Financial industry chain risk identification and analysis method

Abstract

The invention provides a financial industry chain risk identification and analysis method, which comprises the following steps: quantitatively analyzing the influence of purchase settlement mode conversion on the productivity utilization rate of an upstream supplier by comparing the difference of productivity utilization rate data in two time periods before and after settlement mode conversion, including the variation amplitude and the variation direction of the productivity utilization rate; the method comprises the following steps: acquiring fund mobility index data of a downstream enterprise of an industrial chain, the fund mobility index data comprising a cash flow coverage rate and a flow ratio, and evaluating a fund mobility risk level of the downstream enterprise by analyzing the fund mobility index data of the downstream enterprise; and according to a risk conduction critical path analysis result, a supply chain financial service scheme is designed in a targeted manner, including providing financial service suggestions of account receivable financing and inventory pledge financing for high-risk node enterprises, and the service scheme is dynamically adjusted.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for identifying and analyzing risks in a financial industry chain. Background Art

[0002] The financial industry chain refers to the network of enterprises involved in the production and distribution of products and services, including upstream suppliers, downstream customers, and various intermediaries. In the financial industry chain, the change in the raw material procurement model has a significant impact on the production and operation of upstream suppliers. When downstream enterprises change the settlement method of purchase advance payments from spot to acceptance bills, upstream suppliers face a series of chain reactions. First, the use of acceptance bills prolongs the supplier's capital recovery cycle, resulting in a decrease in its inventory turnover rate. Suppliers need more working capital to maintain production and operation, but the cash conversion ability of acceptance bills is relatively weak, which puts higher requirements on the supplier's cash flow management. Secondly, in order to adapt to the changes in payment methods of downstream enterprises, suppliers may need to adjust production plans and inventory management strategies. However, in the case of uncertain market demand, suppliers face the risk of declining capacity utilization. The decline in capacity utilization means that the supplier's fixed cost allocation increases, further compressing its profit margin. At the same time, suppliers may also face the risk of raw material price fluctuations, and the use of acceptance bills as a settlement method limits their ability to hedge risks. Finally, this change in procurement settlement methods may also trigger the transmission of credit risks in the industry chain. If upstream suppliers delay delivery or reduce product quality due to financial pressure, it will affect the production and operation of downstream enterprises and may trigger a chain reaction, which will have an amplifying effect on the credit risk of the entire industrial chain. Therefore, in the financial industrial chain, changes in procurement settlement methods have posed new challenges to the production and operation of upstream suppliers and the credit risk management of the entire industrial chain. Summary of the invention

[0003] The present invention provides a method for identifying and analyzing risks in a financial industry chain, which mainly includes: Analyze the raw material procurement model in the financial industry chain, obtain the inventory turnover rate data of raw material suppliers, extract the change characteristics of the supplier inventory turnover rate data, and obtain time series data reflecting the change trend of the supplier's inventory turnover rate by analyzing the change characteristics; obtain the change data of the raw material procurement advance payment settlement method, the change data includes the specific time node when the procurement settlement method is changed from spot exchange to acceptance bill, and divide the capacity utilization rate data of upstream suppliers based on the time node to obtain the capacity utilization rate data in the two time periods before and after the settlement method change; by comparing the difference in capacity utilization rate data in the two time periods before and after the settlement method change, quantitatively analyze the impact of the change in procurement settlement method on the capacity utilization rate of upstream suppliers, including the change range and direction of capacity utilization rate; obtain the credit rating data of upstream suppliers, combined with the change range and direction of capacity utilization rate , analyze the correlation between the magnitude and direction of changes in capacity utilization and the credit risk of suppliers, and predict the credit risk of suppliers at different capacity utilization levels; obtain the liquidity indicator data of downstream enterprises in the industrial chain, which includes cash flow coverage and current ratio, and evaluate the liquidity risk level of downstream enterprises by analyzing the liquidity indicator data of downstream enterprises; match the credit risk probability distribution of upstream suppliers with the liquidity risk assessment results of downstream enterprises, build an industrial chain credit risk transmission network, calculate the importance of each node enterprise in the risk transmission network, and identify the key path of risk transmission; according to the results of the risk transmission key path analysis, design supply chain financial service solutions in a targeted manner, including providing financial service recommendations for accounts receivable financing and inventory pledge financing to high-risk node enterprises, and dynamically adjust service solutions.

[0004] The technical solution provided by the embodiment of the present invention may have the following beneficial effects: The present invention discloses a method for identifying and analyzing risks in the financial industry chain. By analyzing the impact of changes in raw material procurement patterns on the capacity utilization of upstream suppliers, combined with supplier inventory turnover and credit rating data, a machine learning model is constructed to predict supplier credit risk. At the same time, the liquidity risk of downstream enterprises is assessed, and an industrial chain credit risk transmission network is established. Identify the key paths of risk transmission and design targeted financing services for high-risk node enterprises. The present invention realizes a comprehensive analysis and prediction of industrial chain financial risks, provides data support and decision-making basis for supply chain financial services, effectively reduces financial risks, and improves the overall operational efficiency of the supply chain. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1 The present invention is a flow chart of a method for identifying and analyzing risks in a financial industry chain.

[0006] Figure 2It is a schematic diagram of a financial industry chain risk identification and analysis method of the present invention. DETAILED DESCRIPTION

[0007] In order to further understand the content of the present invention, the present invention is described in detail in conjunction with the accompanying drawings and embodiments. The present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant inventions, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the invention are shown in the accompanying drawings.

[0008] like Figure 1-2 In this embodiment, a method for identifying and analyzing risks in a financial industry chain may specifically include: S101. Analyze the raw material procurement pattern in the financial industry chain, obtain the inventory turnover data of the raw material suppliers, extract the change characteristics of the supplier inventory turnover data, and obtain time series data reflecting the change trend of the supplier inventory turnover rate by analyzing the change characteristics.

[0009] The inventory capital occupancy value is obtained by multiplying the raw material inventory corresponding to the supplier type by the purchase unit price, and the capital occupancy rate is obtained by dividing the inventory capital occupancy value by the total purchase amount; the delivery time and warehousing time are extracted from the supplier delivery record, the transportation cycle data is calculated based on the delivery time and warehousing time, and the average warehousing cycle data is obtained through the warehousing time and the outbound time; the inventory backlog coefficient is obtained by dividing the number of inventory goods by the number of goods sold, and the unit inventory cost is calculated based on the warehousing fees, capital costs, and insurance costs.

