Enterprise risk monitoring method and device, equipment, storage medium and program product

Abstract

The invention provides an enterprise risk monitoring method and device, equipment and a storage medium, and can be applied to the technical field of financial science and technology. The method comprises the following steps: calculating the risk sensitivity of each basic feature in a basic feature set, and screening in the basic feature set based on a plurality of risk sensitivities to obtain a plurality of candidate basic features; a weighted undirected graph is formed based on the incidence relation between every two candidate basic features, and the risk conduction intensity between each node and the adjacent node in the weighted undirected graph is calculated based on the data distribution state indicated by the multiple candidate basic features; obtaining a plurality of passive influence indexes based on the plurality of partial correlation coefficients and the plurality of risk conduction intensities indicated by the weighted undirected graph; fusing the risk sensitivity and the passive influence index corresponding to the plurality of candidate basic features to obtain a plurality of adaptive weights so as to update the corresponding candidate basic features and obtain a plurality of target features; and processing the plurality of target features through a risk monitoring model to obtain risk monitoring data.

Description

TECHNICAL FIELD

[0001] The present disclosure relates to the field of financial technology, and particularly relates to a business risk monitoring method and device, equipment, storage medium and program product. BACKGROUND

[0002] The risk monitoring of small and micro enterprises is a systematic method established by financial institutions to provide financial services to small and micro enterprises, in order to identify various risks brought by the characteristics and experience environment of small and micro enterprises, so as to ensure the safety of their own funds and realize sustainable operation.

[0003] At present, the risk monitoring is mainly performed by analyzing enterprise data through machine learning algorithms. However, the enterprise data is of various types and large in quantity, the influence degree of different types of data on enterprise risk is different, and there may be mutual influence between different types of data. The traditional algorithms do not consider the above data characteristics, resulting in that the accuracy of enterprise risk monitoring needs to be improved. SUMMARY

[0004] In view of the above problems, the present disclosure provides a business risk monitoring method, device, equipment, storage medium and program product.

[0005] According to a first aspect of the present disclosure, a business risk monitoring method is provided, which comprises: in response to receiving a basic feature set of an enterprise, calculating the risk sensitivity of each basic feature in the basic feature set, screening the basic feature set based on the obtained multiple risk sensitivities to obtain multiple candidate basic features, and the basic feature set indicating the comprehensive operation state of the enterprise; based on the correlation relationship between the multiple candidate basic features, a weighted undirected graph is formed, and based on the data distribution state indicated by the multiple candidate basic features, the risk transmission intensity between each node and adjacent node in the weighted undirected graph is iteratively calculated; based on the multiple partial correlation coefficients indicated by the weighted undirected graph and the obtained multiple risk transmission intensities, multiple passive influence indexes are obtained, and the passive influence index represents the cumulative degree of the corresponding node affected by the risk transmission of the adjacent node; the risk sensitivity and passive influence index corresponding to the multiple candidate basic features are fused to obtain multiple adaptive weights, so as to update the corresponding candidate basic features to obtain multiple target features; the multiple target features are processed through a pre-trained risk monitoring model to obtain risk monitoring data, so as to perform enterprise risk early warning according to the risk monitoring data.

[0006] According to an embodiment of the present disclosure, the risk sensitivity of each basic feature in the basic feature set is calculated by: obtaining a corresponding industry relative importance score based on the regulatory importance, industry use frequency and expert importance score of the basic feature, the industry relative importance score representing the importance of the corresponding basic feature in enterprise risk monitoring; obtaining a feature abnormality degree based on the historical abnormality degree, industry abnormality degree and trend abnormality degree of the basic feature, the feature abnormality degree representing the deviation degree of the corresponding basic feature; obtaining a feature volatility degree based on the ratio of the standard deviation and the mean value of the basic feature within a preset time period, the feature volatility degree representing the volatility of the basic feature in the time dimension; obtaining a trend correlation degree based on the synchronization degree of the change trend between the basic feature and other basic features; and fusing the industry relative importance score, the feature abnormality degree, the feature volatility degree and the trend correlation degree to obtain the risk sensitivity.

[0007] According to an embodiment of the present disclosure, the risk transmission strength between each node and adjacent nodes in the weighted undirected graph is iteratively calculated based on the data distribution state indicated by the plurality of candidate basic features, comprising: calculating a corresponding quantile based on the candidate basic feature corresponding to the node and the candidate basic feature corresponding to the adjacent node, the quantile indicating an abnormality benchmark of the data distribution of the corresponding candidate basic feature; calculating the conditional probability and the conditional expectation between each node and adjacent nodes in the weighted undirected graph based on the candidate basic feature corresponding to the node, the candidate basic feature corresponding to the adjacent node and the obtained plurality of quantiles, the conditional probability indicating the possibility of the simultaneous existence of abnormalities of the corresponding node and the adjacent node, and the conditional expectation indicating the abnormality depth of the corresponding node; and fusing the conditional probability and the conditional expectation to obtain the risk transmission strength.

[0008] According to an embodiment of the present disclosure, the weighted undirected graph is formed based on the association relationship between each two of the plurality of candidate basic features, comprising: calculating a partial correlation coefficient between each two of the plurality of candidate basic features through linear regression residual; and constructing the weighted undirected graph by taking the plurality of candidate basic features as nodes and taking the obtained plurality of partial correlation coefficients as edge weights, the partial correlation coefficient representing the association strength between the candidate basic features.

[0009] According to an embodiment of the present disclosure, the partial correlation coefficient between each two of the plurality of candidate basic features is calculated through linear regression residual, comprising: constructing a plurality of linear regression models by taking any one of the two candidate basic features as the dependent variable and the remaining candidate basic features as the independent variables; calculating the residual of each of the plurality of linear regression models, and obtaining the partial correlation coefficient based on the obtained plurality of residuals, the residual indicating the effective information of the corresponding dependent variable excluding the influence of other candidate basic features.

