An intelligent enterprise operation risk assessment method and system

Abstract

The application discloses an intelligent enterprise operation risk assessment method, collects multi-dimensional operation data and industry dynamic benchmark data of an enterprise, pre-processes the collected multi-dimensional operation data and industry dynamic benchmark data of the enterprise, constructs a dynamic multi-model fusion module based on enterprise type, business scene and data integrity characteristics, inputs the pre-processed multi-dimensional operation data of the enterprise into a customized assessment model set, obtains a preliminary risk assessment result, analyzes a risk transmission path through a risk transmission path analysis module, calls an industry dynamic benchmark database, compares the preliminary risk assessment result with benchmark values of enterprises of the same industry and scale and industry risk thresholds horizontally, calibrates a final risk assessment result, receives complaint information and verification data of the final risk assessment result through an enterprise feedback closed-loop unit, generates an enterprise risk assessment data package of the final risk assessment result, a risk transmission path, industry comparison data and model optimization records, and stores the enterprise risk assessment data package into a cloud database.

Description

Technical Field

[0001] This invention relates to the field of risk assessment, and in particular to an intelligent enterprise operation risk assessment method and system, applicable to scenarios such as credit review by financial institutions, risk assessment of enterprise cooperation, and compliance inspection by regulatory authorities. Background Technology

[0002] With the increasing complexity of the market economy and the diversification of business operations, the accuracy and comprehensiveness of business risk assessment have become increasingly important. Various business risk assessment methods and systems already exist in the current technology, such as the assessment method combining decision trees and scoring models based on logistics business characteristics, supplier risk assessment systems based on big data, and business risk analysis models incorporating artificial intelligence.

[0003] However, existing technologies still have the following shortcomings: First, model selection is fixed, with most using a single model or a combination of preset models, making it impossible to dynamically adapt the optimal assessment scheme according to enterprise type, business scenario, and data characteristics, thus limiting the accuracy of the assessment; Second, the assessment dimensions lack horizontal industry comparison, relying solely on the enterprise's own data for analysis, ignoring the impact of industry benchmarks and the market environment, resulting in insufficient objectivity of the assessment results; Third, there is a lack of risk transmission path analysis, only outputting risk level results, unable to trace the source of risk and the scope of impact, which is not conducive to targeted rectification by enterprises; Fourth, there is no complete enterprise feedback and model optimization closed loop, the model parameters are fixed, making it difficult to adapt to changes in enterprise operations and dynamic adjustments in the industry, and the assessment accuracy will decrease after long-term use.

[0004] Therefore, there is an urgent need for an intelligent enterprise business risk assessment method and system that can dynamically adapt to assessment scenarios, combine industry benchmark comparisons, have risk tracing capabilities, and have a continuous optimization mechanism to address the shortcomings of existing technologies. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent enterprise operation risk assessment method and system. Through dynamic multi-model fusion, industry benchmark comparison, risk transmission path analysis, and enterprise feedback closed loop, it achieves more accurate, interpretable, and continuously optimized risk assessment, thus overcoming the shortcomings of existing technologies.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: an intelligent enterprise operation risk assessment method, comprising the following steps: S1: Collect multi-dimensional business data of enterprises and dynamic benchmark data of industries. The multi-dimensional business data of enterprises includes financial data, production and operation data, compliance data, public opinion data and real-time behavior data. The dynamic benchmark data of industries includes industry average data, risk threshold data, policy change data and market trend data. S2: Preprocess the collected multi-dimensional business data of enterprises and dynamic benchmark data of the industry, including data cleaning, outlier removal, standardization transformation and feature labeling; S3: Construct a dynamic multi-model fusion module that, based on enterprise type, business scenario, and data integrity characteristics, adaptively selects at least two of the following models: decision tree model, scoring model, and machine learning model, to generate a customized set of evaluation models. S4: Input the preprocessed multi-dimensional business data of the enterprise into the customized assessment model set to obtain preliminary risk assessment results; S5: Through the risk transmission path analysis module, based on the Bayesian network algorithm, trace the source nodes and transmission paths of each risk item in the preliminary risk assessment results, and determine the core risk factors and their scope of impact; S6: Call the industry dynamic benchmark database, compare the preliminary risk assessment results with the benchmark values ​​of companies of the same size in the same industry and the industry risk threshold, and calibrate to obtain the final risk assessment results. The final risk assessment results include risk level, core risk factors, transmission path and improvement suggestions. S7: Receive the enterprise's appeal information and verification data regarding the final risk assessment results through the enterprise feedback closed-loop unit. After verifying the authenticity of the verification data, update the model weights and feature parameters of the dynamic multi-model fusion module to complete the model iterative optimization. S8: Generate an enterprise risk assessment data package by recording the final risk assessment results, risk transmission path, industry comparison data, and model optimization, and store it in the cloud database.