[0010] Specifically, the raw material inventory corresponding to the supplier type in the raw material supplier database is used to calculate the supplier inventory capital occupancy value by multiplying the purchase unit price by the inventory. The capital occupancy value is divided by the total purchase amount to obtain the capital occupancy rate value. The supplier inventory turnover rate is quantitatively calculated to obtain the inventory turnover basic data set. Data preprocessing is performed on the inventory turnover basic data set. The median filling method is used to process outliers, and the sliding average method is used to process noise data to obtain the preprocessed inventory turnover data set. The supplier delivery time and warehousing time are extracted from the supplier delivery record to calculate the transportation cycle data. The average storage cycle is calculated based on the warehousing time and the outbound time. The transportation cycle data and the storage cycle data are input as feature values ​​into the random forest algorithm to obtain the inventory turnover prediction model. For the preprocessed inventory turnover data set, the capital turnover period is calculated by combining the supplier's accounts receivable turnover days and accounts payable turnover days. The support vector regression method is used to fit the inventory turnover time series data to obtain the inventory turnover rate change curve. The inventory backlog coefficient is obtained by dividing the number of inventory goods by the number of goods sold. The unit inventory cost is calculated by warehousing fees, capital costs, and insurance costs. The inventory backlog coefficient and unit inventory cost are input into the inventory turnover forecasting model to obtain the supplier inventory turnover time series data. The sliding time window method is used to divide the inventory turnover time series data, and the data is sampled by the training set and validation set split ratio. The inventory turnover forecasting model is verified and parameter optimized to obtain the supplier inventory turnover forecast result. The supplier inventory turnover analysis involves the calculation and prediction of multiple related indicators. Take an electronic component supplier as an example. The supplier provides integrated circuit chip products. The monthly raw material inventory is 5,000 pieces, and the purchase unit price is 750 yuan. The inventory capital occupation value is 3.75 million yuan, and the monthly purchase total is 15 million yuan. The capital occupation rate is calculated to be 0.25, indicating that one-quarter of the funds are occupied by inventory goods, which is an important basis for calculating the basic data of inventory turnover. For abnormally fluctuating inventory data, the median filling method is used to process it. For example, if the inventory data of a certain month shows 25,000 pieces that are obviously deviated from the normal range, the median of 6,000 pieces of the three months before and after is calculated for replacement, and then the sliding average method with a 30-day window is used to eliminate short-term fluctuations to make the data smoother. The supplier's delivery records show that it takes an average of 72 hours from shipment to warehousing, including 48 hours of transportation time and 24 hours of loading and unloading and inspection time, which constitutes the transportation cycle data. The average storage time of products in the warehouse is 45 days. The transportation cycle and storage cycle data are input into the random forest model for inventory turnover prediction. The calculation of the capital turnover period needs to consider the accounts receivable turnover days and accounts payable turnover days. The supplier's average accounts receivable collection period is 60 days, the accounts payable payment period is 45 days, and the capital turnover period is 105 days. Combined with the previous inventory turnover data, support vector regression is used to establish a time series model.The inventory backlog analysis shows that the monthly inventory quantity is 4,500 pieces, the average monthly sales quantity is 3,000 pieces, and the calculated inventory backlog coefficient is 1.5, indicating that the inventory backlog is high. The unit inventory cost calculation includes: storage fee of 15 yuan / piece / month, capital cost (calculated at an annualized rate of 6%) of 3.75 yuan / piece / month, and insurance cost of 2 yuan / piece / month, totaling 20.75 yuan / piece / month. The prediction model uses a 90-day sliding time window, based on the data of the last 360 days, and divides the training set and the validation set in a ratio of 8:2. Model verification shows that the average error between the prediction result and the actual inventory turnover rate is within 7%, and the prediction curve can better reflect the changing trend of the supplier's inventory turnover. By optimizing the prediction parameters, such as adjusting the time window size to 60 days, the prediction error is further reduced to within 5%. The prediction results show that the supplier's inventory turnover rate will show a slow upward trend in the next three months, which provides data support for suppliers to improve inventory management.

[0011] S102. Obtain data on changes in the settlement method for advance payments for raw material purchases. The change data includes the specific time point when the purchase settlement method is changed from cash on delivery to acceptance bills. The capacity utilization data of upstream suppliers is divided based on the time point to obtain the capacity utilization data in the two time periods before and after the settlement method is changed.

[0012] The payment method code and payment timestamp information are obtained from the supplier's payment settlement record, and the settlement method change time node data is obtained through the time series clustering method; the single-hot encoding is used to convert the spot exchange and acceptance bill types for the settlement method change time node data, and the settlement method feature data set is obtained by combining the account period length and the payment amount; the machine equipment utilization rate is calculated by dividing the equipment operating time by the standard working time, and the equipment utilization rate is smoothed through a sliding time window to obtain the equipment utilization rate time series data; a deep neural network method is used to take the equipment utilization rate time series data and the raw material utilization efficiency data as input features, and the production completion rate is used as the output label to obtain the capacity utilization rate prediction result.

[0013] Specifically, the payment method code and payment timestamp information are extracted from the supplier's payment settlement record, and the time segmentation point when the payment method changes from spot exchange to acceptance bill is identified by the time series clustering method, so as to obtain the settlement method change time node data. The payment method code is numerically processed, and the spot exchange and acceptance bill types are converted by unique hot encoding. The settlement method change time node data is extracted by combining the account period length and the payment amount, so as to obtain the settlement method feature data set. The machine equipment running time data is extracted from the supplier's production record, and the machine equipment utilization rate is calculated by dividing the equipment running time by the standard working time. The equipment utilization rate data is smoothed by the sliding time window method to obtain the equipment utilization rate time series data. The single batch of raw material quota consumption is obtained according to the raw material purchase order record, and the actual raw material input is extracted from the production record. The raw material utilization coefficient is calculated by dividing the actual input by the quota consumption to obtain the raw material utilization efficiency data. The deep neural network method is used to take the equipment utilization rate time series data and the raw material utilization efficiency data as input features, and the output completion rate is used as the output label to construct the capacity utilization prediction model and obtain the capacity utilization prediction result. The capacity utilization forecast results are segmented according to the time node of the settlement method change. The time window division method is used to divide the data into two time periods before and after the change. The capacity utilization data set in the two time periods is obtained through data standardization. The analysis of supplier settlement method changes involves the processing and association of multiple data dimensions. The payment settlement records of a manufacturing supplier show that the payment method code 1 represents cash payment and 2 represents acceptance bill payment. Through the analysis of the payment records in 2023, it is identified that the payment method change occurred on March 15. The data of 90 days before and after this time point show that the proportion of cash payment has dropped from 85% to 15%. In the feature extraction of settlement data, cash payment is converted to [1,0] and acceptance bill payment is converted to [0,1]. Combined with the account period data, the average account period of cash payment is 15 days, and the account period of acceptance bill payment is extended to 90 days. The data of single payment amount shows that the average cash payment is 500,000 yuan and the average acceptance bill payment is 1.5 million yuan. The machine and equipment operation data records show that the standard daily working time of a single key equipment is 16 hours. The actual operation time data shows that the average operation time in January was 14.4 hours, and the equipment utilization rate was 0.9; the average operation time in February was 12.8 hours, and the equipment utilization rate dropped to 0.8; in March, it further dropped to 11.2 hours, and the equipment utilization rate dropped to 0.7. After the moving average processing of the 30-day window, the equipment utilization curve is smoother, highlighting the downward trend. Raw material consumption analysis shows that the product production quota requires a single batch of 100 kg of raw materials. The actual production records show that the average input in January was 105 kg, and the raw material utilization coefficient was 0.952; the average input in February was 108 kg, and the raw material utilization coefficient dropped to 0.926; the average input in March was 110 kg, and the raw material utilization coefficient further dropped to 0.909.The deep neural network model adopts a three-layer structure. The input layer contains two features: equipment utilization and raw material utilization coefficient. The number of hidden layer nodes is 16, and the output layer predicts capacity utilization. The training data uses data from January to February, and the verification data uses data from March. The model prediction results show that the capacity utilization rate gradually decreases from 0.95 in January to 0.75 in March. The capacity utilization data of the 90 days before the settlement method change time point is standardized and the mean is 0.92, and the standard deviation is 0.03; the data of the 90 days after the change is standardized and the mean is reduced to 0.82, and the standard deviation is increased to 0.05. The data distribution shows the characteristics of downward mean and increased volatility. This trend of change is consistent with the situation that the increase in cash flow pressure on suppliers leads to fluctuations in production and operation conditions. Through the time window analysis method, the degree and duration of the impact of the change in settlement method on the supplier's capacity utilization level are clearly shown.

[0014] S103. By comparing the differences in capacity utilization data in the two time periods before and after the change in settlement method, quantitative analysis is conducted on the impact of the change in procurement settlement method on the capacity utilization of upstream suppliers, including the magnitude and direction of change in capacity utilization.