[0010] According to an embodiment of the present disclosure, for each node in the weighted undirected graph, the passive influence index is obtained in the following manner: based on the partial correlation coefficient between the node and the corresponding neighbor node, the risk transmission strength, and the eigenvector centrality of the corresponding neighbor node, the passive influence index is obtained, and the eigenvector centrality represents the global importance of the corresponding neighbor node in the weighted undirected graph.

[0011] According to an embodiment of the present disclosure, the eigenvector centrality is obtained in the following manner: based on the plurality of partial correlation coefficients indicated by the weighted undirected graph, an adjacency matrix is formed; and based on the maximum eigenvalue in the adjacency matrix and the neighbor matrix, the eigenvector centrality is obtained.

[0012] According to an embodiment of the present disclosure, the basic features are operating features, account behavior features, credit features, or production behavior features.

[0013] A second aspect of the present disclosure provides an enterprise risk monitoring apparatus, comprising: a screening module configured to, in response to receiving a set of basic features of an enterprise, calculate a risk sensitivity of each basic feature in the set of basic features, screen the set of basic features based on the obtained plurality of risk sensitivities to obtain a plurality of candidate basic features, and the set of basic features indicating a comprehensive operation state of the enterprise; a risk transmission module configured to form a weighted undirected graph based on an association relationship between each two of the plurality of candidate basic features, and iteratively calculate a risk transmission strength between each node and a neighboring node in the weighted undirected graph based on a data distribution state indicated by the plurality of candidate basic features; a calculation module configured to obtain a plurality of passive influence indexes based on a plurality of partial correlation coefficients indicated by the weighted undirected graph and the plurality of risk transmission strengths obtained, and the passive influence indexes representing a cumulative degree of risk transmission of the corresponding node by the neighboring node; an updating module configured to fuse the risk sensitivity and the passive influence index corresponding to the plurality of candidate basic features to obtain a plurality of adaptive weights, update the corresponding candidate basic features to obtain a plurality of target features; and a risk monitoring module configured to process the plurality of target features through a pre-trained risk monitoring model to obtain risk monitoring data, and perform enterprise risk early warning according to the risk monitoring data.

[0014] A third aspect of the present disclosure provides an electronic device, comprising: one or more processors; a memory configured to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors perform the above method.

[0015] A fourth aspect of the present disclosure further provides a computer-readable storage medium having stored executable instructions, which, when executed by a processor, cause the processor to perform the above method.

[0016] The fifth aspect of the present disclosure also provides a computer program product comprising a computer program which, when executed by a processor, implements the above method.

[0017] The enterprise risk monitoring method, device, equipment, storage medium and program product provided by the present disclosure can preliminarily screen the basic features capable of sensitively reflecting the enterprise operation risk by calculating the risk sensitivity in the basic feature set; meanwhile, the risk transmission between the candidate basic features is determined by using the data distribution state of the candidate basic features, the correlation between the candidate basic features, etc., and the plurality of target features used for risk monitoring are updated in combination with the risk transmission and correlation between the candidate basic features; therefore, at least part of the problem that the risk monitoring accuracy needs to be improved due to not considering the data features of small and micro enterprises is solved, and the technical effect of improving the risk monitoring accuracy of small and micro enterprises is achieved. BRIEF DESCRIPTION OF DRAWINGS

[0018] The above and other objects, features and advantages of the present disclosure will become more apparent from the following description of embodiments of the present disclosure taken in conjunction with the accompanying drawings, in which:

[0019] Figure 1 An application scenario diagram of the enterprise risk monitoring method, device, equipment, medium and program product according to the embodiments of the present disclosure is schematically shown;

[0020] Figure 2 A flowchart of the enterprise risk monitoring method according to the embodiments of the present disclosure is schematically shown;

[0021] Figure 3 A flowchart of calculating the risk sensitivity according to the embodiments of the present disclosure is schematically shown;

[0022] Figure 4 A structural block diagram of the enterprise risk monitoring device according to the embodiments of the present disclosure is schematically shown; and

[0023] Figure 5 A block diagram of an electronic device suitable for implementing the enterprise risk monitoring method according to the embodiments of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0024] Hereinafter, embodiments of the present disclosure will be described with reference to the accompanying drawings. However, it should be understood that these descriptions are merely exemplary and are not intended to limit the scope of the present disclosure. In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the embodiments of the present disclosure. However, it will be apparent to one skilled in the art that one or more embodiments can be practiced without these specific details. In addition, in the following description, descriptions of well-known structures and techniques have been omitted to avoid unnecessarily obscuring the concept of the present disclosure.

[0025] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. As used herein, the terms "comprises", "comprising", "includes", "including" and the like are specifically intended to be open-ended and to mean that other features, steps, operations, and / or components can be added.

[0026] All terms used herein including technical and scientific terms have the meanings commonly understood by one of ordinary skill in the art unless otherwise defined. It should be noted that the terms used herein are defined as having meanings that are consistent with the context of the specification in which they are used and should not be interpreted in an overly idealized or overly formal way.

[0027] In the case of using expressions similar to "at least one of A, B, and C, etc.", it should be generally interpreted that the meaning of the expression is the same as that of "one or more of A, B, and C" (for example, "a system having at least one of A, B, and C" should include but not be limited to a system having A alone, a system having B alone, a system having C alone, a system having both A and B, a system having both A and C, a system having both B and C, and / or a system having A, B, and C, etc.).

[0028] In the technical solutions of the disclosure, the user information (including but not limited to user personal information, user image information, user equipment information, such as location information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved are information and data authorized by the user or authorized by all parties, and the collection, storage, use, processing, transmission, provision, disclosure and application of related data comply with relevant laws, regulations and standards, do not violate public order and good customs, and provide corresponding operation portal for user to choose authorization or refusal.

[0029] It should be noted that in the embodiments of the disclosure, some software, components, models, etc. may be mentioned, which should be considered as exemplary, and the purpose is only to illustrate the feasibility of the implementation of the technical solutions of the disclosure, but it does not mean that the applicant has or will necessarily use the scheme.

