance inflation factor, and indicators with inflation factors exceeding the critical value are selectively removed. At the same time, considering the differences in credit data under different industry backgrounds, cluster analysis is used to classify credit data according to industry characteristics, and stratified screening is performed within each category based on the correlation coefficient. Finally, a highly correlated credit assessment dataset that is both representative and avoids information redundancy is obtained.

[0048] When conducting time series consistency tests on highly correlated credit assessment data, the following steps are taken: First, a time series model of the credit data is constructed. The moving average and standard deviation of each credit indicator are calculated using the sliding window method, and the stability of the data is assessed by analyzing its fluctuation patterns. Then, the unit root test method is used to determine the stationarity of each credit data series, and non-stationary series are differencing until a stationary state is reached. Next, the autocorrelation function and partial autocorrelation function are used to analyze the autocorrelation structure of the credit data series, and the periodic characteristics of the data are identified by calculating the autocorrelation coefficients of different lag orders. At the same time, the sliding window correlation coefficient analysis method is used to calculate the changes in the correlation coefficients between each credit indicator and the financial chain vulnerability indicator in different time intervals, and the timeliness consistency of the credit data is assessed by analyzing the stability of the correlation coefficients. Finally, by calculating the coefficient of variation and trend consistency index of each credit data series, the reliability and stability of the credit data in the time dimension are comprehensively evaluated, forming a set of time series consistency indicators for credit assessment data.

[0049] When dynamically weighting and fusing credit assessment data based on the time series consistency index, a weight allocation function is first constructed based on the time series consistency index, assigning higher fusion weights to credit data with high stability and strong consistency. Then, an adaptive weighted fusion algorithm is used to dynamically adjust the weight coefficients of each credit data point based on its consistency performance at different time points, ensuring the fusion result reflects the latest trends in credit data. For conflicting credit assessment information, the synthesis rules in DS evidence theory are used for conflict resolution. By calculating the basic probability allocation function and conflict coefficient of each credit data source, the optimal fusion weight allocation scheme is determined. Simultaneously, a sliding time window mechanism is introduced to assign higher weights to recent credit data, and the exponentially weighted moving average method is used to calculate the dynamic weights of each credit indicator. Finally, a weighted average algorithm is used to fuse each credit data point according to its dynamic weights, considering the mutual corroboration relationships between the data points during the fusion process. Credit information supported by multiple credit data sources is appropriately weighted, ultimately generating weighted third-party credit assessment data.

[0050] Step S24 includes the following steps:

[0051] Step S241: Construct a capital chain stress test scenario based on the capital chain vulnerability assessment data and weighted third-party credit assessment data to obtain capital chain stress test scenario data;

[0052] Step S242: Perform multi-scenario cash flow simulation on the cash flow stress test scenario data to obtain multi-scenario cash flow simulation data;

[0053] Step S243: Calculate the probability of capital chain rupture based on multi-scenario cash flow simulation data to obtain the probability distribution data of capital chain rupture.

[0054] Step S244: Fit the evolution trajectory of the vulnerability of the capital chain based on the probability distribution data of the capital chain breakage to obtain the evolution trajectory data of the vulnerability of the capital chain;

[0055] Step S245: Identify vulnerability inflection points in the trajectory data of the vulnerability evolution of the capital chain to obtain vulnerability inflection point data of the capital chain;

[0056] Step S246: Set the early warning threshold for the collapse of the capital chain based on the inflection point data of the capital chain vulnerability, and obtain the early warning threshold data for the collapse of the capital chain.

[0057] In this embodiment of the invention, when constructing a stress test scenario for the capital chain based on capital chain vulnerability assessment data and weighted third-party credit assessment data, the first step is to extract key stress factors affecting the stability of the capital chain from a database of corporate capital chain rupture cases during historical financial crises. These factors include, but are not limited to, the degree of bank credit tightening, the extent of supplier payment term shortening, and the number of days customer payment cycles extended. Then, extreme value theory analysis is used to determine the extreme fluctuation range of each stress factor, and the severity level of the stress test is set by calculating the value at risk of each factor at a confidence level. Finally, the Copula function is used to construct the dependency structure between multiple stress factors, accurately characterizing... The joint probability distribution of simultaneous occurrences of bank credit contraction, supplier credit deterioration, and customer credit decline is studied. During scenario generation, Monte Carlo simulation is used to randomly generate stress scenario combinations that conform to actual distribution characteristics, with each scenario containing a specific combination of credit environment parameters. Simultaneously, the impact of macroeconomic cycle fluctuations is considered, and leading macroeconomic indicators are introduced as external driving variables for scenario construction to establish a transmission path from the macro environment to the enterprise's micro-level cash flow. Ultimately, a multi-level cash flow stress test scenario dataset is formed, including a baseline scenario, a mild stress scenario, a moderate stress scenario, and a severe stress scenario, providing a complete testing environment framework for subsequent cash flow simulations.

[0058] When conducting multi-scenario cash flow simulations based on data from a cash flow stress test, a system dynamics model based on three-stage cash flow forecasting is first constructed to quantitatively model the interrelationships between cash flows from operating, investing, and financing activities. In the operating cash flow simulation, an autoregressive integral moving average model is used to predict changes in sales revenue. The impact of changes in customer credit ratings on accounts receivable turnover is calculated to adjust the cash collection speed, and the payment schedule for accounts payable is dynamically set based on supplier credit assessment data. In the investment cash flow simulation, optimal planning methods are used to determine investment expenditures based on the matching degree between the company's investment plan and financing capabilities. Priority sequences and scale adjustment strategies are established; in the financing cash flow simulation, a correlation model between bank credit lines and credit ratings is established, and the probability distribution of bank credit line reduction under different stress scenarios is calculated using logistic regression; during the simulation, the time step method is used to calculate the daily changes in cash balance, and the cash gap accumulation model is used to record the funding shortage at each time point; for each stress test scenario, a sufficient number of random simulation experiments are conducted to obtain statistically stable cash flow simulation results, and finally, a multi-scenario cash flow simulation dataset containing key indicators such as the time series changes in cash balance, the time point of the funding gap, and the maximum funding gap size is output.