[0015] The inventory turnover time series data are obtained, and the wavelet transform method is used to denoise the time series data to obtain a smoothed data set; the sliding window length and step size parameters are set according to the smoothed data set, and the window segmentation method is used to obtain two time series samples before and after the settlement method is changed; for the two time series samples, a prediction model is established using a long short-term memory network, and the turnover prediction result is obtained through the prediction model; based on the turnover prediction result and the equipment startup time data extracted from the supplier's production record, a capacity prediction model is established using support vector regression, and the capacity utilization rate change trend function is obtained through the capacity prediction model; the change rate of adjacent time points is calculated through the capacity utilization rate change trend function, the change direction is determined according to the positive and negative values ​​of the change rate, and the change amplitude is determined by the absolute value of the change rate.

[0016] Specifically, the time stamp and numerical features are extracted from the inventory turnover time series data, and the wavelet transform method is used to denoise the data to obtain smooth data. The data period characteristics are calculated according to the autocorrelation function to obtain the turnover basic data set. The time dimension is segmented for the turnover basic data set, and the sliding window length and step parameters are set. The window segmentation method is used to obtain two time series samples before and after the settlement method is changed. The long short-term memory network is used to establish a prediction model for the two time series samples. The number of neurons in the input layer is set as the feature dimension, the number of neurons in the hidden layer is set as the time step, and the output layer is a single neuron to obtain the turnover prediction results. The equipment startup time data is extracted from the supplier's production records, the equipment startup rate per unit time is calculated, and the production efficiency data is obtained by combining the product qualification rate. The support vector regression is used to establish the capacity prediction model. According to the prediction results of the support vector regression model, the mean, variance, and skewness characteristics of the capacity utilization rate before and after the period are calculated, and the quantitative indicators of capacity change are obtained through the feature difference. The autoregressive moving average model is established by using the difference equation. The quantitative indicator sequence of capacity change is used as input data. The autoregressive order and the moving average order are set to obtain the capacity utilization rate change trend function. The rate of change of adjacent time points is calculated by the capacity utilization rate change trend function. The direction of change is determined according to the positive and negative values ​​of the rate of change, and the amplitude of change is determined by the absolute value of the rate of change. The inventory turnover time series analysis involves data processing and modeling in multiple links. Taking a certain machinery manufacturing supplier as an example, the original inventory turnover data is recorded in daily units. The data from January to June 2023 show that the daily average fluctuates between 3.5 and 4.2. The db4 wavelet basis function is used for 4-layer decomposition and denoising, and the short-term random fluctuations are removed to obtain a smoother turnover rate data curve. The autocorrelation function calculation shows that the data has a 30-day periodic feature. When the turnover rate data is segmented in the time dimension, a window length of 90 days is set, and the step length is 1 day. 90 samples are obtained before and after the corresponding settlement method change time point. The segmented data showed that the turnover rate before the change was 4.1 and the standard deviation was 0.2; after the change, the mean dropped to 3.7 and the standard deviation increased to 0.3. In the construction of the long short-term memory network model, 7 neurons were set in the input layer to correspond to 7 days of data, 30 neurons were set in the hidden layer to correspond to the monthly cycle, and 1 neuron in the output layer predicted the turnover rate of the next day. The training data used the data of the first 60 days, and the validation set used the data of the last 30 days. The root mean square error of the model prediction was 0.15. The production records showed that the standard working hours of the equipment were 16 hours per day, the median of the actual startup time data was 14 hours, and the startup rate was 0.875. The product quality inspection records showed that the qualified rate dropped from 0.98 before the change to 0.95 after the change, and the comprehensive production efficiency dropped from 0.858 to 0.831. The support vector regression uses the radial basis kernel function, the penalty parameter C is set to 1.0, and the mean square error of the predicted capacity utilization is 0.023.The calculation of capacity characteristics shows that the mean before the change is 0.92, the variance is 0.0025, and the skewness is -0.15. The mean after the change is 0.85, the variance is 0.0049, and the skewness is -0.28. The characteristic differences are -0.07, 0.0024, and -0.13, respectively, indicating that the capacity level has decreased and the volatility has increased. The autoregressive moving average model sets the autoregressive order p=2, the moving average order q=1, and the difference order d=1. The model expression is: , where B is the lag operator, which acts on the time series Xt, is a white noise sequence. The parameter estimation is =0.8, = -0.2, =0.3. The calculation of the change rate of capacity utilization at adjacent time points shows that in the first 10 days after the settlement method was changed, the average daily change rate was -0.5%, and the change direction was downward; in the following 20 days, the average daily change rate converged to -0.2%, still maintaining a downward trend; and finally stabilized, with the absolute value of the average daily change rate being less than 0.1%. The change range gradually transitioned from a significant decline in the early stage to a stable state in the later stage.

[0017] The capacity utilization data of upstream suppliers in the two time periods before and after the change in procurement settlement method are obtained, the mean and standard deviation of the capacity utilization in each time period are calculated, the time series graph and frequency distribution histogram of the capacity utilization are drawn, the changing trend, volatility and distribution characteristics of the capacity utilization are analyzed, and a T test is performed on the capacity utilization data of the two time periods to determine whether there is a significant difference in capacity utilization before and after the change in procurement settlement method.

[0018] The original data of capacity utilization in the supplier's production database is obtained, and the outliers are eliminated by using the box plot method based on the original data to obtain the preprocessed capacity utilization data set; the normality test is performed on the preprocessed capacity utilization data set by using the chi-square goodness of fit test method to obtain the distribution type of the data set; according to the distribution type of the data set, the probability density function is constructed by using the kernel density estimation method, and the data distribution curve is plotted by setting the kernel function type and the bandwidth parameter; for the data distribution curve, if the Levene test statistic determines that the variance is homogeneous, the independent sample T test method is used to calculate the T statistic to obtain the conclusion of the significance of the capacity utilization difference.

[0019] Specifically, the time point of settlement method change is obtained from the supplier's production database, and the original data of capacity utilization is divided into the data series before and after the change by using timestamp marks. The outliers are eliminated by the box plot method to obtain the preprocessed capacity utilization data set. Descriptive statistics are calculated for the preprocessed capacity utilization data set, and the statistical characteristics of the data set are obtained by the arithmetic mean, standard deviation, skewness, and kurtosis formulas. The data dispersion is calculated according to the coefficient of variation to obtain the distribution characteristics of the data set. The normality of the data set is tested by the chi-square goodness of fit test method, the data distribution type is judged by the quantile statistic, and the time series processing method is determined according to the data distribution characteristics to obtain the data processing path. For the determined data processing path, the probability density function is constructed by the kernel density estimation method, and the data distribution curve is drawn by setting the kernel function type and bandwidth parameters to obtain the frequency distribution characteristics of the capacity utilization. The time series data fitting function is established by the support vector regression method, and the nonlinear characteristics are processed by the radial basis kernel function. The rate of change is calculated according to the first-order derivative of the curve to obtain the trend characteristics of the capacity utilization change. The variance homogeneity test was performed on the data sets before and after the change, and the variance homogeneity was judged by the Levene test statistic. The T test type was selected according to the test results to obtain the statistical test scheme. According to the selected T test scheme, the significance level was set as the preset test parameter, and the P value judgment result was obtained by calculating the T statistic and the degree of freedom, and the conclusion of the significance of the difference in capacity utilization was obtained. The capacity utilization data analysis process involves multiple links such as data processing, distribution feature extraction and statistical testing. Taking a certain automotive parts supplier as an example, the supplier changed its settlement method on April 15, 2023. The daily capacity utilization data for 90 days before and after the change were extracted from the original production data. The box plot method was used to set 1.5 times the upper and lower quartile range as the outlier judgment boundary, and 3 outlier points in the data before the change and 5 outlier points in the data after the change were identified. After removing these abnormal data points, a valid sample was obtained. Descriptive statistical analysis was performed on the preprocessed data. The arithmetic mean of the data before the change was 0.87, the standard deviation was 0.052, the skewness was -0.15, and the kurtosis was 2.8; the arithmetic mean of the data after the change was 0.82, the standard deviation was 0.068, the skewness was -0.28, and the kurtosis was 3.2. The coefficient of variation was calculated by dividing the standard deviation by the mean. It was 0.060 before the change and increased to 0.083 after the change, indicating that the data volatility increased. The chi-square goodness of fit test was used to evaluate the data distribution characteristics. The data was divided into 10 intervals to calculate the frequency. At the significance level of 0.05, the chi-square statistic of the data before the change was 14.2, which was less than the critical value of 16.9, and it obeyed the normal distribution; the chi-square statistic of the data after the change was 18.5, which was greater than the critical value, showing non-normal characteristics.The kernel density estimation uses the Gaussian kernel function, and the bandwidth parameter is selected as 0.015 through the cross-validation method. The generated probability density curve shows that the data distribution after the change is flatter and right-skewed, reflecting the characteristics of the decline in capacity utilization and increased volatility. In the support vector regression modeling, the radial basis kernel function is used to deal with nonlinear characteristics, the penalty parameter C is set to 1.0, and the kernel function parameter γ is taken as 0.1. The first-order derivative of the time series curve obtained by fitting shows that the average daily decline rate during the period after the change is 0.03%. When performing the variance homogeneity test, the Levene test statistic F value is 5.82, and the corresponding P value is 0.017, which is less than the significance level of 0.05, indicating that the variances of the two groups of data are not homogeneous. Based on this, the Welch-corrected T test method is selected, and the significance level is set to 0.05. The calculated T statistic is 5.93, the degree of freedom is 167.4, and the corresponding P value is 0.000003, which is much less than the significance level, indicating that there is a significant difference in capacity utilization before and after the settlement method is changed. Specifically, after the change, the average level of capacity utilization decreased by 0.05 percentage points, and the volatility increased significantly.