[0030] The embodiments of the disclosure provide an enterprise risk monitoring method, which calculates the sensitivity of each basic feature to enterprise risk, and screens features sensitive to enterprise risk for enterprise risk monitoring; at the same time, considering the correlation between candidate basic features and risk transmission, the passive influence between candidate basic features is quantified; the target feature is updated by using risk sensitivity and passive influence index, so that the data used for enterprise risk monitoring can not only accurately describe the current risk situation of the enterprise, but also reflect the mutual influence between features, and improve the accuracy and comprehensiveness of enterprise risk monitoring.

[0031] Figure 1An application scenario diagram of the enterprise risk monitoring method according to an embodiment of the present disclosure is shown schematically.

[0032] As shown in the figure, Figure 1 The application scenario 100 according to this embodiment can include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is a medium for providing a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 can include various connection types, such as wired, wireless communication links, or optical fiber cables, etc.

[0033] A user can use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications can be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as shopping applications, web browser applications, search applications, instant messaging tools, email clients, social platform software, etc. (only as examples).

[0034] The first terminal device 101, the second terminal device 102, and the third terminal device 103 can be various electronic devices with display screens and supporting web browsing, including but not limited to smartphones, tablet computers, laptop computers, desktop computers, etc.

[0035] The server 105 can be a server providing various services, such as a background management server supporting a website browsed by a user using the first terminal device 101, the second terminal device 102, and the third terminal device 103 (only as an example). The background management server can analyze and process received user requests and other data, and feed back the processing results (such as web pages, information, or data generated according to user requests, etc.) to the terminal device.

[0036] It should be noted that the enterprise risk monitoring method provided by the embodiments of the present disclosure can generally be executed by the server 105. Accordingly, the enterprise risk monitoring apparatus provided by the embodiments of the present disclosure can generally be arranged in the server 105. The enterprise risk monitoring method provided by the embodiments of the present disclosure can also be executed by a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105. Accordingly, the enterprise risk monitoring apparatus provided by the embodiments of the present disclosure can also be arranged in a server or a server cluster different from the server 105 and capable of communicating with the first terminal device 101, the second terminal device 102, the third terminal device 103, and / or the server 105.

[0037] It should be understood that Figure 1 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0038] The following will be based on Figure 1 The described scene, through Figures 2-3 The enterprise risk monitoring method of the disclosed embodiments is described in detail.

[0039] Figure 2 A flowchart illustrating an enterprise risk monitoring method according to an embodiment of this disclosure is shown schematically.

[0040] like Figure 2 As shown, the enterprise risk monitoring method in this embodiment includes operations S210 to S250.

[0041] In operation S210, in response to receiving the basic feature set of the enterprise, the risk sensitivity of each basic feature in the basic feature set is calculated, and multiple candidate basic features are obtained by filtering based on the obtained multiple risk sensitivities in the basic feature set.

[0042] According to embodiments of this disclosure, the set of basic features indicates the overall operational status of an enterprise and may include data reflecting the operational status of the enterprise from dimensions such as business operations, account behavior, credit, and production behavior; risk sensitivity characterizes the degree to which the corresponding basic features are sensitive to enterprise risks, and basic features with high risk sensitivity are more likely to reflect changes in enterprise risks.

[0043] For example, the industry relative importance score, feature anomaly, feature volatility and trend correlation indicated by basic features are integrated to obtain the corresponding risk sensitivity; the obtained risk sensitivity is grouped according to different dimensions and sorted in descending order, and the basic features ranked at the top in each group are selected as candidate basic features.

[0044] In operation S220, a weighted undirected graph is formed based on the pairwise relationships between multiple candidate basic features. Based on the data distribution state indicated by multiple candidate basic features, the risk transmission strength between each node and its neighboring nodes in the weighted undirected graph is iteratively calculated.

[0045] According to embodiments of this disclosure, a weighted undirected graph uses a graph structure to display the direct and indirect relationships between multiple candidate basic features, and the data distribution state characterizes the numerical distribution of the corresponding candidate basic features.

[0046] For example, a weighted undirected graph is constructed with candidate basic features as nodes and the relationships between pairs of candidate basic features as edges; the conditional probability and conditional expectation between the candidate basic features corresponding to each node and the candidate basic features corresponding to adjacent nodes are calculated, and the conditional probability and conditional expectation are fused to obtain the risk transmission strength.

[0047] In operation S230, multiple passive impact indices are obtained based on multiple partial correlation coefficients indicated by weighted undirected graphs and multiple risk transmission strengths.

[0048] According to embodiments of this disclosure, the partial correlation coefficient characterizes the degree of direct association between each node and its neighboring nodes, while the passive influence index characterizes the cumulative degree to which the corresponding node is affected by the risk transmission from its neighboring nodes.

[0049] For example, multiple partial correlation coefficients and multiple risk transmission intensities of each node corresponding to all its neighboring nodes are fused to obtain multiple passive impact indices.

[0050] In operation S240, the risk sensitivity and passive impact index corresponding to multiple candidate basic features are fused to obtain multiple adaptive weights, which are then used to update the corresponding candidate basic features to obtain multiple target features.

[0051] For example, the risk sensitivity and passive impact index are weighted and summed to obtain the corresponding adaptive weights, and the product of the adaptive weights and the corresponding candidate basic features is used as the corresponding target features.

[0052] When operating the S250, multiple target features are processed through a pre-trained risk monitoring model to obtain risk monitoring data, which is then used to provide enterprise risk warnings.

[0053] For example, multiple target features are concatenated, and the concatenated target features are input into a pre-trained risk monitoring model for processing to obtain risk monitoring data. The risk monitoring model can be a long short-term memory network, a bidirectional long short-term memory network, a decision tree algorithm, a lightweight gradient boosting machine algorithm, etc., and the risk monitoring data can be the risk level.

