Financial risk dynamic monitoring and report generation method based on multi-source data

Abstract

The invention provides a financial risk dynamic monitoring and report generation method based on multi-source data, and the method comprises the steps: obtaining the multi-source data of credit default interchange, carrying out the preprocessing and feature extraction of interest difference data of different deadlines, and obtaining a deadline structure feature vector; obtaining a long and short-term interest difference curve based on long and short-term interest difference data obtained in real time, performing anomaly detection on the morphological change of the long and short-term interest difference curve, and triggering an evaluation process of a credit risk spreading effect when distortion anomaly of the interest difference curve is detected; in the risk assessment and monitoring process, market evaluation is continuously collected, information related to key credit default interchange attributes is extracted, the information serves as supplementary input of a risk conduction mechanism analysis and monitoring strategy, and a risk management strategy is dynamically optimized.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for dynamic monitoring and report generation of financial risks based on multi-source data. Background Art

[0002] Credit default swap spreads are a type of financial derivative used to hedge credit risk, and their spreads reflect the market's assessment of credit risk. In the process of dynamic monitoring and report generation of financial risks based on multi-source data, the term structure characteristics of credit default swap spreads are an important analysis object. When the shape of the long-term and short-term spread curves is distorted, how to accurately assess the contagion effect of credit risk and adjust the monitoring strategy accordingly is a complex technical issue. This issue involves multiple aspects: first, how to timely discover signals of spread curve distortion from massive multi-source heterogeneous data; second, how to quantitatively assess the impact of curve distortion on credit risk and consider the transmission mechanism of risk between different entities and markets; finally, how to transform complex quantitative analysis results into clear and understandable risk reports to provide effective support for decision-making. This series of problems are closely linked and are constrained by multiple factors such as data quality, model capabilities, and computing efficiency, which brings huge challenges to technical implementation. Summary of the invention

[0003] The present invention provides a method for dynamic monitoring and report generation of financial risks based on multi-source data, which mainly includes: Obtain multi-source data on credit default swaps, pre-process and extract features for interest rate spread data of different maturities, and obtain term structure feature vectors; obtain long-term and short-term interest rate spread curves based on real-time long-term and short-term interest rate spread data, perform anomaly detection on the shape changes of long-term and short-term interest rate spread curves, and trigger the evaluation process of credit risk contagion effect when an abnormal distortion of the interest rate spread curve is detected; when triggering the evaluation process of credit risk contagion effect, use the standardized term structure feature vector as the node attribute of the graph, and the abnormal distortion of the interest rate spread curve as the edge attribute, and learn the transmission path and intensity of credit risk between different maturities and market entities through graph neural networks to analyze the credit risk transmission mechanism based on graph neural networks; use the Monte Carlo simulation method to analyze the credit risk transmission mechanism learned by the graph neural network, generate risk contagion scenarios driven by abnormal changes in credit default swap interest rates of different maturities, and evaluate the cumulative losses of credit risk under different scenarios. Distribution, determine the thresholds and warning levels of risk monitoring indicators with key credit default swap attributes as the core; dynamically monitor key credit default swap attributes, and when the credit default swap spread exceeds the preset threshold, or other attributes show significant abnormal changes, trigger the risk warning and emergency response process, and take corresponding risk mitigation measures according to the warning level; in the process of risk assessment and monitoring, continuously collect market evaluations, extract information related to key credit default swap attributes, and use the information as a supplementary input for risk transmission mechanism analysis and monitoring strategies, and dynamically optimize risk management strategies; regularly generate credit risk monitoring reports, summarize risk assessment results, warning situations and response measures, as well as the cumulative loss distribution under different risk contagion scenarios, and use data visualization methods to visualize credit risk monitoring content, including time series graphs of changes in credit default swap attributes, credit risk transmission network graphs, and probability density graphs of cumulative loss distribution.

[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 dynamic monitoring and report generation of financial risks based on multi-source data. The method obtains multi-source heterogeneous data, extracts standardized term structure characteristics, monitors the anomaly of the spread curve in real time, and triggers risk assessment. The risk transmission mechanism is learned by using graph neural network, and the risk contagion scenario is generated by Monte Carlo simulation, the cumulative loss distribution is evaluated, and the monitoring indicators are determined. The key attributes are dynamically monitored, and early warning and emergency response are automatically triggered. At the same time, market evaluation is integrated to optimize risk management strategies. The method realizes comprehensive risk monitoring of the credit default swap market, can timely identify anomalies, evaluate transmission paths, predict potential losses, provide support for risk management decisions, and effectively prevent systemic credit risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Figure 1The present invention is a flowchart of a method for dynamic monitoring and report generation of financial risks based on multi-source data.

[0006] Figure 2 It is a schematic diagram of a method for dynamic monitoring and report generation of financial risks based on multi-source data of the present invention.

[0007] Figure 3 This is another schematic diagram of a method for dynamic monitoring and report generation of financial risks based on multi-source data according to the present invention. DETAILED DESCRIPTION

[0008] 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.

[0009] like Figure 1-3 In this embodiment, a method for dynamic monitoring and report generation of financial risks based on multi-source data may specifically include: S101. Obtain multi-source data of credit default swaps, perform preprocessing and feature extraction on the interest rate spread data of different maturities, and obtain a term structure feature vector.

[0010] The interest rate spread data is obtained from the database according to the time interval, and the initial interest rate spread data set is obtained by the date alignment method and the uniform processing of decimal places; the linear interpolation method is used to fill the detected null value points of the initial interest rate spread data set, and the duplicate data items are removed by distribution characteristics to obtain the cleaned interest rate spread data set; the cleaned interest rate spread data set is grouped according to the term and the moving average method is used to correct the outliers, and the data points exceeding the threshold interval are eliminated by the percentile method to obtain the smoothed interest rate spread data; the smoothed interest rate spread data is normalized by the minimum and maximum standardization method, and the term structure characteristic vector is obtained by fitting and calculation of the support vector regression model.