[0020] S104. Obtain credit rating data of upstream suppliers, analyze the correlation between the magnitude and direction of changes in capacity utilization and the credit risk of suppliers in combination with the magnitude and direction of changes in capacity utilization, and predict the credit risk of suppliers at different levels of capacity utilization.

[0021] The historical rating levels and rating scores are obtained from the supplier credit rating database, and the change amplitude and change direction are calculated according to the capacity utilization time series data; for the historical rating levels and the rating scores, the rating level quantification rules are used to convert the historical rating levels into numerical ratings, and the rating scores are normalized by the maximum and minimum normalization method to obtain standardized rating data; according to the standardized rating data, the Kalman filtering method is used to set the state vector, and the state vector includes the rating score and the change rate in the standardized rating data, and the filtered rating data is obtained by iterative calculation of the observation equation and the state equation; for the filtered rating data, the random forest algorithm is used to construct a credit risk prediction model, and the credit risk of the supplier at different capacity utilization levels is predicted by the credit risk prediction model.

[0022] Specifically, historical rating levels and rating scores are obtained from supplier credit rating records, the change range and change direction are calculated based on the capacity utilization time series data, and the credit rating records are segmented in time series using a monthly sliding window to obtain a credit rating basic data set. Data standardization is performed on the credit rating basic data set, and the letter rating is converted into a numerical rating through the rating level quantification rule. The rating score is standardized using the maximum and minimum value normalization method to obtain a standardized rating data set. The standardized rating data set is processed using the Kalman filter method, and the state vector is set to include the rating score and the change rate. The filtered rating data set is obtained by iterative calculation of the observation equation and the state equation. Based on the filtered rating data set, the principal component analysis method is used to reduce the dimension of the capacity change characteristics and credit rating indicators, and the cumulative contribution rate threshold is set to select the number of principal components to obtain a reduced dimension feature matrix. The random forest algorithm is used to construct a credit risk prediction model, and the number of decision trees and tree depth parameters are set. The feature importance is calculated using the Gini coefficient to obtain a feature weight vector. According to the feature weight vector, key variables are extracted from the reduced dimension feature matrix, and the Monte Carlo method is used to generate capacity utilization samples. The risk probability distribution is obtained by probability density function calculation. According to the risk probability distribution, the capacity utilization quantile is set as the risk level classification standard, and the conditional probability is calculated using the Bayesian method to obtain the credit risk prediction results under different capacity levels. Supplier credit risk prediction involves multidimensional data processing and model construction. Taking a supplier of electronic components as an example, the supplier's credit rating record shows that the rating levels from January to June 2023 are AAA, AA+, AA, AA-, A+, and A, and the corresponding rating scores are 100, 95, 90, 85, 80, and 75, respectively. The capacity utilization data shows that it has gradually decreased from 90% to 75%, with a monthly change range of -2% to -4%. A 30-day sliding window is used to segment the data, and the rating level is quantitatively converted according to the percentage system. AAA corresponds to 100 points, and each small level decreases by 5 points. Through the maximum and minimum value normalization processing, the normalized rating scores are distributed between 0.5 and 1, and the standard deviation is reduced from 0.08 to 0.05. The Kalman filter sets the state vector to include two components: rating score and monthly change rate. The observation equation describes the measurement process of the rating score, and the state equation describes the evolution law of the rating score. The filtering results obtained through iterative calculation show that the noise variance of the rating score is reduced from 0.25 to 0.1, and the signal is smoother. In the principal component analysis, the feature matrix includes 12 indicators such as capacity utilization, change range, rating score, and cash ratio. The calculated eigenvalues ​​show that the cumulative contribution rate of the first three principal components reaches 85%. These three principal components are selected as the features after dimensionality reduction, and the dimension of the reduced feature matrix is ​​reduced from 12 to 3. The random forest model setting includes 100 decision trees, and the maximum depth of each tree is 5 layers.The feature importance calculated by the Gini coefficient shows that the weight of capacity utilization is 0.35, the weight of rating score is 0.30, the weight of cash ratio is 0.20, and the weight of other features accounts for 15%. The Monte Carlo method generates 10,000 sets of capacity utilization samples, and the sample values ​​are evenly distributed between 60% and 95%. The probability density function obtained by kernel density estimation shows that when the capacity utilization rate is lower than 70%, the risk probability shows an exponential growth trend. In the Bayesian method, the 25% quantile of capacity utilization is set to 75%, the 50% quantile is set to 85%, and the 75% quantile is set to 90% as the risk level classification standard. The calculated conditional probability shows that when the capacity utilization rate is lower than 75%, the credit risk probability exceeds 0.8; when it is between 75% and 85%, the risk probability is between 0.4 and 0.8; when it is between 85% and 90%, the risk probability is between 0.2 and 0.4; when it is higher than 90%, the risk probability is lower than 0.2. The confidence interval of the prediction results fluctuates within ±0.05 at a 95% confidence level.

[0023] Calculate the correlation coefficient between the change in capacity utilization and the supplier's credit rating, analyze the linear correlation between the change in capacity utilization and the supplier's credit rating, establish a classification prediction model with the change in capacity utilization and direction as independent variables and supplier credit risk as the dependent variable, and use the trained classification prediction model to predict the probability of supplier credit risk occurring at the target capacity utilization level.

[0024] The original data of capacity utilization are obtained from the capacity change database, the interquartile range method is used to identify outliers, and the capacity utilization preprocessing data set is obtained through the median filling method; the change amplitude and change direction of the capacity utilization preprocessing data set are calculated, and the correlation value between the capacity change and the credit rating is obtained through the Pearson correlation coefficient calculation method; the capacity utilization change amplitude and the change direction are used to construct a feature matrix, and the feature matrix is ​​normalized through the z-score standardization method to obtain a standardized feature matrix; a cross entropy loss function is constructed for the standardized feature matrix, and the parameters of the logistic regression model are optimized through the gradient descent method to obtain a credit risk classification function.