[0007] As a preferred embodiment of the present invention, the model selection logic of the dynamic multi-model fusion module in step S3 includes: S31: Extract enterprise type tags, business scenario tags, and data integrity scores. The enterprise type tags include manufacturing, service, and financial industries, and the business scenario tags include financing applications, cooperation assessments, and annual audits. S32: Preset model adaptation rule base, which records the optimal model combination corresponding to different enterprise types, business scenarios and data integrity scores; S33: Based on the enterprise's type label, business scenario label, and data integrity score, match the model adaptation rule base and select at least two models to form a customized evaluation model set; S34: The output results of the customized evaluation model set are calculated using a weighted voting method. The model weights are dynamically adjusted based on the historical evaluation accuracy, and the higher the accuracy, the greater the weight.

[0008] As a preferred embodiment of the present invention, the workflow of the risk transmission path analysis module in step S5 includes: S51: Construct a risk factor correlation network, wherein the risk factors include financial factors, operational factors, compliance factors, and public opinion factors; S52: Based on the risk items in the preliminary risk assessment results, locate the terminal risk nodes in the associated network; S53: Use Bayesian network algorithms to reverse-engineer the upstream associated nodes of terminal risk nodes to form a risk transmission path chain; S54: Calculate the risk contribution of each node in the path chain. Risk contribution = node influence coefficient × node occurrence probability. Mark nodes with risk contribution higher than the preset threshold as core risk factors.

[0009] As a preferred embodiment of the present invention, the update mechanism of the industry dynamic benchmark database includes: S01: Collect industry data in real time through web crawlers, industry API interfaces, and third-party data service provider channels; S02: Perform timeliness screening on the collected industry data and remove data that exceeds the preset validity period; S03: Classify and statistically analyze industry data according to enterprise type, size, and region, and generate industry average data and risk threshold data; S04: When industry policies change or market trends reach a turning point, an emergency database update is triggered, and risk assessment calibration parameters are adjusted synchronously.

[0010] As a preferred embodiment of the present invention, the workflow of the enterprise feedback closed-loop unit in step S7 includes: S71: Receive the appeal information and corresponding verification data submitted by the enterprise, wherein the verification data includes third-party audit reports, business vouchers, and compliance certification documents; S72: Verify the authenticity of the verification data, including document signature verification, data cross-comparison, and tracing the source of the data through third-party institutions; S73: If the verification passes, extract the valid features from the verification data and adjust the feature weights and decision rules of the corresponding models in the dynamic multi-model fusion module. S74: Feedback the appeal processing results and model adjustment records to the enterprise, and update the enterprise's risk assessment data package.

[0011] As a preferred technical solution of the present invention, the risk level of the final risk assessment result includes five levels: excellent, good, average, poor, and dangerous. Each level corresponds to a specific risk response strategy, which includes continuous monitoring, rectification within a time limit, enhanced review, suspension of cooperation, and risk warning.

