Enterprise abnormal capital collection risk prediction method based on government affair data

Abstract

The invention discloses an enterprise abnormal capital collection risk prediction method based on government affair data, and relates to the technical field of risk prevention and control, and the method comprises the steps: 1, data collection and integration, 2, knowledge graph construction, 3, risk feature rule mining, 4, risk prediction and scoring, and 5, risk prediction and scoring. And step 5, performing result application: applying to risk early warning: performing real-time early warning on high-risk enterprises, providing decision basis for supervision departments, and applying to investment reference: providing enterprise abnormal capital risk assessment results for enterprise partners and investors for investment decision.

Description

Technical Field

[0001] The present invention discloses a method for predicting the risk of abnormal fund-raising of an enterprise based on government affairs data, and relates to the technical field of risk prevention and control. Background Art

[0002] Abnormal fundraising seriously disrupts normal economic and financial order, causing participants to suffer financial losses and even financial hardship, and can easily trigger social instability and numerous public security issues. Currently, research on abnormal fundraising is largely limited to theoretical studies, relatively short-lived, and still in its preliminary stages. Researchers generally tend to analyze issues such as flaws in the financial system, inadequate oversight by relevant departments, and the lack of relevant laws and regulations. However, research on identifying abnormal fundraising in my country using big data technologies and data mining methods is still incomplete. Summary of the Invention

[0003] In response to the problems of the prior art, the present invention provides a method for predicting the risk of abnormal fundraising of enterprises based on government data. The method has the characteristics of strong versatility and simple implementation, and has broad application prospects.

[0004] The specific scheme proposed by the present invention is:

[0005] The present invention provides a method for predicting the risk of abnormal enterprise fundraising based on government affairs data, comprising:

[0006] Step 1: Data collection and integration:

[0007] Obtain multi-source data from various government departments through the data application platform.

[0008] Conduct data integration: standardize the collected multi-source data and build a panoramic portrait of the enterprise.

[0009] Step 2: Knowledge graph construction:

[0010] Entity identification: Identify entities from the enterprise's panoramic portrait, including enterprises, legal persons, investors, directors, supervisors, and senior managers, and affiliated companies.

[0011] Perform relationship extraction: extract the relationship between entities, including investment relationship, guarantee relationship, equity relationship,

[0012] Generate knowledge graph: Represent entities and relationships in the form of a graph, use a graph database to store the knowledge graph data generated, and perform visual display.

[0013] Step 3: Conduct risk feature rule mining:

[0014] Conducting correlation comparison learning: Based on the machine learning decision tree algorithm, the data of known abnormal fundraising enterprises and normal enterprises in the existing data are compared to find the characteristic points of abnormal fundraising enterprises, classify them, and analyze the characteristic differences of different categories.

[0015] Detect anomalies: Based on massive amounts of data, use statistical methods or machine learning algorithms to find abnormal information points that are significantly different from other companies.

[0016] Generate risk feature rules: Generate risk feature rules based on the results of association comparison learning and anomaly detection.

[0017] Step 4: Risk prediction and scoring:

[0018] Extract features: Extract features related to the risk of abnormal fundraising from the company's panoramic portrait and knowledge graph. The relevant features include corporate identity information, business behavior, ethnic relations, and negative public opinion.

[0019] Training model: Based on interpretability, the extracted features are trained using the logistic regression algorithm to build a risk prediction model.

[0020] Risk prediction: Input the characteristic data of the enterprise to be predicted into the trained model to predict whether the enterprise has abnormal fundraising risks.

[0021] Risk scoring: Based on risk feature rules and model prediction results, the company is given a risk score. For example, a comprehensive assessment is conducted based on factors such as the company's group risk, business behavior, ethnic relations, negative public opinion, etc., to give a risk score.

[0022] Step 5: Apply the results:

[0023] Applied to risk warning: Real-time warning for high-risk enterprises, providing decision-making basis for regulatory authorities,

[0024] Application to investment reference: Provide enterprise partners and investors with risk assessment results of abnormal fundraising for investment decision-making.