[0025] Specifically, the capacity utilization data is extracted from the supplier capacity change database, the interquartile range method is used to identify outliers, and the missing values ​​are processed by the median filling method to obtain the capacity utilization preprocessing data set. The change range and change direction are calculated for the capacity utilization preprocessing data set, and the rating score sequence is obtained by using the credit rating record. The correlation value between capacity change and credit rating is obtained by the Pearson correlation coefficient calculation method. According to the change range and change direction of capacity utilization, a feature matrix is ​​constructed, and the feature data is normalized by using the z-score standardization method. The training data set and the validation data set are divided by the 80:2 ratio. For the standardized feature matrix, the cross entropy loss function is used to construct a logistic regression model, and the model parameters are optimized by the gradient descent method to obtain the credit risk classification function. The classification function is tested using the validation data set, and the model performance is evaluated by the accuracy, recall rate, and F1 score to obtain the model evaluation index value. The target capacity utilization level is used to construct the prediction sample, and the risk probability value is calculated by the credit risk classification function to obtain the risk prediction result. For the risk prediction result, the bootstrap method is used to generate a resampled data set, and the interval estimate of the risk probability is obtained by the confidence interval calculation formula. Supplier credit risk classification prediction involves multiple links such as data preprocessing, correlation analysis and model building. Taking a certain automotive parts supplier as an example, the original capacity utilization data shows that the daily record value fluctuates between 55% and 95%. The interquartile range method is used to calculate Q1 to be 75%, Q3 to be 85%, and the interquartile range IQR is 10%. Data less than 60% and greater than 100% are identified as outliers and filled with the median 80%. The capacity utilization change range is calculated by the difference between adjacent time points. Monthly data show that the change range is distributed between -15% and 5%, and the change direction is marked by the positive and negative values ​​of the range. Credit rating data shows that the rating level has dropped from AAA to BBB+, and the corresponding rating score has dropped from 100 to 75. The correlation coefficient between the capacity change range and the credit rating calculated by the Pearson correlation coefficient is 0.82, showing a strong positive correlation. The feature matrix construction includes two dimensions: the capacity utilization change range and the change direction. The change range is expressed in percentage, and the change direction is expressed by 1 and -1. Through z-score standardization, the change amplitude is converted to a standard normal distribution with a mean of 0 and a standard deviation of 1. The 180-day data is divided into a training set of 144 days and a validation set of 36 days in a ratio of 8:2. The logistic regression model uses cross entropy as the loss function, and iteratively optimizes the parameters by gradient descent. The learning rate is set to 0.01 and the number of iterations is 1000. The feature weights obtained after the model converges show that the weight of the capacity change amplitude is 0.75, the weight of the change direction is 0.25, and the model bias term is 0.3. The model evaluation on the validation data set shows that the accuracy rate reaches 85%, the recall rate is 82%, and the F1 score is 0.835.The confusion matrix shows that among the 36 validation samples, there are 20 true positives, 10 true negatives, 3 false positives, and 3 false negatives. For the case of 75% target capacity utilization, the change amplitude of the predicted sample is -10%, the change direction is -1, and the risk probability value is 0.85 when substituted into the classification function. The 95% confidence interval of the risk probability is calculated to be [0.82, 0.88] by resampling 1000 times using the bootstrap method. This prediction method comprehensively considers the impact of the amplitude and direction of capacity changes on credit risk, eliminates the impact of different dimensions through standardization, and the model evaluation indicators show that the prediction results have good accuracy and stability. The calculation of the confidence interval provides the uncertainty range of the risk probability prediction, providing more comprehensive reference information for risk assessment.

[0026] S105. Obtain the liquidity indicator data of the downstream enterprises in the industrial chain. The liquidity indicator data include cash flow coverage ratio and current ratio. By analyzing the liquidity indicator data of the downstream enterprises, the liquidity risk level of the downstream enterprises is assessed.

[0027] The cash flow statement and balance sheet data in the downstream enterprise database are obtained, and the cash flow coverage ratio data is calculated by dividing the cash and cash equivalents by short-term liabilities; the outliers are detected by the box plot method based on the cash flow coverage ratio data, and the outliers are filled by the moving median method to obtain standardized liquidity indicator data; the standardized liquidity indicator data is decomposed by the exponential smoothing method, and the trend item and seasonal item data are obtained by time series decomposition; a support vector regression prediction function is established based on the trend item and seasonal item data, and if the frequency band energy eigenvalue of the prediction function meets the preset threshold after hierarchical clustering grouping, the corresponding liquidity risk level is determined.

[0028] Specifically, the monthly cash flow statement and balance sheet data are extracted from the financial database of downstream enterprises. The cash flow coverage ratio is calculated by dividing cash and cash equivalents by short-term liabilities, and the current ratio is calculated by dividing current assets by current liabilities to obtain the initial indicator data of capital liquidity. The initial indicator data of capital liquidity are standardized, and the outliers are detected and eliminated by the box plot method. The missing values ​​are filled by the moving median method to obtain the processed indicator data of capital liquidity. The trend of the processed indicator data is extracted by the cubic exponential smoothing method, and the smoothing coefficient is set to the preset parameter value. The trend term, seasonal term and residual term data are obtained by time series decomposition. According to the trend decomposition results, the support vector regression method is used to establish the liquidity prediction function, and the nonlinear characteristics are processed by the radial basis kernel function to obtain the liquidity indicator prediction value. Wavelet packet decomposition is performed on the liquidity indicator prediction value, and the decomposition layer number and wavelet basis function type are set. The band energy eigenvalue is obtained by energy entropy calculation. The hierarchical clustering method is used to group the band energy eigenvalues, and the optimal number of clusters is determined by the inter-class distance calculation to obtain the liquidity risk level classification result. A naive Bayes classifier is established based on the historical risk event database, and the characteristic probability is extracted from the liquidity risk level classification results. The probability distribution of the liquidity risk of downstream enterprises is obtained through conditional probability calculation. The liquidity analysis of downstream enterprises involves the processing and risk assessment of multiple financial indicators. Taking an electronic product manufacturing company as an example, the cash and cash equivalent balance of 10 million yuan and short-term liabilities of 20 million yuan are extracted from the monthly financial statements, and the cash flow coverage ratio is calculated to be 0.5; the current assets are 30 million yuan and the current liabilities are 20 million yuan, and the current ratio is calculated to be 1.5. These basic indicators reflect the short-term debt repayment ability of the enterprise. The initial indicator data is standardized, and the upper and lower bounds are set as Q3+1.5IQR and Q1-1.5IQR respectively through the box plot method, and 2 outlier points in the cash flow coverage data, 0.1 and 0.9, and 2 outlier points in the current ratio data, 0.8 and 2.5, are identified. The moving median of the previous and next three months is used to fill these outliers to make the data smoother. The triple exponential smoothing process uses an additive model, setting the horizontal term smoothing coefficient to 0.3, the trend term smoothing coefficient to 0.2, and the seasonal term smoothing coefficient to 0.1. The decomposition results show that the cash flow coverage ratio has obvious seasonality, with a high point of 0.6 in the fourth quarter of each year and a low point of 0.4 in the second quarter. The seasonality of the current ratio is relatively weak, with a fluctuation range of 1.3 to 1.7. The support vector regression uses the radial basis kernel function, with the penalty parameter C set to 1.0 and the kernel function parameter γ set to 0.1. The prediction results show that the cash flow coverage ratio in the next three months shows a downward trend, with the predicted values ​​being 0.45, 0.42, and 0.38 respectively; the current ratio is relatively stable, with the predicted value fluctuating around 1.45. The wavelet packet decomposition uses the db4 wavelet basis function and sets a 3-layer decomposition structure.Energy entropy calculation shows that the energy proportion of cash flow coverage ratio in the high frequency band has increased from 15% to 25%, indicating that short-term fluctuations have intensified; the energy of the current ratio is mainly concentrated in the low frequency band, accounting for about 60%, indicating that the overall trend is stable. Hierarchical clustering uses Euclidean distance measurement, and the optimal number of clusters is determined to be 4 through inter-class distance calculation, corresponding to the four levels of extremely high risk, high risk, medium risk and low risk. The clustering results show that when the cash flow coverage ratio is lower than 0.3, it is extremely high risk, 0.3 to 0.4 is high risk, 0.4 to 0.5 is medium risk, and higher than 0.5 is low risk. The naive Bayes classifier is trained based on historical risk event data and contains 50 historical risk samples. The risk probability distribution calculated based on the current indicator prediction value shows: extremely high risk probability 0.05, high risk probability 0.25, medium risk probability 0.45, low risk probability 0.25. This shows that the company is currently at a medium-to-high risk level and needs to continuously monitor its liquidity status.