[0059] When calculating the probability of a cash flow disruption based on multi-scenario cash flow simulation data, the criteria for determining a cash flow disruption are first clearly defined. A situation where the cash balance is below a safety threshold for thirty consecutive days and cannot be replenished through emergency financing is marked as a cash flow disruption event. Then, a survival analysis method is used to construct a cash flow survival function, and the Kaplan-Meier estimator is used to calculate the cash flow survival probability at each time point, thereby deriving the risk rate function for a cash flow disruption. Next, the Cox proportional hazards model is used to analyze the contribution of each stress factor to the risk of a cash flow disruption, and the key risk drivers are determined by calculating the risk ratio. During the probability calculation process, the kernel density estimation method is used to perform nonparametric fitting on the distribution of cash flow disruption times to obtain a continuous distribution curve of the probability of a cash flow disruption. Simultaneously, the probability weight allocation under different stress scenarios is considered, and the reasonable weights for each scenario are determined by combining expert scoring with historical data backtesting. Finally, a weighted average calculation is used to obtain the cash flow disruption probability distribution data that comprehensively considers various changes in the credit environment.

[0060] When fitting the trajectory of capital chain vulnerability evolution based on the probability distribution data of capital chain rupture, the time series decomposition technique is first used to split the capital chain rupture probability sequence into trend components, periodic components, and random components. The long-term evolution direction of vulnerability is identified by analyzing the slope changes of the trend component. Then, a state-space model is used to establish the dynamic evolution equation of capital chain vulnerability, and weighted third-party credit assessment data is introduced into the model as state variables. The model parameters are estimated and updated in real time using the Kalman filter algorithm. During the trajectory fitting process, spline interpolation is used to smooth the discrete probability observations to generate a continuous capital chain vulnerability evolution curve. At the same time, the path dependence characteristics of vulnerability evolution are considered, and a memory factor is introduced to characterize the influence of historical vulnerability levels on the current state. Finally, the accuracy of trajectory fitting is evaluated based on the goodness-of-fit test method, and the root mean square error and coefficient of determination are used to quantify the fitting quality. The final output is the capital chain vulnerability evolution trajectory data, which clearly shows the evolution path and trend of enterprise capital chain vulnerability under different credit environments.

[0061] When identifying vulnerability inflection points in the evolution trajectory data of financial chain vulnerability, a multi-scale edge detection algorithm is first used to analyze the feature points of the evolution trajectory curve. The possible inflection point locations are located by calculating the rate of curvature change of the curve at each point. Then, the structural breakpoint test method is used to identify structural change points in the trajectory by calculating the sequence F statistic, and to determine the time point when the vulnerability evolution mechanism has changed significantly. In the process of inflection point confirmation, the sliding window method is used to calculate the local trend slope of the trajectory in different time intervals. The trend reversal point is identified by comparing the degree of difference in the slope between adjacent intervals. At the same time, considering the statistical significance of inflection point identification, the confidence interval of each candidate inflection point is calculated by repeated sampling using the bootstrap method, and only reliable inflection points that pass the significance test are retained. For the identified inflection points, their nature is further determined by inflection point type discriminant analysis, distinguishing different types such as vulnerability acceleration inflection points, vulnerability mitigation inflection points, and vulnerability trend reversal inflection points. Finally, a financial chain vulnerability inflection point dataset containing complete information such as inflection point location, inflection point type, and inflection point strength is output.

[0062] When setting early warning thresholds for cash flow collapse based on data on inflection points in cash flow vulnerability, the process begins with a retrospective analysis of historical cash flow collapse cases to determine the time lead relationship and strength correlation between various types of inflection points and the final cash flow collapse. Then, ROC curve analysis is used to evaluate the discriminative effectiveness of different vulnerability levels as early warning thresholds, and the optimal threshold is determined by calculating the sensitivity and specificity of each candidate threshold. During threshold setting, the characteristics of different types of enterprises are considered, and cluster analysis is used to classify enterprises according to industry characteristics, size, and growth stage, setting differentiated early warning thresholds for each category. Simultaneously, a multi-level early warning mechanism is established, dividing the cash flow vulnerability level into three warning levels—attention level, warning level, and danger level—based on the proximity of the threshold to the threshold, and setting corresponding emergency response measures for each level. Finally, retrospective testing verifies the effectiveness of the early warning thresholds, calculating performance indicators such as accuracy, false alarm rate, and false negative rate of the early warning system in historical samples. Based on the test results, the thresholds are optimized and adjusted, ultimately outputting scientifically sound and validated early warning threshold data for cash flow collapse.

[0063] Step S244 includes the following steps:

[0064] Time series trend analysis was performed on the probability distribution data of capital chain rupture to obtain the time series trend curve of capital chain rupture probability.

[0065] Abrupt change point detection was performed on the time series trend curve of the probability of capital chain breakage to obtain abrupt change point data of the probability of capital chain breakage.

[0066] Based on the data on the probability of sudden changes in the funding chain breakage, the evolution stages of funding chain vulnerability are divided to obtain data on the evolution stages of funding chain vulnerability.