[0011] Specifically, the credit default swap spread data of different maturities are obtained from the stock exchange database according to the specified time interval, the interbank market quotation spread data are collected through the financial data platform, and the rating change information is obtained from the international rating agency. The collected spread data are aligned in date and the decimal places are processed uniformly to obtain the initial spread data set. The data quality of the initial spread data set is verified, the missing value detection method is used to identify the data null points, the linear interpolation method is used to fill the detected null points, and the duplicate data items are removed based on the data distribution characteristics to obtain the cleaned spread data set. The cleaned spread data set is grouped according to the term, the five-day moving average method is used to correct the outliers, the percentile method is used to calculate the upper and lower quartile interval thresholds of the spread data of each term group, and the data points exceeding the threshold interval are removed to obtain the smoothed grouped spread data. Based on the smoothed grouped spread data, the autocorrelation analysis method is used to calculate the correlation coefficient matrix between the data of each term group, and the highly correlated term group data are merged based on the correlation coefficient threshold to obtain the term structure data after dimensionality reduction. The minimum and maximum standardization method is used to normalize the term structure data after dimension reduction. The calculation method is to subtract the minimum value of each data point from the group and divide it by the numerical range of the group to obtain the standardized characteristic data. A support vector regression model is constructed for the standardized characteristic data. The Gaussian kernel function is selected as the kernel function. The penalty parameter and kernel function parameter are determined by the cross-validation method. The fitting calculation is performed on the data of each term group to obtain the fitted spread data. The root mean square error method is used to calculate the residual vector of the fitted spread data and the standardized characteristic data, and the principal component analysis method is used to extract features based on the residual vector to obtain the term structure characteristic vector. The spread data of credit default swaps usually comes from multiple market entities, and there are differences in the data formats of different sources. For example, the spread data of the stock exchange is accurate to 4 decimal places, while the interbank market quotation data is accurate to 2 decimal places, so it needs to be processed uniformly. Taking a certain corporate bond as an example, the spread data of its 5-year credit default swap in the exchange is 235.4568 basis points, and the interbank market quotation is 235.46 basis points. After unified processing, the value is 235.46 basis points. In the data quality verification stage, the linear interpolation method is used to fill in the missing values. Assuming that the 5-year credit default swap spread of a certain enterprise on February 1 and February 3 is 245.36 and 255.36 basis points respectively, and the data on February 2 is missing, the spread value on February 2 is 250.36 basis points calculated by linear interpolation. In addition, if repeated data is found on the same day, such as the repeated records of the 5-year spread of a certain enterprise on the same day are 240.35 and 240.36 basis points, the data of 240.36 basis points is retained. In the outlier processing, the 5-day moving average method is used to smooth the data points with large fluctuations.For example, the 5-year interest rate spread of a certain enterprise for 5 consecutive trading days is 260.35, 262.45, 295.36, 265.42, and 268.35 basis points respectively. The data on the third day deviates significantly, and it is corrected by calculating the 5-day moving average of 271.59 basis points. At the same time, the upper and lower quartile intervals are calculated. If the upper quartile value is 280.36 basis points and the lower quartile value is 250.36 basis points, the data points outside the interval will be eliminated. For the interest rate spread data of different terms, highly correlated term groups are identified through autocorrelation analysis. Assuming that the correlation coefficient of the 3-year and 5-year interest rate spreads is calculated to be 0.95, which exceeds the preset threshold of 0.9, the data of these two term groups are merged. In the standardization process, if the maximum value of a term group is 320.36 basis points and the minimum value is 220.36 basis points, then for the data point with a value of 270.36 basis points in the group, its standardized value is 0.5. When constructing the support vector regression model, the Gaussian kernel function is selected as the kernel function, and the kernel function parameter γ=0.01 and the penalty parameter C=100 are determined by cross-validation. For the fitting results, the root mean square error RMSE=√(Σ(y_pred-y_true)² / n) is calculated to evaluate the model effect, where y_pred is the predicted value, y_true is the actual value, and n is the number of samples. Finally, principal component analysis is performed based on the residual vector, and the principal component with an eigenvalue contribution rate of more than 85% is extracted as the term structure eigenvector.

[0012] S102. A long-term and short-term interest rate spread curve is obtained based on the long-term and short-term interest rate spread data obtained in real time, and abnormal detection is performed on the shape changes of the long-term and short-term interest rate spread curve. When an abnormal distortion of the interest rate spread curve is detected, an assessment process for the credit risk contagion effect is triggered.

[0013] According to the preset sampling frequency, long-term interest rate spread data and short-term interest rate spread data are obtained from the market database, and the basic interest rate spread data set is obtained through data standardization processing and timestamp alignment; for the basic interest rate spread data set, the cubic spline interpolation method and the least squares method are used to determine the spline function coefficients to obtain continuous and smooth interest rate spread curve data; the interest rate spread change rate between adjacent term points is calculated for the interest rate spread curve data, and the curve anomaly degree data is obtained through the sliding window method and the autoregressive moving average method; if the curve anomaly degree data exceeds the warning threshold, the risk transmission probability is calculated, and the risk transmission matrix is ​​established through the risk transmission probability to obtain the term structure risk transmission coefficient.