[0029] S106. Match the credit risk probability distribution of upstream suppliers with the liquidity risk assessment results of downstream enterprises, build an industrial chain credit risk transmission network, calculate the importance of each node enterprise in the risk transmission network, and identify the key path of risk transmission.

[0030] Obtain inter-enterprise transaction records in the supplier database, build a node set according to the enterprise identifiers in the transaction records, calculate the connection strength through the transaction amount and transaction frequency, and obtain the initial adjacency matrix of the industrial chain; for the initial adjacency matrix, use the credit risk probability and the capital liquidity risk value to calculate the edge weights, and standardize the edge weights through the maximum and minimum normalization method to obtain the weighted adjacency matrix of the industrial chain; construct a Markov transfer probability matrix according to the weighted adjacency matrix, calculate the eigenvector through the power iteration method, and if the eigenvector residual is less than the convergence threshold, obtain the node importance value; use the node importance value to set the path search starting point, traverse the network path through the depth-first search method, and if the path length is less than the path threshold, add the path to the risk transmission network.

[0031] Specifically, we extract inter-enterprise transaction records from the supplier database, build node sets through enterprise identifiers, calculate the connection strength between enterprises based on transaction amount and transaction frequency, and obtain the initial adjacency matrix of the industrial chain. We preprocess the data for the initial adjacency matrix, standardize the connection strength using the maximum and minimum normalization method, calculate the edge weights using the credit risk probability and the liquidity risk value, and obtain the weighted adjacency matrix of the industrial chain. We use the node degree statistics method to calculate the network density index, and use the in-degree and out-degree calculation formulas to count the number of connections for each node. We obtain the network connectivity index based on the distribution characteristics of the number of connections. We construct a Markov transition probability matrix for the weighted adjacency matrix, set the damping factor parameter value, calculate the eigenvector using the power iteration method, and obtain the initial value of the node importance. We use the page ranking algorithm to iteratively update the initial value of the node importance, set the convergence threshold parameter, and use the residual calculation method to determine the iteration termination condition to obtain the final value of the node importance. We set the path search starting point according to the final value of the node importance, use the depth-first search method to traverse the network path, and prune the path using the path length threshold to obtain the candidate conduction path set. The risk transmission coefficient is calculated for the candidate transmission path set, and the product of the importance of the path nodes is used as the transmission strength. The key path is screened by the transmission strength sorting method to obtain the risk transmission network structure. The risk transmission network analysis of the industrial chain involves multiple links such as network construction, node sorting and path identification. Taking a certain automotive parts industry chain as an example, it includes 5 vehicle manufacturers, 15 first-tier suppliers and 30 second-tier suppliers. The connection strength is calculated based on the monthly transaction records between enterprises. The transaction amount ranges from 100,000 yuan to 10 million yuan, and the transaction frequency ranges from 1 to 20 times per month. The connection strength value is obtained by multiplying the transaction amount by the transaction frequency. The connection strength is normalized and the value is mapped to the [0,1] interval using (x-min) / (max-min). Where x is the original data, min and max represent the minimum and maximum values ​​of all values ​​in the data set, respectively. Combined with the supplier credit risk probability, which takes values ​​from 0.1 to 0.9 and the liquidity risk value, which takes values ​​from 0.2 to 0.8, the edge weights are calculated by weighted average, and the weight coefficients are set to 0.4 and 0.6 respectively. In the constructed 50×50 adjacency matrix, the proportion of non-zero elements is about 15%. The network density calculation shows that the average in-degree of first-tier suppliers is 3.5 and the average out-degree is 2.8; the average in-degree of second-tier suppliers is 2.2 and the average out-degree is 1.5. The overall network density is 0.15, showing typical hierarchical characteristics. The degree distribution follows a power law distribution, and a few core enterprises have a high degree of connectivity. The Markov transition probability matrix is ​​obtained by row normalization, and the damping factor is set to 0.85. The main eigenvector is calculated by the power iteration method, and the initial vector elements are all set to 1 / 50. The iteration is stopped when the residual is less than 0.0001, and it usually takes 20 to 30 iterations to converge.The iterative results of the page ranking algorithm show that the importance of the top 5 nodes are 0.158, 0.142, 0.126, 0.115 and 0.098 respectively. These nodes are mainly first-tier suppliers with large business scale. The importance of the remaining nodes is less than 0.05, showing a long-tail distribution feature. The depth-first search starts with the top 3 nodes in importance, sets the path length threshold to 4, and filters out paths with a sum of importance less than 0.2 through pruning rules. There are about 50 candidate paths obtained by the search, covering the main supply relationships. The risk transmission coefficient is calculated using the geometric mean of the importance of the nodes on the path. The transmission coefficients of the top 5 key paths are 0.132, 0.128, 0.121, 0.115 and 0.108 respectively. These paths usually connect the core enterprises in the industrial chain and reflect the main channels for risk transmission. The key paths identified in this way provide an important basis for industrial chain risk management.

[0032] S107. Based on the results of the risk transmission key path analysis, design supply chain financial service solutions in a targeted manner, including providing financial service recommendations for accounts receivable financing and inventory pledge financing to high-risk node enterprises, and dynamically adjust service solutions.

[0033] According to the calculation formula of the financial indicators of the enterprise's debt-to-asset ratio, current ratio and quick ratio, the weighted average method is used to obtain the enterprise's basic credit score; for the enterprise's basic credit score, the enterprise's comprehensive credit rating is obtained through normalization processing of the enterprise's scale characteristics, industry characteristics and historical credit records; using the enterprise's comprehensive credit rating, the principal component analysis method of fixed asset turnover rate, inventory turnover rate and accounts receivable turnover rate is used to obtain the enterprise's operating status assessment value; based on the enterprise's comprehensive credit rating and the enterprise's operating status assessment value, the radial basis kernel function of support vector regression is used to obtain the enterprise's credit line value.

[0034] Specifically, high-risk node enterprise identifiers are extracted from the key path of risk transmission, and the enterprise's debt-to-asset ratio, current ratio, and quick ratio are calculated using the financial indicator calculation formula, and the enterprise's basic credit score is obtained through the weighted average method. For the basic credit score of the enterprise, a comprehensive scoring indicator is constructed in combination with the enterprise's scale, industry characteristics, and historical credit records, and the enterprise's comprehensive credit rating is obtained through normalization. The fixed asset turnover rate, inventory turnover rate, and accounts receivable turnover rate are used to calculate the enterprise's operating efficiency indicators, and the key features are extracted through the principal component analysis method to obtain the enterprise's operating status assessment value. Based on the enterprise's comprehensive credit rating and operating status assessment value, support vector regression is used to construct a credit limit calculation function, and the radial basis kernel function is used to process nonlinear features to obtain the enterprise's credit limit value.