[0067] Based on the data on the evolution of vulnerability in the capital chain, an analysis of the vulnerability transmission relationship between stages was conducted to obtain vulnerability stage transmission relationship data;

[0068] Based on the Markov chain model, the vulnerability transmission relationship data of the financial chain is fitted to obtain the vulnerability evolution trajectory data of the financial chain.

[0069] In this embodiment of the invention, when performing time series trend analysis on the probability distribution data of capital chain rupture, the original probability sequence is first divided into trend components, periodic components, and random fluctuation components using a seasonal decomposition method. The core factors dominating the change in the probability of capital chain rupture are determined by calculating the variance contribution rate of each component. Next, the trend component is smoothed using an exponentially weighted moving average algorithm. The smoothing coefficient is adjusted to balance the sensitivity and stability of trend identification, thereby eliminating the interference of short-term fluctuations on long-term trend judgment. Then, a polynomial fitting method is used to construct the trend curve equation. The optimal fitting order is determined by minimizing the sum of squared residuals between the fitted curve and the original data points, thus obtaining a time series trend curve of the probability of capital chain rupture that reflects the long-term direction of change in the probability of capital chain rupture. When analyzing the capital chain... When detecting abrupt change points in the trend curve of the probability of capital chain breakage, the sliding window method is first used to calculate the rate of change of the local slope of the trend curve at each point. Candidate points where the trend direction has changed significantly are identified by comparing the differences in slope within adjacent windows. Then, the CUSUM cumulative sum algorithm is used to calculate the cumulative deviation of each point in the sequence from the historical mean. When the cumulative deviation exceeds a preset threshold, it is marked as a statistically significant abrupt change point. At the same time, the Bootstrap resampling technique is used to evaluate the significance level of each candidate abrupt change point. Confidence intervals for abrupt change test are constructed through multiple random samplings to ensure the statistical reliability of the detection results. Finally, cluster analysis is used to merge and optimize adjacent candidate abrupt change points, eliminating false abrupt change signals caused by data noise, and outputting statistically significant data on abrupt change points of the probability of capital chain breakage.

[0070] When dividing the evolutionary stages of capital chain vulnerability based on the data of abrupt change points in the probability of capital chain rupture, the entire time series is first divided into several consecutive time intervals according to the location of the abrupt change points, with each interval representing a relatively stable vulnerability evolutionary stage. Then, the average, variance, and autocorrelation coefficient of the probability of capital chain rupture within each stage are calculated. Adjacent stages with similar statistical characteristics are merged using a clustering algorithm to form evolutionary stages with clear distinguishability. Next, discriminant analysis is used to verify the rationality of the stage division, and the significance level of stage differences is quantified by calculating the Mahalanobis distance of statistical characteristics between stages. Finally, specific risk characteristic labels are defined for each evolutionary stage, such as stable period, volatile period, and deterioration period, forming a clearly structured data set of capital chain vulnerability evolutionary stages. Based on the capital chain rupture probability data, the data is then further divided into several consecutive time intervals, each representing a relatively stable vulnerability evolutionary stage. When analyzing the vulnerability transmission relationships between different stages of the gold chain vulnerability evolution data, the following steps are taken: First, a stage transition frequency matrix is ​​constructed, and the transition probability is calculated by statistically analyzing the number of transitions between each stage in historical data. Then, the Granger causality test is used to analyze the predictive ability of the characteristics of the previous stage on the development of the next stage, identifying the causal relationships between stages. Next, a vector autoregression model is used to quantify the dynamic influence relationship between the core indicators of each stage, and the impact intensity and duration of the impact of a shock in one stage on subsequent stages are analyzed through impulse response functions. At the same time, the moderating effect of changes in the external credit environment on stage transmission is considered, and the influence mechanism of third-party credit assessment data on the stage transmission path is analyzed by introducing interaction terms. Finally, vulnerability stage transmission relationship data that fully describes the internal connections and transmission laws between each evolution stage is obtained.

[0071] When fitting the evolution trajectory of financial chain vulnerability based on the vulnerability stage transmission relationship data using a Markov chain model, a state-space model is first constructed, defining each evolution stage of financial chain vulnerability as a different state of the Markov chain. Then, the state transition probability matrix is ​​calculated based on the stage transition frequency matrix, and multi-step transition probabilities are derived through matrix operations to predict the distribution of vulnerability states in multiple future periods. Next, the Viterbi algorithm is used to find the most likely state transition path, and the most reasonable state sequence under given observation data is determined through dynamic programming. At the same time, the time-varying characteristics of the transition probability are considered, and a time-dependent factor is introduced to enable the Markov model to adapt to the changing transition patterns over time. Finally, a large number of possible evolution paths are generated through Monte Carlo simulation, and the probability prediction interval of financial chain vulnerability evolution is constructed by statistically analyzing the distribution characteristics of these paths, outputting the financial chain vulnerability evolution trajectory data.

[0072] Step S3: Quantify the risk transmission effect of capital chain rupture based on the data of capital chain vulnerability evolution trajectory to obtain the capital chain rupture risk transmission effect index; identify the financial risk antecedent factors based on the capital chain rupture risk transmission effect index to obtain financial risk antecedent factor learning data.

[0073] Step S3 includes the following steps:

[0074] Step S31: Construct a risk transmission network from the data on the evolution trajectory of the vulnerability of the capital chain to obtain the risk transmission network data of the capital chain;

[0075] Step S32: Quantify the risk transmission effect of capital chain rupture based on the data of the capital chain risk transmission network to obtain the capital chain rupture risk transmission effect index;

[0076] Step S33: Identify the antecedent factors of financial risk based on the risk transmission effect index of the broken capital chain and the evolution trajectory data of the capital chain vulnerability, and obtain the set of antecedent factors of financial risk;

[0077] Step S34: Ranking and learning the set of financial risk antecedent factors by importance to obtain financial risk antecedent factor learning data.