[0014] Specifically, long-term interest rate spread data and short-term interest rate spread data are obtained from the trading market database at a fixed frequency. Through data format standardization, data update timestamp alignment, and data validity inspection, data points that exceed two standard deviations of the average value of the previous five days are eliminated to generate long-term and short-term basic interest rate spread data. For the long-term and short-term basic interest rate spread data, the cubic spline interpolation method is used to construct a continuous and smooth interest rate spread curve. The interpolation nodes select the key points of each term. The spline function coefficients are determined by the least squares method to obtain continuous and smooth interest rate spread curve data. The interest rate spread change rate between adjacent term points is calculated for the interest rate spread curve data. The historical fluctuation range of each term point is calculated by the sliding window method. The upper and lower limits of the fluctuation range are set to positive and negative twice the volatility of the previous twenty days to obtain the interest rate spread curve fluctuation characteristic data. For the interest rate spread curve fluctuation characteristic data, the autoregressive moving average method is used to construct a prediction benchmark curve. The autoregressive order is fifth order and the moving average order is third order. The root mean square error between the actual curve and the predicted curve is calculated to obtain the interest rate spread curve abnormality data. According to the abnormal degree data of the interest rate spread curve, the warning threshold is set to twice the root mean square error of the previous thirty days, and the curve interval that exceeds the warning threshold is marked to obtain the risk attention interval data. For the risk attention interval data, the random forest algorithm based on the entropy method is used to calculate the risk transmission probability between different term points. The characteristic variables include the interest rate spread change rate, volatility, and trading volume change rate, and a risk transmission matrix is ​​established. The risk spillover coefficient and risk acceptance coefficient of each term node are calculated through the risk transmission matrix. The risk spillover coefficient represents the ability of the term to transmit risks to other terms, and the risk acceptance coefficient represents the degree to which the term takes on risks of other terms, and the term structure risk transmission assessment results are obtained. In the credit default swap market, abnormal fluctuations in long-term and short-term interest rate spread data often reflect the market's degree of concern about credit risk. Take the credit default swap of a large enterprise as an example. Its 1-year spread is 150 basis points and its 5-year spread is 280 basis points. When data anomalies occur, such as the 1-year spread suddenly rises to 220 basis points, which exceeds the average value of 145 basis points in the previous 5 days by 30 basis points, twice the standard deviation, the data point will be marked as abnormal and removed. When using cubic spline interpolation to construct a continuous and smooth spread curve, 1 year, 2 years, 3 years, 5 years, 7 years and 10 years are selected as key term points, and the spread value of each term point is set to 150, 180, 220, 280, 320 and 350 basis points respectively. The spline function coefficients are solved by the least squares method to obtain the smooth spread curve equation. , where x represents the term, and the coefficients a, b, c, and d are obtained by minimizing the sum of squared errors. When calculating the spread change rate, the spread change rate R between adjacent term points is defined as , where Y2 and Y1 are the interest rate spreads of terms T2 and T1 respectively. For example, the change rate of the interest rate spread between 2-year and 1-year is (180-150) / (2-1)=30 basis points / year. The historical volatility of each term point is calculated using a 20-day window. If the daily average volatility of a term point is 15 basis points, the upper limit of the volatility range is set to 30 basis points and the lower limit is -30 basis points. For the autoregressive moving average model, the fifth-order autoregressive term means using the data of the previous 5 trading days to predict the current value. For example, the calculation formula for the predicted value of the current 5-year interest rate spread is Y_t=0.3Y_t-1+0.25Y_t-2+0.2Y_t-3+0.15Y_t-4+0.1Y_t-5, where Y_t represents the interest rate spread value on day t, Y_t-1 represents the interest rate spread value on day t-1, and so on. The coefficients are obtained by fitting historical data. After adding the third-order moving average term, the influence of the random disturbance term is considered. In the risk transmission analysis, the entropy method is used to calculate the characteristic weights, assuming that the weights of the spread change rate, volatility, and trading volume change rate are 0.4, 0.35, and 0.25, respectively. In the constructed risk transmission matrix, the matrix element P_ij represents the probability of risk transmission from term i to term j, which is calculated by the random forest algorithm. The risk transmission path in the term structure often shows the characteristics of spreading from short-term to long-term. For example, the transmission probability from 2-year to 5-year is 0.65, while the transmission probability from 5-year to 2-year is only 0.25, indicating that abnormal fluctuations in short-term spreads are more likely to trigger adjustments in long-term spreads.

[0015] S103. When triggering the evaluation process of the credit risk contagion effect, the standardized term structure characteristic vector is used as the node attribute of the graph, and the spread curve distortion anomaly is used as the edge attribute. The transmission path and intensity of credit risk between different maturities and market entities are learned through the graph neural network to analyze the credit risk transmission mechanism based on the graph neural network.

[0016] A node attribute matrix is ​​constructed according to the standardized term structure feature vector, an edge attribute matrix is ​​constructed by using the spread curve distorted data, and the initial conduction graph data is obtained by the minimum and maximum normalization method; the weighted Euclidean distance method is used to calculate the similarity score between nodes for the initial conduction graph data, and if the similarity score is lower than the median similarity score of the node pair, the node connection is pruned to obtain the optimized conduction graph data; a graph convolution operator is designed through the Chebyshev polynomial expansion form, a feature extractor is constructed for the nodes in the optimized conduction graph data, and the weighted features of adjacent nodes are calculated to obtain the local structural features; the dot product attention calculation method is used to obtain the attention weight matrix between local structural features, and the node features are weightedly aggregated based on the attention weights to obtain the global conduction features.

[0017] Specifically, the node attribute matrix of the graph is constructed according to the standardized term structure feature vector, the edge attribute matrix is ​​constructed by using the spread curve distortion abnormal data, and the node attribute matrix and the edge attribute matrix are standardized by the minimum and maximum normalization method to obtain the initial transmission graph data. The node metric calculation is performed on the initial transmission graph data, and the similarity scores between nodes of different terms are calculated by the weighted Euclidean distance method. The minimum threshold of the similarity score is set as the median of the similarity scores of all node pairs, and the node connections below the threshold are pruned to obtain the optimized transmission graph data. For each node in the optimized transmission graph data, a three-layer perceptual field graph convolution feature extractor is constructed, and the graph convolution operator is designed using the Chebyshev polynomial expansion form to calculate the weighted features of adjacent nodes and obtain the local structural features of the nodes. For the local structural features of the nodes, the attention weight matrix between nodes is obtained by the dot product attention calculation method, and the node features are weighted and aggregated based on the attention weights to obtain the global transmission features. A four-layer graph neural network structure is constructed, each layer includes a graph convolution layer and an attention layer, and features are transferred between adjacent layers using skip connections. The network parameters are trained through node classification tasks to obtain the risk transmission map. According to the risk transmission map, the shortest path and path weight between nodes are extracted, the cumulative transmission probability and transmission delay on the path are calculated, and the credit risk transmission path and transmission intensity are quantitatively characterized to obtain the risk transmission mechanism matrix. In the credit market, the term structure characteristic vector reflects the risk characteristics of debts with different maturities. Taking a certain corporate bond as an example, the standardized characteristic values ​​of its 1-year, 3-year and 5-year maturities are 0.3, 0.5 and 0.7 respectively. These characteristic values ​​are mapped to the 0-1 interval through minimum and maximum normalization processing as the attribute values ​​of the nodes in the graph. At the same time, the degree of distortion of the spread curve, such as the sudden increase of the 1-3 year spread from 50 basis points to 120 basis points, this abnormal change is standardized and used as the attribute value of the edge connecting the two term nodes. When constructing the graph structure, the similarity between nodes is calculated by weighted Euclidean distance, and the weight is set according to the distance of the term. The weight of adjacent terms is 0.6, the weight of one term interval is 0.3, and the weight of two terms interval is 0.1. Assume that the feature vectors of two nodes are [0.3, 0.4, 0.5] and [0.4, 0.5, 0.6] respectively, and the calculated weighted distance is 0.15. If the median of all node distances is 0.2, the connection between the two nodes is retained. Graph convolution feature extraction uses third-order Chebyshev polynomial expansion, and the polynomial coefficients are obtained through training to obtain [0.8, -0.3, 0.1]. For the feature update of the central node, the feature mean of the first-order neighbors is calculated first. If the feature value of the neighbor node is [0.4, 0.5], the first-order feature is 0.45, the second-order feature is 0.42, and the third-order feature is 0.40. Finally, these features are weighted and combined to obtain the new feature of the node.In the attention mechanism, the attention score is calculated by dot product. For example, the dot product of the feature vectors of nodes i and j is 0.85. After softmax normalization, the attention weight is 0.3, which indicates the strength of the influence of node j on node i during information transmission. In the four-layer graph neural network, the output dimensions of each layer are set to 64, 32, 16 and 8 respectively. The features are transmitted between adjacent layers through jump connections to avoid information loss in deep networks. The Dijkstra algorithm is used to extract the risk transmission path, and the negative logarithm of the transmission probability is used as the path weight. Assuming that the shortest path from 1 year to 5 years passes through 3 years, the transmission probabilities of each segment are 0.8 and 0.7 respectively, then the cumulative transmission probability is 0.56, and the transmission delay is 2 term units.