[0054] When the risk level is low, a risk warning will be issued to the user; when the risk level is medium, a higher risk warning will be issued to the user; and when the risk level is high, a high risk warning will be issued to the user.

[0055] Through the above operations S210 to S250, the risk sensitivity is calculated to perform preliminary screening in the basic feature set, which reduces the amount of data processing and provides data that can sensitively reflect the enterprise's operational risks for subsequent calculations. By utilizing the data distribution status of candidate basic features, the correlation between candidate basic features, and weighted undirected graphs, the degree of risk transmission between candidate basic features is quantified. The target features that comprehensively consider risk sensitivity and risk transmission are used for risk monitoring, which improves the comprehensiveness and accuracy of enterprise risk monitoring.

[0056] Figure 3 A flowchart illustrating the calculation of risk sensitivity according to an embodiment of this disclosure is shown schematically.

[0057] Risk sensitivity can measure the sensitivity and importance of a basic feature in reflecting a company's potential risks. In order to accurately select the candidate basic features with the most risk indication significance from multiple basic features provided by the company, a comprehensive calculation is performed based on the abnormal patterns of the company's own data, industry comparisons, and the relationships between features to characterize the risk sensitivity of features from different dimensions. In some embodiments, calculating the risk sensitivity of each basic feature in the set of basic features includes operations S310 to S350.

[0058] When operating S310, based on the regulatory importance of basic features, industry usage frequency, and expert importance scores, the corresponding industry relative importance score is obtained.

[0059] According to embodiments of this disclosure, regulatory importance indicates the frequency with which a corresponding fundamental feature is mentioned and emphasized in financial institution regulatory documents; industry usage frequency represents the prevalence of its adoption and use in external review reports and credit review templates of similar enterprises; expert importance score refers to an empirical score based on domain experts (such as credit reviewers and industry analysts); and industry relative importance score represents the importance of the corresponding fundamental feature in enterprise risk monitoring. A feature that receives widespread attention from regulators, peers, and experts has a stronger indicative significance for risk.

[0060] For example, the industry relative importance score I of the i-th basic feature i Represented as:

[0061] ;

[0062] In the formula, Reg i Represents the regulatory importance of the i-th basic feature, Reg i ∈[0,1], Freq i Represents the industry usage frequency of the i-th basic feature, Freq i ∈[0,1],exp iRepresents the expert importance score of the i-th basic feature, exp i ∈[0,1], a, b, c represent harmonic weights, a+b+c=1.

[0063] Here, regulatory importance, industry usage frequency, and expert importance scores are pre-stored in the feature knowledge base, with each basic feature having its own matching regulatory importance, industry usage frequency, and expert importance score.

[0064] When operating S320, the corresponding feature anomalies are obtained based on the historical anomalies, industry anomalies, and trend anomalies of the basic features.

[0065] According to embodiments of this disclosure, historical anomaly and trend anomaly represent the degree of deviation of the corresponding basic feature from its own historical pattern, industry anomaly represents the degree of deviation of the corresponding basic feature from the industry norm, and feature anomaly represents the degree of deviation of the corresponding basic feature.

[0066] For each basic feature, the historical anomaly is further obtained based on the current observed value of the basic feature and its mean and standard deviation within a historical time window; the industry anomaly is obtained based on the current observed value of the basic feature and the median and median absolute deviation of the basic feature values ​​among companies of the same size and industry; the trend anomaly is obtained based on the linear regression slope of the basic feature within a preset short-term time window and a preset long-term time window; and the maximum value among the historical anomaly, industry anomaly, and trend anomaly is taken as the feature anomaly.

[0067] For example, the feature anomaly degree A of the i-th basic feature i Represented as:

[0068] A i =max(A 1i A 2i A 3i );

[0069] The historical anomaly degree A of the i-th basic feature 1i Represented as:

[0070] A 1i =|(X i -μ i ) / σ i ;

[0071] In the formula, X i μ represents the current observation value of the i-th basic feature. i Let σ represent the mean of the i-th basic feature over a historical time window (e.g., the past 24 months). i This represents the standard deviation of the i-th basic feature within the same historical time window;

[0072] The industry anomaly degree A of the i-th basic feature 2i Represented as:

[0073] A 2i =|(X i -median i ) / MAD i |;

[0074] In the formula, median i MAD represents the median of basic characteristic i for companies of the same industry and size. i This represents the median absolute deviation of basic feature i among companies of the same industry and size.

[0075] The trend anomaly A of the i-th basic feature 3i Represented as:

[0076] A 3i =|Trend 1i –Trend 2i | / (|Trend 2i |+ε);

[0077] In the formula, Trend 1i Trend represents the linear regression slope of the i-th basic feature over a short time window (e.g., the last 6 months). 2i ε represents the linear regression slope of the i-th basic feature over a long-term time window (e.g., the past 24 months), and ε represents a very small constant used to prevent division by zero errors.

[0078] In operation S330, the characteristic volatility is obtained based on the ratio of the standard deviation to the mean of the basic characteristic within a preset time period.

[0079] According to embodiments of this disclosure, the characteristic volatility characterizes the volatility of a basic characteristic over time and is used to measure the recent abnormal increase in volatility of the corresponding basic characteristic. A sudden increase in volatility (regardless of direction) often indicates an increase in uncertainty in the operating environment or a weakening of internal control.

[0080] For example, the characteristic volatility V of the i-th basic feature i Represented as:

[0081] ;

[0082] In the formula, σ i_r μ represents the standard deviation of the i-th basic feature over a recent time window (e.g., the past 3-6 months). i_r Let represent the mean of the i-th basic feature over a recent time window.

[0083] In operation S340, the trend correlation is obtained based on the degree of synchronization between the changing trends of the basic features and other basic features.

[0084] According to embodiments of this disclosure, trend correlation is used to measure the degree of synchronization between the changing trend of a corresponding basic feature and the changing trend of other features. A feature that deviates from or is highly correlated with the trends of many other core features may be on the critical path of risk transmission.