[0078] In this embodiment of the invention, when constructing a risk transmission network based on the evolution trajectory data of cash flow vulnerability, a topological framework for cash flow risk transmission is first built based on the reconciliation relationships between various accounting items in the company's financial statements and the cash flow paths in the business process. Cash flow-related elements are abstracted as network nodes, including asset nodes such as accounts receivable, inventory, and fixed assets; liability nodes such as short-term loans and accounts payable; and profit and loss nodes such as sales revenue and operating costs. Then, the Granger causality test method is used to analyze the time lead-lag relationship of changes in cash indicators at each node, calculating the predictive power of historical data from one node on the current value of another node. The significance level is used to determine whether a transmission relationship exists between nodes. Then, the impulse response function of the vector autoregression model is used to quantify the dynamic influence intensity between nodes. The transmission weight is calculated by analyzing the cumulative response of other nodes when one node is subjected to a unit impact. At the same time, the changes in the transmission path under different business cycles are considered. The stability of the transmission relationship is identified by analyzing the rolling time window, and only the transmission path that is consistently significant within multiple time windows is retained. Finally, the centrality index of each node is calculated based on complex network theory, including degree centrality, betweenness centrality, and proximity centrality, to identify key risk hub nodes in the network and form complete data on the transmission network of capital chain risks.

[0079] When quantifying the risk transmission effect of capital chain rupture based on capital chain risk transmission network data, the SIR infectious disease model is first used to simulate the spread of risk in the network. The initial capital chain rupture risk is set as the source of infection, and the risk transmission range is simulated by calculating the infection probability of each node. Then, the Gini coefficient method is used to assess the unevenness of risk transmission, and the risk aggregation effect is quantified by calculating the degree of inequality in the distribution of risk among nodes. Next, a cascading failure model of risk transmission is constructed to simulate the scale of the chain reaction caused by the correlation between nodes when a key node experiences a capital chain rupture. The systemic risk level is assessed by calculating the number of affected nodes and the proportion of capital scale. At the same time, the time dimension of risk transmission is considered. The time length and speed indicators required for the risk to travel from occurrence to transmission to each node are calculated to assess the timeliness of risk transmission. Finally, multiple dimensions such as transmission range, transmission intensity, and transmission speed are integrated, and the entropy weight method is used to determine the weight of each dimension. A comprehensive index that fully reflects the risk transmission effect of capital chain rupture is obtained by weighted summation. This index not only considers the breadth of risk transmission but also includes the depth and speed characteristics of risk transmission. When identifying pre-risk factors for financial risks based on the risk transmission effect index of broken capital chains and the evolution trajectory data of capital chain vulnerability, the following steps are taken: First, principal component analysis is used to reduce the dimensionality of multidimensional financial indicators. Representative core indicators are selected by calculating the loading coefficients of each indicator on the principal components. Then, the feature importance assessment method of the decision tree algorithm is used to identify the key factors with the greatest impact on the risk transmission effect by calculating the information gain brought by each feature in the process of constructing the decision tree. Next, through time series lead-lag analysis, the average lead time of changes in each financial indicator relative to the deterioration of capital chain vulnerability is calculated to select pre-risk indicators with early warning value. At the same time, considering the interaction between indicators, feature interaction effect analysis using the random forest algorithm is used to identify indicator combinations with synergistic enhancement effects. Finally, a candidate set of pre-risk factors for financial risks is constructed based on the selected core indicators. The business rationality and practical effectiveness of the pre-risk factors are ensured through domain expert review and historical case verification, forming a scientific and complete set of pre-risk factors for financial risks.

[0080] When ranking and learning the set of financial risk antecedent factors, the following approach is first used: First, the random forest algorithm is employed to measure variable importance, assessing the relative importance of each antecedent factor by calculating its contribution to the decline in model accuracy. Then, the gradient boosting decision tree algorithm is used to evaluate feature importance, supplementing and validating the importance ranking by counting the number of times each feature acts as a split point across all decision trees. Next, the SHAP value analysis method is used to quantify the marginal contribution of each antecedent factor to the risk prediction results, obtaining a more robust importance assessment by calculating the Shapley value of each feature on different samples. Simultaneously, considering the stability and operability of the antecedent factors, the volatility of the importance ranking of each factor over different time periods is calculated, prioritizing factors with high and stable importance. Finally, a learning model of the mapping relationship between antecedent factors and cash flow risk is constructed based on historical data. A neural network algorithm is used to capture complex nonlinear relationships, and the model parameters are continuously updated by adding new sample data, enabling the system to adaptively learn changes in risk patterns, ultimately outputting financial risk antecedent factor learning data.

[0081] Step S32 includes the following steps:

[0082] Step S321: Perform node importance analysis on the risk transmission network data of the capital chain to obtain node importance data of the risk transmission network;

[0083] Step S322: Evaluate the criticality of the risk transmission path based on the importance data of the nodes in the risk transmission network to obtain the criticality data of the risk transmission path;

[0084] Step S323: Calculate the risk transmission speed based on the criticality data of the risk transmission path to generate a risk transmission speed index; calculate the risk transmission range of the risk transmission speed index to obtain a risk transmission range index.