[0018] S104. Use the Monte Carlo simulation method to analyze the credit risk transmission mechanism learned by the graph neural network, generate risk contagion scenarios driven by abnormal changes in credit default swaps with different maturities, evaluate the cumulative loss distribution of credit risks under different scenarios, and determine the risk monitoring indicator thresholds and warning levels with key credit default swap attributes as the core.

[0019] The drift rate and volatility parameters are calculated based on the historical data of credit spreads, and the geometric Brownian motion equation is used to generate a set of random fluctuation paths of the credit spread, wherein the geometric Brownian motion equation includes a drift term and a disturbance term. For the set of random fluctuation paths, the exponentially weighted moving average method is used to calculate the dynamic fluctuation interval value, and the credit status mutation point is identified by the ratio of the dynamic fluctuation interval value to the standard deviation to obtain a multi-dimensional risk feature sequence. For the multi-dimensional risk feature sequence, the Gaussian kernel density estimation method is used to fit the risk loss distribution function, and the bandwidth parameter of the Gaussian kernel density estimation method is obtained by the cross-validation method. For the risk loss distribution function, the generalized Pareto distribution is used to fit the tail risk characteristics, and the risk loss quantile value under the confidence interval is obtained, and the risk warning threshold is determined by the hierarchical clustering method.

[0020] Specifically, based on the risk transmission mechanism extracted by the graph neural network, a Monte Carlo simulator with abnormal profit changes is constructed, using the geometric Brownian motion equation Generate a random fluctuation path of the credit spread of the reference entity, where S represents the credit spread, dS represents the change in the credit spread within the time interval dt, the drift rate μ and the volatility σ are estimated based on historical data, dW is the increment of Brownian motion, which is a random variable used to introduce randomness, and the spread path set is obtained by adjusting the spreads of different terms in a linked manner through the Markov transition probability matrix. For the spread path set, the exponentially weighted moving average method is used to calculate the dynamic fluctuation range, and the upper and lower limits are set to two times the standard deviation to identify the credit status mutation point. The term structure imbalance index is calculated by combining the ratio of short-term spreads to long-term spreads. The liquidity fluctuation characteristics are characterized by the daily trading volume change rate to obtain a multi-dimensional risk feature sequence. A hierarchical level is constructed for the multi-dimensional risk feature sequence, and the risk level is divided according to credit rating, term structure, and liquidity level. The Gaussian kernel density estimation method is used to fit the risk loss distribution function in each level, and the kernel function bandwidth is selected by cross-validation to obtain a risk distribution function family. Based on the risk distribution function family, the generalized Pareto distribution is used to fit the tail risk, and the risk loss quantiles under the 99%, 97.5%, and 95% confidence intervals are calculated to generate loss threshold sequences with different confidence levels. A hierarchical clustering algorithm is used for the loss threshold sequence. The number of clusters is determined by the maximization criterion of inter-class distance. The risk level intervals are divided based on the cluster center to obtain the hierarchical risk warning threshold. According to the risk warning threshold, the conditional entropy method is used to calculate the contribution of each dimensional risk indicator to the warning result. The indicators with the highest contribution are selected to construct the monitoring indicator set, and the red, orange, and yellow warning level boundary values ​​are set to obtain the risk warning indicator system. The random fluctuation process of the credit default swap spread is simulated by the geometric Brownian motion equation. Taking a company's 5-year credit default swap as an example, the drift rate μ=0.02 and the volatility σ=0.15 are estimated based on historical data. The initial spread is set to 200 basis points and the time step is 1 day. 1000 random paths are generated through Monte Carlo simulation. In the linkage adjustment of adjacent term spreads, the correlation between terms is described by the Markov transfer matrix. For example, the probability of transition from 3-year to 5-year is 0.7, indicating that the change in 3-year spread will be transmitted to 5-year with a probability of 70%. In the process of risk feature identification, the dynamic fluctuation range is calculated by exponential weighted moving average method, and the data of the last 5 trading days are given weights [0.4, 0.25, 0.15, 0.12, 0.08]. When the spread on a certain day exceeds the interval of two standard deviations, it is marked as a credit mutation point. The term structure imbalance index is calculated by the ratio of short-term spread to long-term spread. For example, the 2-year spread is 180 basis points, the 5-year spread is 240 basis points, and the imbalance index is 0.75, which is lower than the normal level of 0.85. The liquidity characteristics are characterized by the daily trading volume change rate. For example, the daily trading volume drops by 30% compared with the previous day, indicating that the liquidity has deteriorated significantly. In the risk loss distribution estimation, the Gaussian kernel density function is used, and the optimal bandwidth h=0.08 is selected through cross-validation.For tail risk, the generalized Pareto distribution is used for fitting, and its distribution function is. , where the location parameter ξ=0.3, the scale parameter β=50, and the calculated risk loss quantile at the 99% confidence level is 350 basis points. In the risk level classification, the inter-class distance is calculated by the hierarchical clustering algorithm, and the difference between different risk levels is measured by the Euclidean distance. Assuming that the loss threshold sequence is [150, 200, 250, 300, 350] basis points, it is divided into three level intervals of [150, 200], [200, 300] and [300, 350] by the maximum inter-class distance criterion. The contribution of risk indicators is calculated based on the conditional entropy method. The larger the entropy value, the stronger the indicator's ability to distinguish. For example, the conditional entropy of the credit spread change rate is 0.85, the conditional entropy of the term structure imbalance index is 0.72, and the conditional entropy of the liquidity indicator is 0.65. Based on this, the top three indicators with the highest contribution are selected as core monitoring indicators. The final red warning threshold was set at 300 basis points, the orange warning threshold was set at 200 basis points, and the yellow warning threshold was set at 150 basis points, forming a complete risk warning system.