[0035]

[0036] ，C reditRepresents the final credit limit, R1 represents the enterprise's comprehensive credit rating, R2 represents the operating status assessment value, k1 and k2 represent the corresponding weight coefficients, and β represents the basic credit limit. For the enterprise's accounts receivable and inventory data, the decision tree algorithm is used to establish a financing plan classifier, set the tree depth and leaf node number parameters, and calculate the optimal split point through the Gini coefficient to obtain the type of financing plan. According to the type of financing plan, a set of monitoring indicators is set, including capital turnover rate indicators, asset pledge rate indicators, and cash flow coverage rate indicators. The risk monitoring standard is obtained through the indicator warning threshold setting rules. A monthly sliding window is used to track the monitoring indicators, and the risk status changes are calculated through the indicator volatility. The financing plan adjustment parameters are obtained by combining the historical data distribution characteristics. According to the risk status changes and adjustment parameters, the new financing quota and guarantee requirements are calculated, and the adjusted financing service plan is generated through the financing plan update rules. The design of supply chain financial service solutions involves matching corporate credit assessment with financing plans. Take a certain auto parts manufacturing company as an example. The company has total assets of 50 million yuan and total liabilities of 30 million yuan, with a calculated asset-liability ratio of 60%; current assets of 20 million yuan, current liabilities of 15 million yuan, with a calculated current ratio of 1.33; quick assets of 12 million yuan, and a quick ratio of 0.8. By setting weights of 0.4, 0.3, and 0.3 to calculate the weighted average, the basic credit score is 0.75. Combined with the company's operating data, the annual operating income is 40 million yuan, the company scale index is 0.8; the prosperity index of the auto parts industry is 0.85; the on-time repayment rate in the past 12 months is 98%, and the historical credit record index is 0.9. The comprehensive credit rating is calculated using weights of 0.3, 0.3, and 0.4, and the score is 0.82, corresponding to the AA credit rating. The calculation of operating efficiency indicators shows that the original value of fixed assets is 20 million yuan, the annual operating income is 40 million yuan, and the fixed asset turnover rate is 2.0; the annual sales cost is 30 million yuan, the average inventory is 8 million yuan, and the inventory turnover rate is 3.75; the accounts receivable is 9 million yuan, and the accounts receivable turnover rate is 4.44. The first principal component extracted by principal component analysis explains 90% of the variance, and the operating status assessment value is 0.78. Support vector regression uses radial basis kernel function, parameter γ is set to 0.1, and combined with the business scale of the enterprise, the base credit limit of 10 million yuan is calculated. Through the credit rating and operating status adjustment coefficient of 0.9, the final credit limit is determined to be 9 million yuan. The decision tree is set to a maximum depth of 4 layers and a minimum number of leaf node samples of 20. The characteristics of accounts receivable show that the account period is within 90 days, accounting for 75%, and the amount concentration is 0.3; the inventory characteristics show that raw materials account for 40% and finished products account for 45%. The classification results recommend the use of accounts receivable pool financing of 6 million yuan and inventory pledge financing of 3 million yuan.In the monitoring indicator settings, the capital turnover rate indicator is calculated by dividing the monthly operating income by the average occupied funds, and the warning value is set to 1.5; the asset pledge rate indicator is calculated by dividing the pledged asset valuation by the financing amount, and the warning value is set to 1.8; the cash flow coverage rate indicator is calculated by dividing the monthly operating cash inflow by the financing repayment amount, and the warning value is set to 1.2. A 3-month sliding window is used to monitor indicator changes, and a risk warning is triggered when the indicator volatility exceeds 20%. The latest monitoring shows that the capital turnover rate is 1.6, the asset pledge rate is 2.0, and the cash flow coverage rate is 1.3. All indicators are within a safe range. Based on the current risk status, it is recommended to maintain the existing financing plan and dynamically monitor the company's operating conditions.

[0037] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying and analyzing risks in a financial industry chain, characterized in that: The method comprises: Analyze the raw material procurement model in the financial industry chain, obtain the inventory turnover rate data of raw material suppliers, extract the change characteristics of the supplier inventory turnover rate data, and obtain time series data reflecting the change trend of the supplier's inventory turnover rate by analyzing the change characteristics; obtain the change data of the raw material procurement advance payment settlement method, the change data includes the specific time node when the procurement settlement method is changed from spot exchange to acceptance bill, and divide the capacity utilization rate data of upstream suppliers based on the time node to obtain the capacity utilization rate data in the two time periods before and after the settlement method change; by comparing the difference in capacity utilization rate data in the two time periods before and after the settlement method change, quantitatively analyze the impact of the change in procurement settlement method on the capacity utilization rate of upstream suppliers, including the change range and direction of capacity utilization rate; obtain the credit rating data of upstream suppliers, combined with the change range and direction of capacity utilization rate , analyze the correlation between the magnitude and direction of changes in capacity utilization and the credit risk of suppliers, and predict the credit risk of suppliers at different capacity utilization levels; obtain the liquidity indicator data of downstream enterprises in the industrial chain, which includes cash flow coverage and current ratio, and evaluate the liquidity risk level of downstream enterprises by analyzing the liquidity indicator data of downstream enterprises; match the credit risk probability distribution of upstream suppliers with the liquidity risk assessment results of downstream enterprises, build an industrial chain credit risk transmission network, calculate the importance of each node enterprise in the risk transmission network, and identify the key path of risk transmission; according to the results of the risk transmission key path analysis, design supply chain financial service solutions in a targeted manner, including providing financial service recommendations for accounts receivable financing and inventory pledge financing to high-risk node enterprises, and dynamically adjust service solutions.

2. The method according to claim 1, characterized in that The raw material procurement model in the financial industry chain is analyzed to obtain the inventory turnover rate data of the raw material supplier, and the change characteristics of the supplier inventory turnover rate data are extracted. By analyzing the change characteristics, time series data reflecting the change trend of the supplier inventory turnover rate is obtained, including: The inventory capital occupancy value is obtained by multiplying the raw material inventory corresponding to the supplier type by the purchase unit price, and the capital occupancy rate value is obtained by dividing the inventory capital occupancy value by the total purchase amount; Extract the delivery time and warehousing time from the supplier's delivery record, calculate the transportation cycle data based on the delivery time and warehousing time, and obtain the average storage cycle data through the warehousing time and outbound time; The inventory backlog coefficient is obtained by dividing the number of inventory items by the number of goods sold, and the unit inventory cost is calculated based on warehousing costs, capital costs and insurance costs.

3. The method according to claim 1, characterized in that The change data of the settlement method of raw material purchase advance payment is obtained, and the change data includes the specific time node when the purchase settlement method is changed from spot exchange to acceptance bill. The capacity utilization data of the upstream supplier is divided based on the time node to obtain the capacity utilization data in the two time periods before and after the settlement method is changed, including: Obtain payment method code and payment timestamp information from supplier payment settlement records, and obtain settlement method change time node data through time series clustering method; For the data of the settlement method change time node, one-hot encoding is used to convert the spot exchange and acceptance bill types, and the settlement method feature data set is obtained by combining the account period length and payment amount; Calculate the machine equipment utilization rate by dividing the equipment operation time by the standard working time, and smooth the equipment utilization rate through a sliding time window to obtain equipment utilization rate time series data; A deep neural network method is adopted, the equipment utilization time series data and the raw material utilization efficiency data are taken as input features, and the production completion rate is taken as the output label to obtain the capacity utilization prediction result.