[0012] An intelligent enterprise operational risk assessment system includes: a data acquisition module for collecting multi-dimensional operational data and industry dynamic benchmark data, comprising a financial data interface, an IoT data acquisition unit, a public opinion monitoring unit, and an industry data crawling unit; a data preprocessing module for cleaning, outlier removal, standardization, and feature labeling of the collected data; a dynamic multi-model fusion module for adaptively selecting the optimal model combination based on enterprise type, business scenario, and data characteristics, generating a customized assessment model set, and outputting preliminary risk assessment results; and a risk transmission path analysis module for tracing the source of risk and forming a transmission path analysis based on a Bayesian network algorithm. The system includes: a risk assessment path and identification of core risk factors; an industry dynamic benchmark database for storing and updating industry average data, risk threshold data, policy change data, and market trend data in real time, providing a basis for horizontal comparison and calibration; an assessment result generation module for combining preliminary risk assessment results, risk transmission paths, and industry benchmark data to generate a final risk assessment result that includes risk level, core risk factors, transmission paths, and improvement suggestions; a corporate feedback closed-loop unit for receiving corporate appeal information and verification data, completing authenticity verification, and driving model iteration and optimization; and a cloud storage module for storing corporate risk assessment data packages, model parameters, industry benchmark data, and feedback processing records.

[0013] As a preferred technical solution of the present invention, the dynamic multi-model fusion module has a built-in model library, which includes decision tree model, logistic regression scoring model, random forest model, gradient boosting tree model and neural network model. Each model has pre-stored adaptation parameters corresponding to different industries and scenarios.

[0014] As a preferred technical solution of the present invention, it also includes a risk warning module, which is used to send graded warning signals to enterprises and related parties based on the final risk assessment results and risk transmission paths. The warning signals include SMS warnings, system pop-up warnings, and email warnings.

[0015] Compared with the prior art, the beneficial effects that this invention can achieve are: 1. By using a dynamic multi-model fusion module, an evaluation solution can be customized based on enterprise type, business scenario, and data characteristics, which solves the problem of fixed existing technical models and improves the evaluation accuracy. 2. The risk transmission path analysis module clearly identifies the source of risk and the transmission process, breaking through the limitation of existing technologies that "only provide results without providing reasons," thus facilitating targeted rectification by enterprises; 3. Introduce an industry dynamic benchmark database for horizontal comparison to avoid the limitations of data from a single company, making the evaluation results more valuable. 4. The enterprise feedback closed-loop unit enables dynamic updates of model parameters, adapting to industry changes and enterprise operational adjustments, resulting in stronger long-term stability; it also covers multiple industries and scenarios, meeting the risk assessment needs of different entities such as financial institutions, enterprise cooperation, and regulatory inspections. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating the overall process of the intelligent enterprise operation risk assessment method of the present invention. Figure 2 This is a flowchart of the risk transmission path analysis for this invention; Figure 3 This is a flowchart illustrating the industry dynamic benchmark database update process of this invention. Figure 4 This is a flowchart of the closed-loop feedback process for the enterprise in this invention. Detailed Implementation

[0017] To make the technical means, creative features, objectives, and effects of this invention easier to understand, the invention is further described below with reference to specific embodiments. However, the following embodiments are merely preferred embodiments of this invention and not all of them. Other embodiments obtained by those skilled in the art based on the embodiments described herein without creative effort are all within the protection scope of this invention.

[0018] Example 1: Specific Implementation of Intelligent Enterprise Operation Risk Assessment Method Step S1: Data Acquisition Enterprise multi-dimensional operational data collection: Obtaining financial data such as balance sheets, profit and loss statements, and cash flow statements through enterprise financial system interfaces; collecting production and operation data such as equipment utilization rate, production cycle time, and electricity consumption characteristics through IoT devices; obtaining compliance data such as administrative penalties and qualification certifications through government platform APIs; collecting public opinion data such as media reports and social media comments through web crawlers; and obtaining real-time behavioral data such as employee attendance rate and working hours through enterprise attendance systems and office monitoring.

[0019] Industry dynamic benchmark data collection: Obtain industry average data such as average debt-to-asset ratio and average revenue growth rate of sub-sectors such as manufacturing, service and finance through industry association APIs; collect industry risk thresholds and policy change notices through the official websites of regulatory authorities; obtain market trends and competitor operating data through third-party data service providers.