[0085] Furthermore, for each basic feature i, for all other basic features j except itself, calculate the absolute correlation coefficient between the change values ​​of basic feature i and basic feature j, and then calculate the average value.

[0086] For example, the trend correlation T of the i-th basic feature i Represented as:

[0087] T i =Σ j |corr(ΔX i ,ΔX j )| / (M-1);

[0088] In the formula, ΔX i Let ΔX represent the first-order difference sequence of the i-th basic feature. j Let represent the first-order difference sequence of the j-th basic feature, corr(·,·) represents the calculation of the Pearson correlation coefficient between the two difference sequences, and M represents the total number of basic features.

[0089] When operating the S350, the risk sensitivity is obtained by integrating the industry relative importance score, characteristic anomaly degree, characteristic volatility and trend correlation.

[0090] For example, the industry relative importance score, feature anomaly, feature volatility, and trend correlation are respectively min-max normalized, and the normalized industry relative importance score, feature anomaly, feature volatility, and trend correlation are weighted and summed to obtain the risk sensitivity, with the sum of the corresponding weights being 1.

[0091] Through the above operations, the massive amount of basic features are compressed into a set of "multiple candidate basic features" that are more indicative of risk by using risk sensitivity calculations, which greatly reduces the complexity and noise interference of subsequent calculations. Through a multi-dimensional comprehensive evaluation system, it is ensured that the selected candidate basic features can comprehensively and keenly depict the current risk status of the enterprise, providing an accurate and reasonable data foundation for subsequent steps.

[0092] In some embodiments, the basic features are operational features, account behavior features, credit features, or production behavior features. That is, the set of basic features may include basic features of multiple dimensions such as operational features, account behavior features, credit features, and production behavior features.

[0093] For example, before performing operation S210, the obtained raw dataset is cleaned and time-series aligned, such as outlier handling, missing value imputation, time serial port alignment, and standardization, to obtain a basic feature set.

[0094] For example, operational characteristics may include operating revenue, gross profit margin, tax payment, number of social security contributors and growth rate over the past N years; account behavior characteristics may include summaries, counterparties, amounts, frequencies, periodic characteristics (such as regularity of salary payments), and the proportion of large transactions at night from the daily transaction records of the company's corporate accounts over the past M months; credit characteristics may include historical overdue records, inquiry frequency, watch / bad credit labels, and third-party commercial credit scores from credit reports; and production behavior characteristics may include energy consumption per unit of output and stability of production cycles. The above data may be provided by the company with its authorization.

[0095] Based on this, data is provided for enterprise risk monitoring from multiple dimensions, facilitating a comprehensive assessment of enterprise risks.

[0096] Candidate basic features are not isolated, but rather mutually influential and restrictive. To intuitively demonstrate the complex relationships between candidate basic features, abstract candidate basic features are transformed into a visual network that can reflect the association patterns of each candidate basic feature within the current time period. In some embodiments, forming a weighted undirected graph based on the pairwise associations between multiple candidate basic features includes: calculating partial correlation coefficients between multiple candidate basic features through linear regression residuals; constructing a weighted undirected graph using multiple candidate basic features as nodes, direct associations between multiple candidate basic features as edges, and the obtained partial correlation coefficients as edge weights.

[0097] According to embodiments of this disclosure, partial correlation coefficients characterize the strength of direct association between pairwise candidate basic features, filtering out indirect associations transmitted through third-party variables.

[0098] When constructing a weighted undirected graph, using partial correlation coefficients as edge weights can more accurately characterize the direct dependency structure between features and avoid false associations caused by common influencing factors (such as macroeconomic fluctuations).

[0099] Traditional correlation coefficient calculations include indirect associations and confounding factors, leading to inaccurate node connections in the resulting weighted undirected graph. To construct a more accurate weighted undirected graph and provide a reliable foundation for risk transmission analysis, the direct correlation between pairwise candidate basic features is calculated using linear regression residuals. In some embodiments, calculating partial correlation coefficients between multiple candidate basic features using linear regression residuals includes: constructing multiple linear regression models with any one of the candidate basic features as the dependent variable and the remaining candidate basic features as independent variables; calculating the residuals corresponding to each of the multiple linear regression models; and obtaining partial correlation coefficients based on the obtained residuals.

[0100] According to embodiments of this disclosure, the residuals indicate the effective information of the corresponding dependent variable after excluding the influence of other candidate basic features.

[0101] For any pairwise candidate fundamental features c i and c j For example, the partial correlation coefficient ρ {ij|R} Represented as:

[0102] ρ {ij|R} =corr(e i ,e j );

[0103] The linear regression models are expressed as follows:

[0104] c i =Σ {k≠i,j} β k ×c k +ε i ;

[0105] c j =Σ {k≠i,j} β k ×c k +ε j ;

[0106] In the formula, e i Represents candidate basic features c i The residuals corresponding to the linear regression model, i.e., c i Unable to be c k The explanation part, e j Represents candidate basic features c j The residuals corresponding to the linear regression model, c k Let β represent the k-th candidate basic feature. k ε represents the weight corresponding to the k-th candidate basic feature. i ε j This represents a constant term.

[0107] For example, using multiple candidate basic features as nodes, when a partial correlation coefficient is greater than a threshold, an edge is established between nodes i and j, and the absolute value of the partial correlation coefficient is used as the edge weight to form a weighted undirected graph.

[0108] To quantify the risk transmission effect between candidate basic features, in some embodiments, the risk transmission strength between each node and its neighboring nodes in a weighted undirected graph is iteratively calculated based on the data distribution state indicated by multiple candidate basic features. This includes: calculating the corresponding quantiles based on the candidate basic features corresponding to the node and the candidate basic features corresponding to the neighboring nodes; calculating the conditional probability and conditional expectation between each node and its neighboring nodes in the weighted undirected graph based on the candidate basic features corresponding to the node, the candidate basic features corresponding to the neighboring nodes, and the obtained multiple quantiles; and fusing the conditional probability and conditional expectation to obtain the risk transmission strength.