[0021] S105. Dynamically monitor key credit default swap attributes. When the credit default swap spread exceeds the preset threshold, or other attributes show significant abnormal changes, the risk warning and emergency response process will be triggered, and corresponding risk mitigation measures will be taken according to the warning level.

[0022] A monitoring indicator matrix is ​​constructed based on credit rating change data, term spread data and market transaction volume data, and the monitoring indicator matrix is ​​subjected to dimensionality reduction processing through principal component analysis, and abnormal score data is obtained by Mahalanobis distance calculation; for the abnormal score data, an exponentially weighted moving average control chart is used for abnormality inspection, and a multi-level warning threshold is set according to the weighted moving average and the weighted standard deviation to obtain graded warning data; for the graded warning data, a cumulative sum control chart is used to calculate the trend characteristic value, and the direction of indicator change is determined by the trend characteristic value. If the indicator change directions are the same, the linkage warning data is obtained; for the linkage warning data, a Gaussian kernel function support vector machine is used for classification, and a risk knowledge graph is constructed according to the classification results. The risk response rule base is queried through the risk knowledge graph to obtain a risk mitigation plan.

[0023] Specifically, a multidimensional monitoring indicator matrix is ​​constructed based on the changes in credit ratings, the fluctuation range of term spreads, and the rate of change of market trading volume. The principal component analysis method is used to reduce the dimensionality of the multidimensional indicators. The indicator anomaly score is calculated by the Mahalanobis distance, where the Mahalanobis distance uses the inverse matrix of the covariance matrix as the metric matrix to obtain real-time anomaly monitoring data. For the real-time anomaly monitoring data, an exponentially weighted moving average control chart is used for anomaly testing, and the weight coefficient is set to λ=0.2. The weighted mean and weighted standard deviation are calculated, and three times the standard deviation is set as the upper limit of the red warning, two times the standard deviation as the upper limit of the orange warning, and one times the standard deviation as the upper limit of the yellow warning to obtain the graded warning data. The trend characteristics of the graded warning data are calculated using the cumulative sum control chart, and the continuous rising point threshold and the cumulative distance threshold are set to identify the persistent abnormal trend. The linkage warning data is obtained by combining the same-direction change range of multiple indicators. The linkage warning data is classified using the Gaussian kernel function support vector machine. The kernel function parameter σ is determined by cross-validation. The classifier is trained based on historical risk event samples to obtain warning level discrimination data. According to the warning level discrimination data, a risk knowledge graph is constructed. The nodes represent the risk event types, and the edges represent the risk transmission relationships. The risk response rules are stored and queried through the graph database to obtain risk disposal data. For the risk disposal data, matching is performed based on the preset risk mitigation measures library. The corresponding mitigation plan is selected according to the abnormal degree of indicators, warning level and risk type, and the execution priority is set to obtain the risk response plan. Taking the credit default swap contract of a large enterprise as an example, the construction of multidimensional monitoring indicators includes three dimensions: the frequency of credit rating changes, the amplitude of term spread changes, and the rate of change of market trading volume. These indicators are reduced in dimension through principal component analysis. The first principal component explains 75% of the variance, indicating the comprehensive credit risk level, and the second principal component explains 15% of the variance, reflecting the characteristics of term structure changes. When calculating the Mahalanobis distance, the inverse matrix of the sample covariance matrix is ​​used as the metric matrix. When the Mahalanobis distance between the observed value and the mean exceeds 10.0, it is judged as abnormal. In the anomaly detection stage, the weight coefficient λ of the exponentially weighted moving average control chart is 0.2, which means that the weight of the most recent data is 0.2, the previous data is 0.16, and the previous data is 0.128, decreasing in sequence. Assuming that the weighted mean at a certain point in time is 250 basis points and the weighted standard deviation is 30 basis points, the upper limit of the red warning is 340 basis points, the upper limit of the orange warning is 310 basis points, and the upper limit of the yellow warning is 280 basis points. In the cumulative sum control chart, it is set to trigger a trend warning when 5 consecutive points rise and the cumulative distance exceeds 50 basis points. The linkage analysis of multiple indicators shows that when the credit rating is downgraded by one level, the 5-year spread increases by 80 basis points, and the daily trading volume decreases by 40%, this combination of characteristics of deterioration in the same direction often indicates the accelerated exposure of credit risk. In the support vector machine classifier, the Gaussian kernel function is used. , where x and y are sample vectors from the dataset and can be feature vectors of any dimension, is the Euclidean distance between x and y, exp is an exponential function, and the width parameter of the Gaussian kernel function σ=0.5 is determined by five-fold cross validation, and the classification accuracy reaches 85%. The construction of the risk knowledge graph is based on historical risk event data. The nodes include risk types such as credit rating downgrades, term structure distortions, and liquidity depletion. The weight of the edge represents the probability of risk transmission. When the risk of credit rating downgrade is identified, the relevant risk transmission path is queried through the graph database to calculate the cumulative risk exposure. At the risk disposal level, for red warning situations, such as the 5-year spread breaking through 340 basis points, the highest level of response is initiated, including limiting the scale of transactions and increasing margin requirements; for orange warnings, such as the spread rising for 5 consecutive days and accumulating more than 50 basis points, a medium-level response is taken, such as increasing the monitoring frequency and adjusting the limit; for yellow warnings, such as the continuous shrinkage of trading volume, general preventive measures are taken, such as strengthening market observation and preparing emergency funds.