4. The method according to claim 1, characterized in that: By comparing the difference in capacity utilization data between the two time periods before and after the settlement method change, the impact of the change in procurement settlement method on the capacity utilization of upstream suppliers is quantitatively analyzed, including the magnitude and direction of change in capacity utilization, including: Obtaining inventory turnover rate time series data, and performing denoising processing on the time series data using a wavelet transform method to obtain a smoothed data set; According to the smoothed data set, the sliding window length and step size parameters are set, and the window segmentation method is used to obtain two time series samples before and after the settlement method is changed; For the two time series samples, a prediction model is established using a long short-term memory network, and a turnover rate prediction result is obtained through the prediction model; Based on the turnover rate prediction results and the equipment startup time data extracted from the supplier's production records, a capacity prediction model is established using support vector regression, and a capacity utilization rate change trend function is obtained through the capacity prediction model; The rate of change of adjacent time points is calculated through the capacity utilization rate change trend function, the direction of change is determined according to the positive and negative values ​​of the rate of change, and the amplitude of change is determined by the absolute value of the rate of change; it also includes: obtaining the capacity utilization data of upstream suppliers in the two time periods before and after the change of procurement settlement method, calculating the mean and standard deviation of the capacity utilization in each time period, drawing the time series graph and frequency distribution histogram of the capacity utilization, analyzing the change trend, volatility and distribution characteristics of the capacity utilization, and performing a T test on the capacity utilization data of the two time periods to determine whether there is a significant difference in capacity utilization before and after the change of procurement settlement method.

5. The method according to claim 4, characterized in that The method of obtaining the capacity utilization data of upstream suppliers in the two time periods before and after the change of the purchase settlement method, calculating the mean and standard deviation of the capacity utilization in each time period, drawing a time series graph and a frequency distribution histogram of the capacity utilization, analyzing the changing trend, volatility and distribution characteristics of the capacity utilization, and performing a T test on the capacity utilization data of the two time periods to determine whether there is a significant difference in the capacity utilization before and after the change of the purchase settlement method includes: Obtaining original capacity utilization data from the supplier's production database, and removing outliers using a box plot method based on the original data to obtain a preprocessed capacity utilization data set; For the preprocessed capacity utilization data set, a normality test is performed using the chi-square goodness of fit test method to obtain the distribution type of the data set; according to the distribution type of the data set, a probability density function is constructed using the kernel density estimation method, and a data distribution curve is drawn by setting the kernel function type and bandwidth parameters; for the data distribution curve, if the Levene test statistic determines that the variance is homogeneous, the independent sample T test method is used to calculate the T statistic to obtain a conclusion on the significance of the capacity utilization difference.

6. The method according to claim 1, characterized in that The above method obtains the credit rating data of upstream suppliers, combines the change range and direction of capacity utilization, analyzes the correlation between the change range and direction of capacity utilization and the credit risk of suppliers, and predicts the credit risk of suppliers at different capacity utilization levels, including: Obtain historical rating levels and rating scores from the supplier credit rating database, and calculate the magnitude and direction of change based on the capacity utilization time series data; With respect to the historical rating levels and the rating scores, the historical rating levels are converted into numerical ratings using rating level quantification rules, and the rating scores are normalized using a maximum and minimum value normalization method to obtain standardized rating data; According to the standardized rating data, a Kalman filter method is used to set a state vector, wherein the state vector includes the rating score and the change rate in the standardized rating data, and the filtered rating data is obtained by iterative calculation of the observation equation and the state equation; For the filtered rating data, a random forest algorithm is used to construct a credit risk prediction model, and the credit risk of suppliers at different capacity utilization levels is predicted through the credit risk prediction model; it also includes: calculating the correlation coefficient between the change in capacity utilization and the supplier's credit rating, analyzing the linear correlation between the change in capacity utilization and the supplier's credit rating, taking the change in capacity utilization and the direction of change as independent variables and the supplier's credit risk as the dependent variable, establishing a classification prediction model, and using the trained classification prediction model to predict the probability of supplier credit risk occurring at the target capacity utilization level.

7. The method according to claim 6, characterized in that The calculation of the correlation coefficient between the capacity utilization rate change range and the supplier's credit rating, the analysis of the linear correlation between the capacity utilization rate change range and the supplier's credit rating, the capacity utilization rate change range and change direction as independent variables, the supplier's credit risk as the dependent variable, the establishment of a classification prediction model, and the use of the trained classification prediction model to predict the probability of supplier credit risk at the target capacity utilization level, including: The original data of capacity utilization are obtained from the capacity change database, the interquartile range method is used to identify outliers, and the capacity utilization preprocessing data set is obtained through the median filling method; the change amplitude and change direction of the capacity utilization preprocessing data set are calculated, and the correlation value between the capacity change and the credit rating is obtained through the Pearson correlation coefficient calculation method; the capacity utilization change amplitude and the change direction are used to construct a feature matrix, and the feature matrix is ​​normalized through the z-score standardization method to obtain a standardized feature matrix; a cross entropy loss function is constructed for the standardized feature matrix, and the parameters of the logistic regression model are optimized through the gradient descent method to obtain a credit risk classification function.

8. The method according to claim 1, characterized in that The acquisition of the capital liquidity indicator data of the downstream enterprises in the industrial chain, the capital liquidity indicator data includes the cash flow coverage ratio and the current ratio, and the capital liquidity risk level of the downstream enterprises is evaluated by analyzing the capital liquidity indicator data of the downstream enterprises, including: Obtain cash flow statement and balance sheet data from the downstream enterprise database, and calculate cash flow coverage ratio data by dividing cash and cash equivalents by short-term liabilities; Using the box plot method to detect outliers based on the cash flow coverage data, and filling the outliers using the moving median method to obtain standardized liquidity indicator data; The standardized liquidity index data is decomposed by using an exponential smoothing method, and trend item and seasonal item data are obtained by time series decomposition; A support vector regression prediction function is established based on the trend item and seasonal item data. If the frequency band energy characteristic value of the prediction function meets the preset threshold after hierarchical clustering grouping, the corresponding liquidity risk level is determined.

9. The method according to claim 1, characterized in that: The credit risk probability distribution of upstream suppliers is matched with the liquidity risk assessment results of downstream enterprises to build a credit risk transmission network for the industrial chain, calculate the importance of each node enterprise in the risk transmission network, and identify the key paths of risk transmission, including: Obtain inter-enterprise transaction records from the supplier database, construct a node set based on the enterprise identifiers in the transaction records, calculate the connection strength based on the transaction amount and transaction frequency, and obtain the initial adjacency matrix of the industrial chain; For the initial adjacency matrix, edge weights are calculated using the credit risk probability and the capital liquidity risk value, and the edge weights are standardized using the maximum and minimum value normalization method to obtain an industrial chain weighted adjacency matrix; A Markov transition probability matrix is ​​constructed according to the weighted adjacency matrix, and a eigenvector is calculated by a power iteration method, and if the eigenvector residual is less than a convergence threshold, a node importance value is obtained; The node importance value is used to set the path search starting point, and the network path is traversed through the depth-first search method. If the path length is less than the path threshold, the path is added to the risk conduction network.

10. The method according to claim 1, characterized in that According to the results of the risk transmission key path analysis, the supply chain financial service plan is designed in a targeted manner, including providing financial service suggestions for accounts receivable financing and inventory pledge financing to high-risk node enterprises, and dynamically adjusting the service plan, including: According to the calculation formula of the financial indicators of the enterprise's debt-to-asset ratio, current ratio and quick ratio, the weighted average method is used to obtain the enterprise's basic credit score; For the basic credit score of the enterprise, the comprehensive credit rating of the enterprise is obtained by normalizing the enterprise scale characteristics, industry characteristics and historical credit records; Using the comprehensive credit rating of the enterprise, the principal component analysis method of fixed asset turnover rate, inventory turnover rate and accounts receivable turnover rate is used to obtain the enterprise operating status assessment value; According to the comprehensive credit rating of the enterprise and the evaluation value of the enterprise's operating conditions, the value of the enterprise's credit limit is obtained through radial basis kernel function processing of support vector regression.

Images