[0085] Step S324: Calculate the transmission effect intensity of the risk transmission path criticality data based on the risk transmission range index to obtain the risk transmission effect intensity data;

[0086] Step S325: Perform multi-level transmission effect accumulation processing based on the risk transmission effect intensity data to obtain risk multi-level transmission accumulation effect data;

[0087] Step S326: Quantify the risk transmission effect index of capital chain rupture based on the cumulative effect data of risk multi-level transmission, and obtain the risk transmission effect index of capital chain rupture.

[0088] In this embodiment of the invention, when analyzing the node importance of the risk transmission network data of the capital chain, the improved PageRank algorithm is first used to calculate the relative importance of each node in the network. The global influence of each node is determined by iteratively calculating the number and quality of points to each node by other important nodes. At the same time, combined with the local characteristics of the nodes, the degree centrality index is used to count the number of edges directly connected to each node, the betweenness centrality index is used to calculate the frequency of the node appearing on the shortest path of other node pairs, and the proximity centrality index is used to measure the average shortest distance from the node to all other nodes in the network. Then, the functional characteristics of the nodes in the capital flow process are considered. By analyzing the criticality of the accounting subjects represented by the nodes in the capital cycle, different functional weights are assigned. Finally, the entropy weight method is used to weight and fuse multiple centrality indicators. The objective weight is determined by calculating the information entropy value of each indicator to avoid the bias of subjective assignment. Finally, the node importance data of the risk transmission network that comprehensively considers the importance of the network topology and business functions is obtained.

[0089] When assessing the criticality of risk transmission paths based on the importance data of nodes in a risk transmission network, the K-shortest path algorithm is first used to identify all important risk transmission paths in the network. Different path length thresholds are set to filter transmission paths with practical significance. Then, a path criticality evaluation index system is constructed, including dimensions such as the number of important nodes traversed by the path, the average node importance score on the path, and path redundancy. Next, the TOPSIS comprehensive evaluation method is used to calculate the closeness of each path to the ideal solution. By constructing positive and negative ideal solutions, the relative distance from each path to these two reference points is calculated to assess its criticality. Simultaneously, the stability factor of the paths is considered; by analyzing the persistence of paths in different time windows, higher criticality weights are assigned to long-term stable paths. Finally, the paths are sorted and classified according to their criticality scores, dividing them into three levels: core paths, important paths, and general paths, forming complete risk transmission path criticality data. When calculating the risk transmission speed based on the criticality data of the risk transmission path, the Dijkstra algorithm is first used to find the shortest transmission time path on each critical path. The basic transmission time is determined by analyzing the capital flow cycle between nodes on the path. Then, a node processing delay factor is introduced to consider the difference in response time of different types of nodes to risk signals. For example, the risk transmission speed of accounts receivable nodes is usually faster than that of fixed asset nodes. Next, a dynamic calculation model for risk transmission speed is constructed. By analyzing the statistical characteristics of the actual time of risk transmission along each path in historical data, the mean and variance of the path transmission speed are calculated. After generating the risk transmission speed index, the SIR model is further used to simulate the diffusion process of risk in the network. By setting different initial infected nodes and transmission parameters, the proportion of nodes infected within a fixed time range is statistically analyzed to obtain the risk transmission range index. At the same time, the clustering characteristics of the network structure are considered. The local clustering effect of risk transmission is analyzed by calculating the modularity index of the network, thus improving the evaluation dimension of the risk transmission range.

[0090] When calculating the transmission effect intensity of risk transmission path criticality data based on the risk transmission range index, a framework for assessing the intensity of the transmission effect is first constructed, decomposing the transmission effect into two dimensions: breadth effect and depth effect. The breadth effect is directly measured by the risk transmission range index, while the depth effect is assessed by analyzing the degree of change in node status after risk transmission. Then, an influence measurement method is used to calculate the average impact of each critical path on affected nodes during risk transmission, and the depth of transmission is quantified by analyzing the weighted average of the changes in node status before and after risk transmission. Next, the concept of path load capacity is introduced, considering the carrying capacity of each path in actual capital flow, and paths with larger carrying capacity are assigned higher effect intensity weights. Finally, a nonlinear mapping relationship from path criticality and transmission range to transmission effect intensity is established through a radial basis function neural network, and the network parameters are optimized by training historical data to obtain accurate risk transmission effect intensity data. When processing the cumulative effects of multi-level transmission based on the intensity data of risk transmission effects, we first construct a cascade effect model of risk transmission to simulate the process of risk propagating step by step along the network path from the initial node. Then, we use convolution operations to calculate the cumulative effect of multi-level transmission, and calculate the residual strength of risk after multi-level transmission by setting attenuation coefficients for different transmission levels. Next, we consider the synergistic and offsetting effects in the transmission process, analyze the risk superposition law at the intersection of multiple transmission paths, and calculate the resultant force effect of multi-path risk transmission using vector synthesis. At the same time, we introduce a time accumulation factor to analyze the time delay effect of risk in the transmission process, and calculate the cumulative impact of risk in a specific time period through integral operations. Finally, we use system dynamics to establish a stock-flow model of risk transmission, and simulate the dynamic accumulation process of risk in multi-level transmission by setting different transmission rates and feedback mechanisms, thus obtaining comprehensive data on the cumulative effect of multi-level risk transmission that reflects the complex characteristics of risk transmission.