[0024] S106. During the risk assessment and monitoring process, continuously collect market evaluations and extract information related to key credit default swap attributes, use the information as a supplementary input for risk transmission mechanism analysis and monitoring strategies, and dynamically optimize risk management strategies.

[0025] According to preset rules, market evaluation texts are obtained from financial websites, and named entities in the texts are extracted through the conditional random field method to obtain initial text data. For the initial text data, word frequency and inverse document frequency are used to construct text vectors, and text features are extracted through a neural network containing two convolutional layers to obtain text semantic data. The information reliability score of the text semantic data is calculated according to the rating agency level, release time interval, and historical accuracy. If the reliability score is higher than a preset threshold, a text knowledge graph is constructed through entity relationship extraction to obtain information knowledge data. For the information knowledge data, an attention mechanism is used to identify risk-related entities, and the risk transmission path is predicted through a long short-term memory network to obtain risk trend data, and the risk transmission probability is dynamically updated using the Markov decision process.

[0026] Specifically, market evaluation texts are obtained from financial news websites, rating agency announcements, and research reports according to preset information collection rules, where the collection rules include a list of issuer names, rating change keywords, and rating outlook keywords. The text is preprocessed using a bidirectional maximum matching word segmentation algorithm, and named entities in the text are extracted using a conditional random field method to obtain initial text data. For the initial text data, a text vector based on word frequency and inverse document frequency is constructed, and a word vector model with a sliding window size of 5 is used to calculate text semantic similarity. A neural network with two convolutional layers is used to extract text features, and text sentiment is classified into three categories to obtain text semantic data. The information reliability score of the text semantic data is calculated according to the rating agency level of the information source, the interval between the release time and the historical accuracy. A threshold of reliability score higher than 0.75 is set for screening, and a text knowledge graph is constructed through entity relationship extraction to obtain information knowledge data. A risk information matrix is ​​constructed based on the information knowledge data, and the attention mechanism is used to identify risk-related entities in the text. The risk transmission probability is calculated through the co-occurrence frequency and semantic distance between entities to obtain risk-related data. The early warning score standard is set for risk-related data, and the long short-term memory network is used to predict the evolution trend of the risk transmission path. The network input dimension is the same as the number of risk entities, and the number of hidden nodes is set to twice the input dimension to obtain risk trend data. According to the risk trend data, the risk response rules are constructed using the Markov decision process, and the risk transmission probability is dynamically updated based on the latest market evaluation information. The risk management plan is generated in combination with the risk warning threshold. In the text analysis of the financial market, the preset information collection rules contain a specific keyword list, such as rating change related words including "downgrade", "upgrade", "maintain", "watch list", etc., and rating outlook related words including "positive", "negative", "stable", etc. Take the credit default swap issued by a certain enterprise as an example. When the text "the company's debt scale continues to expand, cash flow pressure increases, and the rating outlook is adjusted to negative" is detected, the conditional random field method is used to identify "company" as the issuer entity and "negative" as the rating outlook entity. In the process of constructing the text vector, the word frequency-inverse document frequency (TF-IDF) method is used to calculate the word weight. Assume that "rating downgrade" appears 3 times in a document, the total number of words in the document is 100, and 10 documents out of 1000 contain this word, then its TF value is 0.03, IDF value is log(1000 / 10)=2, and the final weight is 0.06. Text features are extracted through a two-layer convolutional neural network. The first layer uses 64 3×3 convolution kernels, and the second layer uses 32 3×3 convolution kernels to classify the text into three categories: positive, neutral, and negative. In the evaluation of information reliability, different weights are given to information from different sources, such as the weight of information released by rating agencies is 0.9, the weight of research institute reports is 0.7, and the weight of news reports is 0.5.In terms of timeliness, the weight of the information released on the day is 1.0, and it decays by 0.05 every day. The historical accuracy is calculated by the accuracy of the previous prediction. For example, if the prediction accuracy of a certain information source in the past was 85%, it is given a weight of 0.85. The weighted average of the three dimensions gives the final information reliability score. In the risk information matrix, the correlation strength between entities is calculated by co-occurrence frequency and semantic distance. For example, "real estate" and "liquidity risk" appear together 50 times in the text, and each appears 100 times and 80 times respectively, then the co-occurrence frequency is 50 / √(100×80)=0.56. The risk transmission trend is predicted by the long short-term memory network, with the input dimension of 20 risk entities, 40 hidden nodes, and a prediction window of 5 trading days. In the construction of risk response rules, the Markov decision process is adopted. The state space contains four states: normal, attention, warning, and danger. The corresponding risk management measures are selected according to the current state and market evaluation information. For example, when the market evaluation shows a continuous negative trend and the negative information reliability score exceeds 0.8, the trigger status will be transferred from "attention" to "warning", and the risk management plan will be adjusted accordingly.

[0027] S107. Generate credit risk monitoring reports regularly, summarize risk assessment results, early warning situations and response measures, as well as the cumulative loss distribution under different risk contagion scenarios, and use data visualization methods to visualize credit risk monitoring content, including time series graphs of changes in credit default swap attributes, credit risk transmission network graphs, and probability density graphs of cumulative loss distribution.

[0028] The weighted average method is used to calculate the sliding mean of the indicator based on the credit status data, term structure data and liquidity data to obtain a risk indicator sequence; a rolling time window is used to perform segmented statistical calculations on the risk indicator sequence, and the indicator characteristics are grouped by a hierarchical clustering method to obtain a risk feature sequence; a Gaussian kernel function is used to perform kernel density estimation fitting on the risk feature sequence, and cumulative loss sample data is generated by Latin hypercube sampling to obtain risk probability data; a conduction network diagram is constructed for the risk probability data, the gravity coefficients between nodes are set, and the node positions are determined by iterative optimization to obtain network layout data.