[0020] Step S2: Data Preprocessing Data cleaning: removing missing values ​​from financial data and invalid comments from public opinion data; Outlier removal: The 3σ criterion is used to remove extreme outliers in production data (the 3σ criterion is also known as the Raida criterion. It first assumes that a set of test data contains only random errors, calculates and processes it to obtain the standard deviation, and determines an interval according to a certain probability. Any error exceeding this interval is not a random error but a gross error, and data containing such errors should be removed). Standardization conversion: Converting financial indicators of different magnitudes (such as revenue and net profit) into standardized scores (0-100 points). Feature labeling: The data is labeled with "financial", "operational", "compliance" and "public opinion" categories to provide a basis for model selection.

[0021] Step S3: Dynamic Multi-Model Fusion Extracting company tags: A manufacturing company applied for financing, and its data integrity score was 92 points, corresponding to the tags "manufacturing + financing application + high data integrity"; Matching model combination: Query the model adaptation rule base. The optimal combination corresponding to this tag is "gradient boosting tree model (weight 0.4) + scoring model (weight 0.3) + random forest model (weight 0.3)". Determine the calculation method: Use a weighted voting method to fuse the model output results, and dynamically adjust the weights based on the historical evaluation accuracy.

[0022] Step S4: Preliminary Risk Assessment Results The preprocessed data was input into three models, which yielded risk levels of "poor", "poor", and "moderate" respectively. The weighted voting method is used to calculate: (0.4 × Poor + 0.3 × Poor + 0.3 × Average) = The preliminary risk assessment result is "Poor".

[0023] Step S5: Risk Transmission Path Analysis Constructing a correlation network: Construct a risk factor correlation network based on financial factors (accounts receivable, cash flow), operational factors (equipment utilization rate), and compliance factors (qualification certificates); Identifying the terminal risk node: Based on the preliminary assessment results, the terminal risk node was identified as the "cash flow gap"; Reverse derivation path: By tracing upstream related nodes through Bayesian network algorithms, a transmission path chain is formed: "customer default → delayed accounts receivable collection → cash flow gap → increased operational risk"; Contribution calculation: Risk contribution of delayed accounts receivable collection = 0.9 (node ​​impact coefficient) × 0.94 (node ​​occurrence probability) = 85%, marked as core risk factor.

[0024] Step S6: Industry Benchmark Comparison and Calibration Database query: The industry average debt-to-asset ratio in the manufacturing financing scenario is 50%, the risk threshold is 70%, and the company's debt-to-asset ratio is 68%. Calibration Analysis: The company's debt-to-equity ratio is higher than the industry average but not to the risk threshold. Considering the core risk factor (delayed accounts receivable collection is a seasonal factor in the industry), the preliminary result of "poor" is calibrated to "average". The final result is marked "Accounts receivable collection needs to be closely monitored," with the corresponding improvement suggestion being "Establish a customer credit rating system and optimize the accounts receivable collection mechanism."

[0025] Step S7: Enterprise Feedback Closed-Loop Optimization The company appealed: It submitted a third-party credit report proving that its core customers had no default records and that the delay in accounts receivable was due to seasonal factors in the industry. Verification: The system confirms the validity of the appeal by verifying the signature on the credit report and cross-referencing seasonal industry data; Model Adjustment: The weight of the "Accounts Receivable Collection Cycle" feature in the gradient boosting tree model has been adjusted from 0.25 to 0.18; Feedback result: The appeal to the company was approved, the risk assessment result was updated to "good", and the company's risk assessment data package in the cloud database was updated simultaneously.

[0026] Step S8: Data storage Packaged data includes: final risk assessment result (good), risk transmission path diagram, industry comparison data, complaint handling records, and model adjustment parameters; Storage method: Generate enterprise risk assessment data packages, use Alibaba Cloud OSS for encrypted storage, set access control, and support subsequent queries and secondary assessment calls.