[0091] When quantifying the risk transmission effect index of capital chain rupture based on the cumulative effect data of multi-level risk transmission, a multi-dimensional evaluation index system is first constructed, including core indicators such as transmission range coverage, transmission intensity peak, transmission speed extreme value, and cumulative effect scale. Then, principal component analysis is used to reduce the dimensionality of multiple evaluation indicators, extracting the main components that can represent most of the information, and determining their weight in the comprehensive index by calculating the variance contribution rate of each principal component. Next, fuzzy comprehensive evaluation method is used to handle the uncertainty in the evaluation process. By establishing the membership function and fuzzy relation matrix of each indicator, the comprehensive fuzzy score of the risk transmission effect is calculated. At the same time, considering the differences in the characteristics of different enterprises, cluster analysis is used to classify enterprises into different types, and a customized index calculation model is established for each type. Finally, the final index is synthesized by the weighted geometric mean method, and the discriminant validity of the index is verified by multi-dimensional scaling analysis. The predictive accuracy of the index is verified by backtesting with historical data, and the capital chain rupture risk transmission effect index is finally obtained.

[0092] Step S4: Based on the learning data of the financial risk antecedent factors, perform multi-level early warning threshold analysis and generate early warning signals to obtain the financial risk early warning level, and send the financial risk early warning level to the management decision terminal to execute the enterprise financial risk early warning;

[0093] Step S4 includes the following steps:

[0094] Step S41: Standardize the learning data of the financial risk precursor factors by performing early warning indicator processing to obtain standardized risk precursor factor data;

[0095] Step S42: Calculate multi-level early warning thresholds based on the standardized risk pre-factor data to obtain risk early warning threshold interval data, wherein the risk early warning threshold interval data includes safety threshold, attention threshold, early warning threshold and danger threshold;

[0096] Step S43: Map the standardized risk pre-factor data to the risk warning threshold range data to obtain the financial risk warning level signal;

[0097] Step S44: Send the financial risk warning level signal to the management decision terminal to execute the enterprise financial risk warning.

[0098] Step S43 includes the following steps:

[0099] Step S431: Construct warning level mapping rules based on the risk warning threshold range data to obtain a warning level mapping rule library;

[0100] Step S432: Based on the warning level mapping rule base, perform real-time matching analysis on the standardized risk precondition factor data to generate a preliminary warning level signal;

[0101] Step S433: Correct the preliminary warning level signal using an expert knowledge base to obtain the corrected warning level signal;

[0102] Step S434: Automatically generate a warning report for the corrected warning level signal to obtain a financial risk warning level signal and corresponding decision-making suggestions.

[0103] In this embodiment of the invention, when standardizing the early warning indicators of the learning data of financial risk antecedent factors, the min-max standardization method is first used to linearly transform the original values ​​of each antecedent factor indicator to the range of zero to one. The data is then dimensionless by calculating the difference between each indicator value and its minimum value, divided by the difference between its maximum and minimum values. For indicators with outliers, the Z-score standardization method is used, which eliminates the influence of data distribution skew by subtracting the indicator's average value from each indicator value and then dividing by the indicator's standard deviation. Simultaneously, considering the differences in the importance of different indicators in risk warning, the entropy weight method is used to calculate the objective weight of each indicator. The degree of dispersion of each indicator value is determined by calculating its information entropy value in the overall data, and indicators with greater dispersion are assigned higher weights. Next, the standardized data undergoes a consistency check to ensure that indicators with different dimensions are comparable in the comprehensive evaluation, ultimately resulting in standardized risk antecedent factor data with unified dimensions and weight distribution.

[0104] When calculating multi-level early warning thresholds based on standardized risk pre-factor data, the first step is to use percentile statistics to determine the risk distribution characteristics of each indicator. Risk level boundaries are defined by calculating the different percentile values ​​of each indicator in historical data. For the safety threshold, the 75th percentile of each indicator during historical normal operation is selected as the benchmark value; indicators above this value are considered to be in a safe state. The attention threshold is the 85th percentile of each indicator in historical data; a risk attention mechanism is activated when an indicator exceeds this value but does not reach the early warning threshold. The early warning threshold is calculated dynamically, based on the relative level of each indicator within the industry and combined with the company's historical performance. Cluster analysis is used to divide the indicator data of similar companies into different risk clusters, and the lower boundary of the higher-risk cluster is taken as the early warning line. The danger threshold is determined through extreme value theory analysis, calculating the extreme values ​​of each indicator before historical crisis events, and taking the average extreme value of the indicator in multiple crisis events as the danger threshold. Simultaneously, considering the synergistic effect between indicators, canonical correlation analysis is used to determine the composite threshold when multiple indicators deteriorate simultaneously, ultimately forming a statistically significant risk warning threshold range data containing four levels.

[0105] When mapping standardized risk pre-factor data to early warning levels based on risk warning threshold range data, a rule base for mapping early warning levels based on fuzzy comprehensive evaluation is first constructed. By establishing the membership function of each indicator to different early warning levels, the precise indicator values ​​are transformed into the degree of membership to each early warning level. Then, a multilayer perceptron neural network model is used to realize the nonlinear mapping from standardized indicator data to early warning levels. By training the relationship between risk events and indicator performance in historical data, the network can automatically learn complex risk patterns. In the real-time matching analysis process, the nearest neighbor classification algorithm is used to calculate the similarity between the current indicator data and the typical features of each historical early warning level. By matching with the standard patterns in the early warning level mapping rule base, a preliminary early warning level signal is generated. At the same time, time series analysis technology is introduced to consider the dynamic trend of indicator data, and indicators that continue to deteriorate are given higher early warning weights to ensure the timeliness and accuracy of early warning results. Example

[0106] Please see Figure 2 As shown, an intelligent financial risk early warning system based on management decision-making is presented. The system includes a capital chain vulnerability assessment data acquisition module, a vulnerability evolution trajectory data acquisition module, a risk pre-factor collection module, and a financial risk early warning terminal.

[0107] The modules described above are connected via wired and / or wireless means to enable data transmission between them.