[0029] Specifically, a monitoring indicator matrix is ​​created based on the credit status change data, term structure data, and liquidity change data. A multidimensional array is used to store the early warning records and disposal measures of different risk types. The sliding mean and change rate of each indicator are calculated by the weighted average method to obtain the risk indicator sequence. The risk indicator sequence is segmented and statistically analyzed using a rolling time window. The mean, variance, and skewness of each indicator on different time scales are calculated. The indicator characteristics are grouped by the hierarchical clustering method to obtain the risk feature sequence. For the risk feature sequence, the Gaussian kernel function is used to perform kernel density estimation to fit the loss distribution under different risk scenarios. The cumulative loss samples are generated by Latin hypercube sampling. The risk level is divided by setting 90%, 95%, and 99% confidence intervals to obtain the risk probability data. Based on the risk probability data, a transmission network diagram is constructed. The attraction coefficient and repulsion coefficient between nodes are set by the stress-based force-directed layout algorithm. The node positions are determined by iterative optimization to obtain the network layout data. The network layout data is visualized, and a time series line chart is drawn by the spline interpolation method. The loss distribution is displayed by the kernel density curve. The risk transmission intensity is displayed based on the heat map to obtain the chart component data. The monitoring report template is constructed according to the chart component data, and the convolutional neural network is used to optimize the spatial layout of the chart. The responsive grid system is used to adaptively layout the screens of different sizes to obtain the monitoring report content. In the process of generating the risk monitoring report, the monitoring indicator matrix is ​​first constructed. Taking a certain enterprise credit default swap as an example, its credit status indicators include credit rating, default distance, and debt-to-asset ratio. The term structure indicators include short-term spread, long-term spread, and term spread ratio. The liquidity indicators include daily trading volume, bid-ask spread, and trading frequency. A 30-day sliding window is used to calculate the mean and change rate of each indicator, and the weight is set according to the importance of the indicator, such as a credit rating weight of 0.4, a default distance weight of 0.3, and a debt-to-asset ratio weight of 0.3. When extracting risk features, multiple time scales are used for each indicator to calculate statistics, such as using 5 days, 10 days, and 20 days as windows to calculate statistics. Taking the credit spread as an example, its 5-day mean is 280 basis points, the variance is 45, and the skewness is 0.8, indicating that it is right-skewed in the short term. The hierarchical clustering method is used to cluster similar indicator features into a group. The clustering distance uses the Euclidean distance and the connection method uses the average connection. In the loss distribution fitting, the Gaussian kernel function is used. , the bandwidth parameter σ is selected by cross-validation. 10,000 sample points are generated by Latin hypercube sampling, and the risk values ​​under different confidence intervals are calculated, such as 350 basis points for the loss value under 95% confidence interval and 420 basis points for the loss value under 99% confidence interval. The visualization of the risk transmission network adopts the force-directed layout algorithm, with the attraction coefficient between nodes set to 0.3 and the repulsion coefficient set to 0.7. The node positions are optimized through 1,000 iterations. The size of the node indicates the risk intensity, the thickness of the edge indicates the transmission probability, and the depth of the color indicates the transmission speed. For the visualization of time series data, cubic spline interpolation is used to ensure the smoothness of the curve, and the interpolation node spacing is 1 trading day. In the layout of the monitoring report, an 8-layer convolutional neural network is used to optimize the chart layout. The input is the position and size information of the chart, and the output is the optimized layout parameters. The responsive grid system adopts a 12-column layout, and automatically adjusts the number of columns under different screen sizes, such as maintaining 12 columns on a large screen, reducing to 8 columns on a medium screen, and reducing to 4 columns on a small screen to maintain the display effect of the chart. All charts use a unified color scheme, with red representing high risk, yellow representing medium risk, and green representing low risk. The gradient of colors reflects the continuous change of risk.

[0030] 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 dynamic monitoring and report generation of financial risks based on multi-source data, characterized in that: The method comprises: Obtain multi-source data on credit default swaps, pre-process and extract features for interest rate spread data of different maturities, and obtain term structure feature vectors; obtain long-term and short-term interest rate spread curves based on real-time long-term and short-term interest rate spread data, perform anomaly detection on the shape changes of long-term and short-term interest rate spread curves, and trigger the evaluation process of credit risk contagion effect when an abnormal distortion of the interest rate spread curve is detected; when triggering the evaluation process of credit risk contagion effect, use the standardized term structure feature vector as the node attribute of the graph, and the abnormal distortion of the interest rate spread curve as the edge attribute, and learn the transmission path and intensity of credit risk between different maturities and market entities through graph neural networks to analyze the credit risk transmission mechanism based on graph neural networks; use the Monte Carlo simulation method to analyze the credit risk transmission mechanism learned by the graph neural network, generate risk contagion scenarios driven by abnormal changes in credit default swap interest rates of different maturities, and evaluate the cumulative losses of credit risk under different scenarios. Distribution, determine the thresholds and warning levels of risk monitoring indicators with key credit default swap attributes as the core; dynamically monitor key credit default swap attributes, and when the credit default swap spread exceeds the preset threshold, or other attributes show significant abnormal changes, trigger the risk warning and emergency response process, and take corresponding risk mitigation measures according to the warning level; in the process of risk assessment and monitoring, continuously collect market evaluations, extract information related to key credit default swap attributes, and use the information as a supplementary input for risk transmission mechanism analysis and monitoring strategies, and dynamically optimize risk management strategies; regularly generate credit risk monitoring reports, summarize risk assessment results, warning situations and response measures, as well as the cumulative loss distribution under different risk contagion scenarios, and use data visualization methods to visualize credit risk monitoring content, including time series graphs of changes in credit default swap attributes, credit risk transmission network graphs, and probability density graphs of cumulative loss distribution.

2. The method according to claim 1, characterized in that The method of obtaining multi-source data of credit default swaps, preprocessing and feature extraction of interest rate spread data of different maturities, and obtaining a term structure feature vector includes: Obtain interest rate spread data from the database according to the time interval, and obtain the initial interest rate spread data set through date alignment method and uniform decimal processing; A linear interpolation method is used to fill the detected null value points for the initial interest rate spread data set, and duplicate data items are removed by distribution characteristics to obtain a cleaned interest rate spread data set; The cleaned interest rate spread data set is grouped by term, and outliers are corrected using a moving average method. Data points exceeding a threshold range are eliminated using a percentile method to obtain smoothed interest rate spread data. The smoothed interest rate spread data is normalized using the minimum and maximum standardization method, and the term structure characteristic vector is obtained by fitting and calculating the support vector regression model.