[0027] Example 2: Specific Components of an Intelligent Enterprise Operation Risk Assessment System Data acquisition module: includes financial data interface (connecting to financial software such as Yonyou and Kingdee), IoT data acquisition unit (connecting to device sensors and smart meters), public opinion monitoring unit (deploying web crawlers), and industry data crawling unit (connecting to industry association APIs); Data preprocessing module: Data processing algorithms are written in Python, integrating functions such as missing value imputation, outlier removal, and standardization transformation; Dynamic multi-model fusion module: The built-in model library is trained using the TensorFlow framework, the model adaptation rule library is stored in a MySQL database, and the weighted voting algorithm is implemented in Java; Risk transmission path analysis module: Implements Bayesian network algorithm based on PyTorch, and visualizes risk transmission path; Industry dynamic benchmark database: MongoDB is used to store unstructured industry data, MySQL is used to store structured statistical data, and an automatic update mechanism is set up at 3 am every day; Assessment Result Generation Module: Generates a Word format assessment report and visual charts, including risk level, core risk factors, transmission path, and improvement suggestions; Enterprise feedback closed-loop unit: Develop a web-based appeal portal, integrate a document verification interface (connecting to a third-party electronic signature verification platform), and synchronize the model parameter adjustment interface to the dynamic multi-model fusion module in real time; Cloud storage module: Alibaba Cloud OSS can be used to store enterprise risk assessment data packages, and access control and data encryption mechanisms can be set up; Risk warning module: When the risk level is "poor" or "dangerous", it automatically sends SMS warnings to credit review personnel of financial institutions and email warnings to enterprises.

[0028] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent enterprise operational risk assessment method, characterized in that: Includes the following steps: S1: Collect multi-dimensional business data of enterprises and dynamic benchmark data of industries. The multi-dimensional business data of enterprises includes financial data, production and operation data, compliance data, public opinion data and real-time behavior data. The dynamic benchmark data of industries includes industry average data, risk threshold data, policy change data and market trend data. S2: Preprocess the collected multi-dimensional business data of enterprises and dynamic benchmark data of the industry, including data cleaning, outlier removal, standardization transformation and feature labeling; S3: Construct a dynamic multi-model fusion module that, based on enterprise type, business scenario, and data integrity characteristics, adaptively selects at least two of the following models: decision tree model, scoring model, and machine learning model, to generate a customized set of evaluation models. S4: Input the preprocessed multi-dimensional business data of the enterprise into the customized assessment model set to obtain preliminary risk assessment results; S5: Through the risk transmission path analysis module, based on the Bayesian network algorithm, trace the source nodes and transmission paths of each risk item in the preliminary risk assessment results, and determine the core risk factors and their scope of impact; S6: Call the industry dynamic benchmark database, compare the preliminary risk assessment results with the benchmark values ​​of companies of the same size in the same industry and the industry risk threshold, and calibrate to obtain the final risk assessment results. The final risk assessment results include risk level, core risk factors, transmission path and improvement suggestions. S7: Receive the enterprise's appeal information and verification data regarding the final risk assessment results through the enterprise feedback closed-loop unit. After verifying the authenticity of the verification data, update the model weights and feature parameters of the dynamic multi-model fusion module to complete the model iterative optimization. S8: Generate an enterprise risk assessment data package by recording the final risk assessment results, risk transmission path, industry comparison data, and model optimization, and store it in the cloud database.

2. The intelligent enterprise operation risk assessment method according to claim 1, characterized in that: The model selection logic of the dynamic multi-model fusion module in step S3 includes: S31: Extract enterprise type tags, business scenario tags, and data integrity scores. The enterprise type tags include manufacturing, service, and financial industries, and the business scenario tags include financing applications, cooperation assessments, and annual audits. S32: Preset model adaptation rule base, which records the optimal model combination corresponding to different enterprise types, business scenarios and data integrity scores; S33: Based on the enterprise's type label, business scenario label, and data integrity score, match the model adaptation rule base and select at least two models to form a customized evaluation model set; S34: The output results of the customized evaluation model set are calculated using a weighted voting method. The model weights are dynamically adjusted based on the historical evaluation accuracy, and the higher the accuracy, the greater the weight.