[0108] The cash flow vulnerability assessment data acquisition module collects enterprise financial and operational data through a multi-source data acquisition system to obtain enterprise financial and operational data; it then performs a cash flow vulnerability assessment analysis on the enterprise financial and operational data to obtain cash flow vulnerability assessment data.

[0109] Vulnerability Evolution Trajectory Data Acquisition Module: Based on the vulnerability assessment data of the capital chain, third-party credit assessment data is imported and processed to obtain third-party credit assessment data; based on the third-party credit assessment data, dynamic simulation of the enterprise's capital chain vulnerability is performed to obtain the capital chain vulnerability evolution trajectory data;

[0110] Risk Antecedent Factor Acquisition Module: Based on the evolution trajectory data of capital chain vulnerability, the risk transmission effect of capital chain rupture is quantified to obtain the capital chain rupture risk transmission effect index; based on the capital chain rupture risk transmission effect index, financial risk antecedent factors are identified to obtain financial risk antecedent factor learning data;

[0111] Financial risk early warning terminal: Based on the learning data of the aforementioned financial risk pre-factors, it performs multi-level early warning threshold analysis and generates early warning signals to obtain the financial risk early warning level, and sends the financial risk early warning level to the management decision terminal to execute the enterprise financial risk early warning.

[0112] It should be noted that the interval and threshold sizes are set for ease of comparison. The size of the threshold depends on the amount of sample data and the base number set by those skilled in the art for each set of sample data, as long as it does not affect the proportional relationship between the parameter and the quantized value. Furthermore, the above formulas are all dimensionless calculations, and the formulas are derived from software simulations using a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0113] It should be understood that in the various embodiments of this application, the order of the above-mentioned processes does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0114] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0115] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An intelligent financial risk early warning method based on management decision-making, characterized in that, Includes the following steps: Step S1: Collect enterprise financial operation data through a multi-source data acquisition system to obtain enterprise financial operation data; conduct a cash flow vulnerability assessment and analysis on the enterprise financial operation data to obtain cash flow vulnerability assessment data; Step S2: Based on the vulnerability assessment data of the capital chain, the credit assessment data of the third-party institution is introduced and processed to obtain the third-party credit assessment data; based on the third-party credit assessment data, the capital chain breakage probability mutation point is detected and the evolution stage is divided, and the enterprise capital chain vulnerability is dynamically simulated by combining the Markov chain model to obtain the capital chain vulnerability evolution trajectory data. Step S3: Based on the evolution trajectory data of the vulnerability of the capital chain, the risk transmission effect of the capital chain breakage is quantified by constructing a risk transmission network and evaluating the criticality of the transmission path, so as to obtain the capital chain breakage risk transmission effect index; the financial risk antecedent factors are identified according to the capital chain breakage risk transmission effect index, so as to obtain the financial risk antecedent factor learning data. Step S4: Based on the learning data of the financial risk antecedent factors, perform multi-level early warning threshold analysis and generate early warning signals to obtain the financial risk early warning level, and send the financial risk early warning level to the management decision terminal to execute the enterprise financial risk early warning.

2. The intelligent financial risk early warning method based on management decision-making according to claim 1, characterized in that, Step S1 includes the following steps: Step S11: Collect enterprise financial operation data through a multi-source data acquisition system to obtain enterprise financial operation data. The multi-source data acquisition system includes a financial statement data acquisition module, a bank transaction data acquisition module, a tax declaration data acquisition module, and an operating data acquisition module. Step S12: Perform data cleaning and standardization on the enterprise's financial operation data to obtain standardized financial operation data; Step S13: Conduct a cash flow vulnerability assessment and analysis on the standardized financial operating data to obtain preliminary cash flow vulnerability assessment data; Step S14: Extract multi-dimensional vulnerability indicators from the preliminary cash flow vulnerability assessment data to obtain the cash flow vulnerability assessment data.

3. The intelligent financial risk early warning method based on management decision-making according to claim 1, characterized in that, Step S2 includes the following steps: Step S21: Obtain credit assessment data from third-party institutions, wherein the credit assessment data from third-party institutions includes bank credit rating data, supplier credit assessment data, and customer credit rating data; Step S22: Based on the vulnerability assessment data of the capital chain, conduct a credibility analysis of the third-party credit assessment data to obtain the credibility weight of the third-party credit assessment data; Step S23: Perform weighted fusion processing on the third-party credit assessment data according to the credibility weight of the third-party credit assessment data to obtain weighted third-party credit assessment data; Step S24: Based on the vulnerability assessment data of the capital chain and the weighted third-party credit assessment data, conduct a dynamic simulation of the vulnerability of the enterprise's capital chain to obtain the trajectory data of the vulnerability of the capital chain.

4. The intelligent financial risk early warning method based on management decision-making according to claim 3, characterized in that, Step S23 includes the following steps: Step S231: Based on the vulnerability assessment data of the capital chain, conduct a correlation analysis of the third-party credit assessment data to obtain the correlation coefficient of the credit assessment data; Step S232: Based on the correlation coefficient of the credit assessment data, perform correlation screening on the credit assessment data of third-party institutions to obtain high-correlation credit assessment data; Step S233: Perform a time series consistency test on the highly correlated credit assessment data to obtain the time series consistency index of the credit assessment data; Step S234: Perform dynamic weighted fusion of credit assessment data based on the time series consistency index of credit assessment data to obtain weighted third-party credit assessment data.