3. The method according to claim 1, characterized in that The long-term and short-term interest rate spread curve is obtained based on the long-term and short-term interest rate spread data obtained in real time, and abnormal detection is performed on the shape change of the long-term and short-term interest rate spread curve. When abnormal distortion of the interest rate spread curve is detected, the credit risk contagion effect assessment process is triggered, including: Obtain long-term interest rate spread data and short-term interest rate spread data from the market database according to the preset sampling frequency, and obtain the basic interest rate spread data set through data standardization and timestamp alignment; For the basic interest rate spread data set, cubic spline interpolation method and least square method are used to determine the spline function coefficients to obtain continuous and smooth interest rate spread curve data; Calculate the interest rate change rate between adjacent term points for the interest rate spread curve data, and obtain the curve abnormality degree data by using the sliding window method and the autoregressive moving average method; If the curve abnormality data exceeds the warning threshold, the risk transmission probability is calculated, and the risk transmission matrix is ​​established through the risk transmission probability to obtain the term structure risk transmission coefficient.

4. The method according to claim 1, characterized in that: In the evaluation process of triggering the credit risk contagion effect, the standardized term structure feature vector is used as the node attribute of the graph, and the spread curve distortion anomaly is used as the edge attribute. The transmission path and intensity of credit risk between different terms and market entities are learned through the graph neural network to analyze the credit risk transmission mechanism based on the graph neural network, including: The node attribute matrix is ​​constructed according to the standardized term structure feature vector, the edge attribute matrix is ​​constructed by using the spread curve distortion data, and the initial transmission graph data is obtained by using the minimum and maximum normalization method; The weighted Euclidean distance method is used to calculate the similarity score between nodes for the initial conduction map data. If the similarity score is lower than the median of the similarity score of the node pair, the node connection is pruned to obtain the optimized conduction map data. A graph convolution operator is designed through the Chebyshev polynomial expansion form, a feature extractor is constructed for the nodes in the optimized conduction graph data, and weighted features of adjacent nodes are calculated to obtain local structural features; The dot product attention calculation method is used to obtain the attention weight matrix between local structural features, and the node features are weighted aggregated based on the attention weights to obtain the global conduction features.

5. The method according to claim 1, characterized in that The Monte Carlo simulation method is used to analyze the credit risk transmission mechanism learned by the graph neural network, generate risk contagion scenarios driven by abnormal changes in credit default swap interest rates of different maturities, evaluate the cumulative loss distribution of credit risk under different scenarios, and determine the risk monitoring indicator thresholds and warning levels with key credit default swap attributes as the core, including: Calculating drift rate and volatility parameters based on historical credit spread data, and using a geometric Brownian motion equation to generate a set of random fluctuation paths of the credit spread, wherein the geometric Brownian motion equation includes a drift term and a disturbance term; For the random fluctuation path set, the dynamic fluctuation interval value is calculated by using the exponentially weighted moving average method, and the credit status mutation point is identified by the ratio of the dynamic fluctuation interval value to the standard deviation to obtain a multi-dimensional risk feature sequence; For the multidimensional risk feature sequence, a Gaussian kernel density estimation method is used to fit the risk loss distribution function, and the bandwidth parameter of the Gaussian kernel density estimation method is obtained by a cross-validation method; For the risk loss distribution function, the generalized Pareto distribution is used to fit the tail risk characteristics, the risk loss quantile value under the confidence interval is obtained, and the risk warning threshold is determined by the hierarchical clustering method.

6. The method according to claim 1, characterized in that The key credit default swap attributes are dynamically monitored. When the credit default swap spread exceeds the preset threshold, or other attributes show significant abnormal changes, the risk warning and emergency response process is triggered, and corresponding risk mitigation measures are taken according to the warning level, including: Constructing a monitoring indicator matrix based on credit rating change data, term spread data, and market transaction volume data, performing dimensionality reduction processing on the monitoring indicator matrix through principal component analysis, and obtaining abnormal score data by using Mahalanobis distance calculation; For the abnormal score data, an exponentially weighted moving average control chart is used to perform an abnormality test, and a multi-level warning threshold is set according to the weighted moving average and the weighted standard deviation to obtain graded warning data; For the graded warning data, a trend characteristic value is calculated using a cumulative sum control chart, and the direction of indicator change is determined by the trend characteristic value. If the indicator change directions are the same, linkage warning data is obtained; The linkage warning data is classified by using a Gaussian kernel support vector machine, and a risk knowledge graph is constructed according to the classification results. The risk response rule base is queried through the risk knowledge graph to obtain a risk mitigation plan.

7. The method according to claim 1, characterized in that In the risk assessment and monitoring process, the market evaluation is continuously collected, information related to key credit default swap attributes is extracted, and the information is used as a supplementary input for risk transmission mechanism analysis and monitoring strategy to dynamically optimize risk management strategy, including: Obtain market evaluation text from a financial website according to preset rules, extract named entities in the text using a conditional random field method, and obtain initial text data; For the initial text data, a text vector is constructed using word frequency and inverse document frequency, and text features are extracted through a neural network containing two convolutional layers to obtain text semantic data; Calculating the information reliability score of the text semantic data according to the rating agency level, release time interval and historical accuracy; if the reliability score is higher than a preset threshold, constructing a text knowledge graph through entity relationship extraction to obtain information knowledge data; For the information knowledge data, the attention mechanism is used to identify risk-related entities, the risk transmission path is predicted through the long short-term memory network, the risk trend data is obtained, and the Markov decision process is used to dynamically update the risk transmission probability.

8. The method according to claim 1, characterized in that The credit risk monitoring report is generated regularly, summarizing the risk assessment results, early warning situations and response measures, as well as the cumulative loss distribution under different risk contagion scenarios, and using data visualization methods to visualize the credit risk monitoring content, including a time series diagram of changes in credit default swap attributes, a credit risk transmission network diagram, and a probability density diagram of the cumulative loss distribution, including: The sliding mean of the indicator is calculated using the weighted average method based on the credit status data, term structure data and liquidity data to obtain the risk indicator series; A rolling time window is used to perform segmented statistical calculations on the risk indicator sequence, and the indicator features are grouped by a hierarchical clustering method to obtain a risk feature sequence; The risk characteristic sequence is fitted with a Gaussian kernel function for kernel density estimation, and cumulative loss sample data is generated through Latin hypercube sampling to obtain risk probability data; A conduction network diagram is constructed for the risk probability data, the gravitational coefficients between nodes are set, the node positions are determined through iterative optimization, and the network layout data is obtained.

Images