[0109] According to embodiments of this disclosure, quantiles indicate the anomalous baseline of the distribution of the corresponding candidate basic feature data, conditional probability indicates the probability that the corresponding node and its neighboring nodes are both anomalous, and conditional expectation indicates the anomalous depth of the corresponding node.

[0110] For example, the risk transmission intensity T {j→i} Represented as:

[0111] T {j→i} =P(c i {i,10} |c j {j,10} )×E[(Q {i,10} -c i ) / σ i |c j {j,10} ];

[0112] In the formula, P(c i {i,10} |c j {j,10} ) represents conditional probability, characterizing when c j abnormal c i Also, the possibility of an anomaly, E[(Q {i,10} -c i ) / σ i |c j {j,10} ] represents conditional expectation, and represents when c j abnormal c i abnormal depth, Q {i,10} Indicates feature c i The 10th percentile, Q {j,10} Indicates feature c j The 10th percentile. ​​​​​​

[0113] Risk transmission intensity T {j→i} It can represent c j The abnormal deterioration of feature c i The likelihood and extent of abnormal deterioration.

[0114] The passive impact index measures the cumulative degree of risk transmission from neighboring features to candidate basic features. It is used to accurately calculate the passive impact index by utilizing partial correlation coefficients, risk transmission strength, and the eigenvector centrality of the corresponding neighboring nodes. In some embodiments, for each node in the weighted undirected graph, the passive impact index is obtained in the following way: based on the partial correlation coefficients between the node and its corresponding neighboring nodes, the risk transmission strength, and the eigenvector centrality of the corresponding neighboring nodes.

[0115] According to embodiments of this disclosure, the eigenvector centrality characterizes the global importance of the corresponding neighbor node in the weighted undirected graph, and the importance of the node in the weighted undirected graph is measured based on the importance of the neighbor node.

[0116] For example, the passive influence index D of the i-th candidate basic feature i Represented as:

[0117] ;

[0118] In the formula, N(i) represents the set of neighboring nodes, and C j The eigenvector centrality of neighbor node j is represented.

[0119] Based on this, the calculation of the passive influence index comprehensively considers the correlation strength, transmission effect, and neighbor importance, ensuring the accuracy and rationality of risk transmission analysis and enabling it to better capture the risk transmission mechanism between features.

[0120] In some embodiments, the eigenvector centrality is obtained by forming an adjacency matrix based on multiple partial correlation coefficients indicated by a weighted undirected graph; and by obtaining the eigenvector centrality based on the largest eigenvalue in the adjacency matrix and the neighbor matrix.

[0121] For example, for the weighted undirected graph G=(V,E,W) obtained in the aforementioned steps, an adjacency matrix A=[w] is formed based on the absolute values ​​of its corresponding partial correlation coefficients, i.e., the edge weights. {ij} For unconnected pairs of nodes, w {ij} =0; the eigenvector centrality vector C satisfies the following formula:

[0122] ;

[0123] In the formula, λ represents the largest eigenvalue in the neighbor matrix.

[0124] The above formula is solved by power iteration to obtain the eigenvector centrality vector C.

[0125] In the embodiments of this disclosure, the eigenvector centrality is calculated using the neighbor matrix through the global structure of the entire weighted undirected graph, taking into account the mutual influence of all nodes and edges in the weighted undirected graph, rather than just local connectivity information; this makes it easier to identify features that may not be abnormal, but are located among multiple important features and are likely to become key channels for risk transmission.

[0126] Based on the above-mentioned enterprise risk monitoring methods, this disclosure also provides an enterprise risk monitoring device. The following will be combined with... Figure 4 The device is described in detail.

[0127] Figure 4 A schematic block diagram of an enterprise risk monitoring device according to an embodiment of the present disclosure is shown.

[0128] like Figure 4 As shown, the enterprise risk monitoring device 400 in this embodiment includes a screening module 410, a risk transmission module 420, a calculation module 430, an update module 440, and a risk monitoring module 450.

[0129] The filtering module 410 is used to respond to the received basic feature set of an enterprise, calculate the risk sensitivity of each basic feature in the basic feature set, and filter the basic feature set based on the obtained multiple risk sensitivities to obtain multiple candidate basic features. The basic feature set indicates the overall operational status of the enterprise. In one embodiment, the filtering module 410 can be used to perform the operation S210 described above, which will not be repeated here.

[0130] The risk transmission module 420 is used to form a weighted undirected graph based on the pairwise relationships between multiple candidate basic features, and iteratively calculate the risk transmission strength between each node and its adjacent nodes in the weighted undirected graph based on the data distribution state indicated by the multiple candidate basic features. In one embodiment, the risk transmission module 420 can be used to perform the operation S220 described above, which will not be repeated here.

[0131] The calculation module 430 is used to weight the correlation coefficients of multiple parts indicated by the undirected graph and the obtained multiple risk transmission intensities to obtain multiple passive influence indices. The passive influence index represents the cumulative degree of risk transmission from adjacent nodes to the corresponding node. In one embodiment, the calculation module 430 can be used to perform the operation S230 described above, which will not be repeated here.

[0132] The update module 440 is used to fuse the risk sensitivity and passive impact index corresponding to multiple candidate basic features to obtain multiple adaptive weights, thereby updating the corresponding candidate basic features and obtaining multiple target features. In one embodiment, the update module 440 can be used to perform the operation S240 described above, which will not be repeated here.

[0133] The risk monitoring module 450 is used to process multiple target features through a pre-trained risk monitoring model to obtain risk monitoring data, and to conduct enterprise risk warnings based on the risk monitoring data. In one embodiment, the risk monitoring module 450 can be used to perform the operation S250 described above, which will not be repeated here.