5. The intelligent financial risk early warning method based on management decision-making according to claim 3, characterized in that, Step S24 includes the following steps: Step S241: Construct a capital chain stress test scenario based on the capital chain vulnerability assessment data and weighted third-party credit assessment data to obtain capital chain stress test scenario data; Step S242: Perform multi-scenario cash flow simulation on the cash flow stress test scenario data to obtain multi-scenario cash flow simulation data; Step S243: Calculate the probability of capital chain rupture based on multi-scenario cash flow simulation data to obtain the probability distribution data of capital chain rupture. Step S244: Fit the evolution trajectory of the vulnerability of the capital chain based on the probability distribution data of the capital chain breakage to obtain the evolution trajectory data of the vulnerability of the capital chain; Step S245: Identify vulnerability inflection points in the trajectory data of the vulnerability evolution of the capital chain to obtain vulnerability inflection point data of the capital chain; Step S246: Set the early warning threshold for the collapse of the capital chain based on the inflection point data of the capital chain vulnerability, and obtain the early warning threshold data for the collapse of the capital chain.

6. The intelligent financial risk early warning method based on management decision-making according to claim 5, characterized in that, Step S244 includes the following steps: Time series trend analysis was performed on the probability distribution data of capital chain rupture to obtain the time series trend curve of capital chain rupture probability. Abrupt change point detection was performed on the time series trend curve of the probability of capital chain breakage to obtain abrupt change point data of the probability of capital chain breakage. Based on the data on the probability of sudden changes in the funding chain breakage, the evolution stages of funding chain vulnerability are divided to obtain data on the evolution stages of funding chain vulnerability. Based on the data on the evolution of vulnerability in the capital chain, an analysis of the vulnerability transmission relationship between stages was conducted to obtain vulnerability stage transmission relationship data; Based on the Markov chain model, the vulnerability transmission relationship data of the financial chain is fitted to obtain the vulnerability evolution trajectory data of the financial chain.

7. The intelligent financial risk early warning method based on management decision-making according to claim 1, characterized in that, Step S3 includes the following steps: Step S31: Construct a risk transmission network from the data on the evolution trajectory of the vulnerability of the capital chain to obtain the risk transmission network data of the capital chain; Step S32: Quantify the risk transmission effect of capital chain rupture based on the data of the capital chain risk transmission network to obtain the capital chain rupture risk transmission effect index; Step S33: Identify the antecedent factors of financial risk based on the risk transmission effect index of the broken capital chain and the evolution trajectory data of the capital chain vulnerability, and obtain the set of antecedent factors of financial risk; Step S34: Ranking and learning the set of financial risk antecedent factors by importance to obtain financial risk antecedent factor learning data.

8. The intelligent financial risk early warning method based on management decision-making according to claim 7, characterized in that, Step S32 includes the following steps: Step S321: Perform node importance analysis on the risk transmission network data of the capital chain to obtain node importance data of the risk transmission network; Step S322: Evaluate the criticality of the risk transmission path based on the importance data of the nodes in the risk transmission network to obtain the criticality data of the risk transmission path; Step S323: Calculate the risk transmission speed based on the criticality data of the risk transmission path to generate a risk transmission speed index; calculate the risk transmission range of the risk transmission speed index to obtain a risk transmission range index. Step S324: Calculate the transmission effect intensity of the risk transmission path criticality data based on the risk transmission range index to obtain the risk transmission effect intensity data; Step S325: Perform multi-level transmission effect accumulation processing based on the risk transmission effect intensity data to obtain risk multi-level transmission accumulation effect data; Step S326: Quantify the risk transmission effect index of capital chain rupture based on the cumulative effect data of risk multi-level transmission, and obtain the risk transmission effect index of capital chain rupture.

9. The intelligent financial risk early warning method based on management decision-making according to claim 1, characterized in that, Step S4 includes the following steps: Step S41: Standardize the learning data of the financial risk precursor factors by performing early warning indicator processing to obtain standardized risk precursor factor data; Step S42: Calculate multi-level early warning thresholds based on the standardized risk pre-factor data to obtain risk early warning threshold interval data, wherein the risk early warning threshold interval data includes safety threshold, attention threshold, early warning threshold and danger threshold; Step S43: Map the standardized risk pre-factor data to the risk warning threshold range data to obtain the financial risk warning level signal; Step S44: Send the financial risk warning level signal to the management decision terminal to execute the enterprise financial risk warning.

10. An intelligent financial risk early warning system based on management decision-making, characterized in that, It is implemented based on any one of claims 1-9, a method for intelligent financial risk early warning based on management decision-making, comprising: The cash flow vulnerability assessment data acquisition module collects enterprise financial and operational data through a multi-source data acquisition system to obtain enterprise financial and operational data; it then performs a cash flow vulnerability assessment analysis on the enterprise financial and operational data to obtain cash flow vulnerability assessment data. Vulnerability Evolution Trajectory Data Acquisition Module: Based on the vulnerability assessment data of the capital chain, third-party credit assessment data is introduced and processed to obtain third-party credit assessment data; based on the third-party credit assessment data, dynamic simulation of the enterprise's capital chain vulnerability is performed to obtain the capital chain vulnerability evolution trajectory data; Risk Antecedent Factor Acquisition Module: Based on the evolution trajectory data of capital chain vulnerability, the risk transmission effect of capital chain rupture is quantified to obtain the capital chain rupture risk transmission effect index; based on the capital chain rupture risk transmission effect index, financial risk antecedent factors are identified to obtain financial risk antecedent factor learning data; Financial risk early warning terminal: Based on the learning data of the aforementioned financial risk pre-factors, it performs multi-level early warning threshold analysis and generates early warning signals to obtain the financial risk early warning level, and sends the financial risk early warning level to the management decision terminal to execute the enterprise financial risk early warning.

Images