3. The intelligent enterprise operation risk assessment method according to claim 1, characterized in that: The workflow of the risk transmission path analysis module described in step S5 includes: S51: Construct a risk factor correlation network, wherein the risk factors include financial factors, operational factors, compliance factors, and public opinion factors; S52: Based on the risk items in the preliminary risk assessment results, locate the terminal risk nodes in the associated network; S53: Use Bayesian network algorithms to reverse-engineer the upstream associated nodes of terminal risk nodes to form a risk transmission path chain; S54: Calculate the risk contribution of each node in the path chain. Risk contribution = node influence coefficient × node occurrence probability. Mark nodes with risk contribution higher than the preset threshold as core risk factors.

4. The intelligent enterprise operation risk assessment method according to claim 1, characterized in that: The update mechanism of the industry dynamic benchmark database includes: S01: Collect industry data in real time through web crawlers, industry API interfaces, and third-party data service provider channels; S02: Perform timeliness screening on the collected industry data and remove data that exceeds the preset validity period; S03: Classify and statistically analyze industry data according to enterprise type, size, and region, and generate industry average data and risk threshold data; S04: When industry policies change or market trends reach a turning point, an emergency database update is triggered, and risk assessment calibration parameters are adjusted synchronously.

5. The intelligent enterprise operation risk assessment method according to claim 1, characterized in that: The workflow of the enterprise feedback closed-loop unit described in step S7 includes: S71: Receive the appeal information and corresponding verification data submitted by the enterprise, wherein the verification data includes third-party audit reports, business vouchers, and compliance certification documents; S72: Verify the authenticity of the verification data, including document signature verification, data cross-comparison, and tracing the source of the data through third-party institutions; S73: If the verification passes, extract the valid features from the verification data and adjust the feature weights and decision rules of the corresponding models in the dynamic multi-model fusion module. S74: Feedback the appeal processing results and model adjustment records to the enterprise, and update the enterprise's risk assessment data package.

6. The intelligent enterprise operation risk assessment method according to claim 1, characterized in that: The final risk assessment results include five risk levels: excellent, good, average, poor, and dangerous. Each level corresponds to a specific risk response strategy, which includes continuous monitoring, rectification within a specified period, enhanced review, suspension of cooperation, and risk warning.

7. An intelligent enterprise operational risk assessment system, characterized in that: include: Data acquisition module: used to collect multi-dimensional business data and industry dynamic benchmark data of enterprises. The data acquisition module includes a financial data interface, an Internet of Things data acquisition unit, a public opinion monitoring unit, and an industry data crawling unit. Data preprocessing module: used to clean the collected data, remove outliers, standardize the data, and label the data. Dynamic multi-model fusion module: used to adaptively select the optimal model combination based on enterprise type, business scenario and data characteristics, generate a customized set of evaluation models and output preliminary risk assessment results; Risk transmission path analysis module: used to trace the source of risk, form the transmission path chain and identify the core risk factors based on Bayesian network algorithm; Industry Dynamics Benchmark Database: Used to store and update industry average data, risk threshold data, policy change data, and market trend data in real time, providing a basis for horizontal comparison and calibration; Assessment Result Generation Module: This module combines preliminary risk assessment results, risk transmission paths, and industry benchmark data to generate a final risk assessment result that includes risk level, core risk factors, transmission paths, and improvement suggestions. Enterprise feedback closed-loop unit: used to receive enterprise appeal information and verification data, complete authenticity verification and drive model iterative optimization; Cloud storage module: Used to store enterprise risk assessment data packages, model parameters, industry benchmark data, and feedback processing records.

8. The intelligent enterprise operation risk assessment system according to claim 7, characterized in that: The dynamic multi-model fusion module has a built-in model library, which includes decision tree models, logistic regression scoring models, random forest models, gradient boosting tree models, and neural network models. Each model has pre-stored adaptation parameters corresponding to different industries and scenarios.

9. The intelligent enterprise operation risk assessment system according to claim 7, characterized in that: It also includes a risk warning module, which is used to send tiered warning signals to enterprises and related parties based on the final risk assessment results and risk transmission paths. The warning signals include SMS warnings, system pop-up warnings, and email warnings.

Images