[0134] According to embodiments of this disclosure, the screening module 410 includes a first calculation submodule, a second calculation submodule, a third calculation submodule, a fourth calculation submodule, and a fusion submodule. The first calculation submodule is used to obtain a corresponding industry relative importance score based on the regulatory importance, industry usage frequency, and expert importance score of the basic feature. The industry relative importance score represents the importance of the corresponding basic feature in enterprise risk monitoring. The second calculation submodule is used to obtain a corresponding feature anomaly based on the historical anomaly, industry anomaly, and trend anomaly of the basic feature. The feature anomaly represents the degree of deviation of the corresponding basic feature. The third calculation submodule is used to obtain feature volatility based on the ratio of the standard deviation to the mean of the basic feature within a preset time period. The feature volatility represents the volatility of the basic feature in the time dimension. The fourth calculation submodule is used to obtain trend correlation based on the degree of synchronization of the changing trends between the basic feature and other basic features. The fusion submodule is used to fuse the industry relative importance score, feature anomaly, feature volatility, and trend correlation to obtain risk sensitivity.

[0135] According to embodiments of this disclosure, the risk transmission module 420 includes a first acquisition submodule, a second acquisition submodule, and a third acquisition submodule. The first acquisition submodule is used to calculate the corresponding quantiles based on the candidate basic features corresponding to the node and the candidate basic features corresponding to the adjacent nodes. The quantiles indicate the anomaly baseline of the corresponding candidate basic feature data distribution. The second acquisition submodule is used to calculate the conditional probability and conditional expectation between each node and its adjacent nodes in the weighted undirected graph based on the candidate basic features corresponding to the node, the candidate basic features corresponding to the adjacent nodes, and the obtained multiple quantiles. The conditional probability indicates the possibility that the corresponding node and its adjacent nodes are both anomalous, and the conditional expectation indicates the anomalous depth of the corresponding node. The third acquisition submodule is used to fuse the conditional probability and the conditional expectation to obtain the risk transmission intensity.

[0136] According to embodiments of this disclosure, any multiple modules among the screening module 410, risk transmission module 420, calculation module 430, update module 440, and risk monitoring module 450 can be combined into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this disclosure, at least one of the screening module 410, risk transmission module 420, calculation module 430, update module 440, and risk monitoring module 450 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or any other reasonable means of integrating or packaging circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the screening module 410, risk transmission module 420, calculation module 430, update module 440 and risk monitoring module 450 may be implemented at least partially as a computer program module, which can perform corresponding functions when the computer program module is run.

[0137] Figure 5 A block diagram of an electronic device suitable for implementing an enterprise risk monitoring method according to an embodiment of the present disclosure is shown schematically.

[0138] like Figure 5 As shown, an electronic device 500 according to an embodiment of the present disclosure includes a processor 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage portion 508 into a random access memory (RAM) 503. The processor 501 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 501 may also include onboard memory for caching purposes. The processor 501 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of the present disclosure.

[0139] RAM 503 stores various programs and data required for the operation of electronic device 500. Processor 501, ROM 502, and RAM 503 are interconnected via bus 504. Processor 501 performs various operations of the method flow according to embodiments of the present disclosure by executing programs in ROM 502 and / or RAM 503. It should be noted that the programs may also be stored in one or more memories other than ROM 502 and RAM 503. Processor 501 may also perform various operations of the method flow according to embodiments of the present disclosure by executing programs stored in said one or more memories.

[0140] According to embodiments of this disclosure, the electronic device 500 may further include an input / output (I / O) interface 505, which is also connected to a bus 504. The electronic device 500 may also include one or more of the following components connected to the I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 510 as needed so that computer programs read from it can be installed into the storage section 508 as needed.

[0141] This disclosure also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs that, when executed, implement the method according to the embodiments of this disclosure.

[0142] According to embodiments of this disclosure, the computer-readable storage medium may be a non-volatile computer-readable storage medium, such as including, but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this disclosure, the computer-readable storage medium may be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this disclosure, the computer-readable storage medium may include ROM 502 and / or RAM 503 and / or one or more memories other than ROM 502 and RAM 503 described above.

[0143] Embodiments of this disclosure also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to cause the computer system to implement the item recommendation method provided in the embodiments of this disclosure.

[0144] When the computer program is executed by the processor 501, it performs the functions defined in the system / apparatus of this disclosure embodiments. According to embodiments of this disclosure, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0145] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and may be downloaded and installed via the communication section 509, and / or installed from a removable medium 511. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0146] In such an embodiment, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by processor 501, it performs the functions defined in the system of this disclosure embodiment. According to embodiments of this disclosure, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0147] According to embodiments of this disclosure, program code for executing the computer programs provided in embodiments of this disclosure can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can execute entirely on a user's computing device, partially on a user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0148] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0149] Those skilled in the art will understand that the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways, even if such combinations or combinations are not explicitly described in this disclosure. In particular, the features described in the various embodiments and / or claims of this disclosure can be combined or combined in various ways without departing from the spirit and teachings of this disclosure. All such combinations and / or combinations fall within the scope of this disclosure.

[0150] The embodiments of this disclosure have been described above. However, these embodiments are for illustrative purposes only and are not intended to limit the scope of this disclosure. Although various embodiments have been described above, this does not mean that the measures in the various embodiments cannot be used advantageously in combination. The scope of this disclosure is defined by the appended claims and their equivalents. Various substitutions and modifications can be made by those skilled in the art without departing from the scope of this disclosure, and all such substitutions and modifications should fall within the scope of this disclosure.

Claims

1. A method for enterprise risk monitoring, characterized in that, The method includes: In response to receiving a set of basic characteristics of an enterprise, the risk sensitivity of each basic characteristic in the set of basic characteristics is calculated, and multiple candidate basic characteristics are obtained by filtering the set of basic characteristics based on the obtained multiple risk sensitivities. The set of basic characteristics indicates the overall operating status of the enterprise. Based on the pairwise relationships between the multiple candidate basic features, a weighted undirected graph is formed. Based on the data distribution state indicated by the multiple candidate basic features, the risk transmission intensity between each node and its adjacent nodes in the weighted undirected graph is iteratively calculated. Based on the multiple partial correlation coefficients indicated by the weighted undirected graph and the multiple risk transmission intensities obtained, multiple passive impact indices are obtained, and the passive impact indexes characterize the cumulative degree of risk transmission from adjacent nodes to the corresponding node; The risk sensitivity and passive impact index corresponding to the multiple candidate basic features are fused to obtain multiple adaptive weights, which are then used to update the corresponding candidate basic features to obtain multiple target features. The multiple target features are processed by a pre-trained risk monitoring model to obtain risk monitoring data, which is then used to provide enterprise risk warnings.

2. The method according to claim 1, characterized in that, The calculation of the risk sensitivity of each basic feature in the set of basic features includes: Based on the regulatory importance, industry usage frequency, and expert importance scores of the aforementioned basic features, a corresponding industry relative importance score is obtained, which represents the degree of importance of the corresponding basic feature in enterprise risk monitoring. Based on the historical anomaly degree, industry anomaly degree, and trend anomaly degree of the basic features, the corresponding feature anomaly degree is obtained, and the feature anomaly degree characterizes the degree of deviation of the corresponding basic features. Based on the ratio of the standard deviation to the mean of the basic feature within a preset time period, the feature volatility is obtained, which characterizes the volatility of the basic feature in the time dimension. The trend correlation is obtained based on the degree of synchronization between the changing trends of the aforementioned basic features and other basic features; The risk sensitivity is obtained by fusing the industry relative importance score, the feature anomaly degree, the feature volatility, and the trend correlation.

3. The method according to claim 1, characterized in that, The iterative calculation of the risk transmission strength between each node and its neighboring nodes in the weighted undirected graph, based on the data distribution state indicated by the multiple candidate basic features, includes: Based on the candidate basic features corresponding to the node and the candidate basic features corresponding to the adjacent nodes, the corresponding quantiles are calculated. The quantiles indicate the abnormal baseline of the distribution of the corresponding candidate basic feature data. Based on the candidate basic features corresponding to the node, the candidate basic features corresponding to the neighboring node, and the obtained multiple quantiles, the conditional probability and conditional expectation between each node and its neighboring nodes in the weighted undirected graph are calculated. The conditional probability indicates the possibility that the corresponding node and its neighboring nodes are both anomalous, and the conditional expectation indicates the anomalous depth of the corresponding node. The risk transmission strength is obtained by fusing the conditional probability and the conditional expectation.

4. The method according to claim 1, characterized in that, The formation of the weighted undirected graph based on the pairwise relationships between the multiple candidate basic features includes: The partial correlation coefficients between each pair of the multiple candidate basic features are calculated using linear regression residuals; A weighted undirected graph is constructed using the multiple candidate basic features as nodes and the obtained partial correlation coefficients as edge weights. The partial correlation coefficients characterize the correlation strength between the candidate basic features.

5. The method according to claim 4, characterized in that, The calculation of partial correlation coefficients between the multiple candidate basic features using linear regression residuals includes: Multiple linear regression models are constructed by using any one of the pairwise candidate basic features as the dependent variable and the remaining candidate basic features as independent variables. The residuals corresponding to the multiple linear regression models are calculated respectively. Based on the obtained residuals, partial correlation coefficients are obtained. The residuals indicate the effective information of the corresponding dependent variable after excluding the influence of other candidate basic features.

6. The method according to claim 1, characterized in that, For each node in the weighted undirected graph, the passive influence index is obtained using the following method: The passive influence index is obtained based on the partial correlation coefficient between the node and its corresponding neighbor nodes, the risk transmission strength, and the eigenvector centrality of the corresponding neighbor nodes. The eigenvector centrality characterizes the global importance of the corresponding neighbor nodes in the weighted undirected graph.

7. The method according to claim 6, characterized in that, The eigenvector centrality is obtained using the following method: An adjacency matrix is ​​formed based on the multiple partial correlation coefficients indicated by the weighted undirected graph; The eigenvector centrality is obtained based on the largest eigenvalue in the adjacency matrix and the neighbor matrix.

8. The method according to claim 1, characterized in that, The basic characteristics are operational characteristics, account behavior characteristics, credit characteristics, or production behavior characteristics.

9. A business risk monitoring device, characterized in that, The device includes: A filtering module is used to respond to receiving a set of basic characteristics of an enterprise, calculate the risk sensitivity of each basic characteristic in the set of basic characteristics, and filter the basic characteristics based on the obtained multiple risk sensitivities to obtain multiple candidate basic characteristics. The set of basic characteristics indicates the overall operating status of the enterprise. The risk transmission module is used to form a weighted undirected graph based on the pairwise correlation between the multiple candidate basic features, and to iteratively calculate the risk transmission intensity between each node and its adjacent nodes in the weighted undirected graph based on the data distribution state indicated by the multiple candidate basic features. The calculation module is used to obtain multiple passive impact indices based on multiple partial correlation coefficients indicated by the weighted undirected graph and multiple risk transmission intensities. The passive impact index represents the cumulative degree of risk transmission from adjacent nodes to the corresponding node. The update module is used to fuse the risk sensitivity and passive impact index corresponding to the multiple candidate basic features to obtain multiple adaptive weights, thereby updating the corresponding candidate basic features and obtaining multiple target features; and The risk monitoring module is used to process the multiple target features through a pre-trained risk monitoring model to obtain risk monitoring data, and to conduct enterprise risk warning based on the risk monitoring data.

10. An electronic device, comprising: One or more processors; Storage device for storing one or more programs. The characteristic is that, when the one or more programs are executed by the one or more processors, the one or more processors perform the method according to any one of claims 1 to 8.

11. A computer-readable storage medium having executable instructions stored thereon, characterized in that, When executed by the processor, this instruction causes the processor to perform the method according to any one of claims 1 to 8.

12. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the method according to any one of claims 1 to 8.

Images