[0025] Furthermore, when collecting data in step 1 of the method for predicting the risk of abnormal fundraising of enterprises based on government data, multi-source data is obtained from various government departments through the data application platform. The multi-source data includes market entity information, basic information of legal persons, basic information of directors, supervisors and senior managers, basic information of affiliated enterprises, paid-in and subscribed capital information of investors, social security participation information of enterprises, basic information of annual reports of enterprises, basic information of shareholders of enterprises, basic information of enterprise investments, information on the list of abnormal operations, administrative penalty information, dishonest debtors, restrictions on high consumption, information on terminated cases, enterprise tax information, and online public opinion information.

[0026] Furthermore, in step 1 of the method for predicting the risk of abnormal fundraising of enterprises based on government data, standardized governance is carried out, including noise removal and missing value processing in source data, unification of data formats and coding standards, association and fusion of data from different sources, formulation of indicator strategies and processing based on standardized data relationships, thereby constructing a complete panoramic portrait of the enterprise.

[0027] Furthermore, when detecting outliers in step 3 of the method for predicting the risk of abnormal fundraising of enterprises based on government data, based on massive data, the mean, variance, distribution analysis methods in statistical methods or the isolation forest algorithm and local anomaly factor algorithm in machine learning algorithms are used to discover abnormal information points that are significantly different from other enterprises.

[0028] The present invention also provides a device for predicting the risk of abnormal corporate fundraising based on government data, including a collection and integration module, a knowledge graph construction module, a rule mining module, a risk prediction and scoring module, and an application module.

[0029] The collection and integration module collects and integrates data:

[0030] Obtain multi-source data from various government departments through the data application platform.

[0031] Conduct data integration: standardize the collected multi-source data and build a panoramic portrait of the enterprise.

[0032] The knowledge graph construction module constructs the knowledge graph:

[0033] Entity identification: Identify entities from the enterprise's panoramic portrait, including enterprises, legal persons, investors, directors, supervisors, and senior managers, and affiliated companies.

[0034] Perform relationship extraction: extract the relationship between entities, including investment relationship, guarantee relationship, equity relationship,

[0035] Generate knowledge graph: Represent entities and relationships in the form of a graph, use a graph database to store the knowledge graph data generated, and perform visual display.

[0036] The rule mining module mines risk feature rules:

[0037] Conducting correlation comparison learning: Based on the machine learning decision tree algorithm, the data of known abnormal fundraising enterprises and normal enterprises in the existing data are compared to find the characteristic points of abnormal fundraising enterprises, classify them, and analyze the characteristic differences of different categories.

[0038] Detect anomalies: Based on massive amounts of data, use statistical methods or machine learning algorithms to find abnormal information points that are significantly different from other companies.

[0039] Generate risk feature rules: Generate risk feature rules based on the results of association comparison learning and anomaly detection.

[0040] Risk Prediction and Scoring Module Risk Prediction and Scoring:

[0041] Extract features: Extract features related to the risk of abnormal fundraising from the company's panoramic portrait and knowledge graph. The relevant features include corporate identity information, business behavior, ethnic relations, and negative public opinion.

[0042] Training model: Based on interpretability, the extracted features are trained using the logistic regression algorithm to build a risk prediction model.

[0043] Risk prediction: Input the characteristic data of the enterprise to be predicted into the trained model to predict whether the enterprise has abnormal fundraising risks.

[0044] Risk scoring: Based on risk feature rules and model prediction results, the company is given a risk score. For example, a comprehensive assessment is conducted based on factors such as the company's group risk, business behavior, ethnic relations, negative public opinion, etc., to give a risk score.

[0045] Apply the module to apply the results:

[0046] Applied to risk warning: Real-time warning for high-risk enterprises, providing decision-making basis for regulatory authorities,

[0047] Application to investment reference: Provide enterprise partners and investors with risk assessment results of abnormal fundraising for investment decision-making.

[0048] Furthermore, when the collection and integration module of the enterprise abnormal fundraising risk prediction method based on government data collects data, multi-source data is obtained from various government departments through the data application platform. The multi-source data includes market entity information, basic information of legal persons, basic information of directors, supervisors and senior managers, basic information of affiliated enterprises, paid-in and subscribed capital information of investors, social security participation information of enterprises, basic information of annual reports of enterprises, basic information of shareholders of enterprises, basic information of enterprise investments, information on the list of abnormal operations, administrative penalty information, dishonest debtors, restrictions on high consumption, information on terminated cases, enterprise tax information, and online public opinion information.

[0049] Furthermore, the collection and integration module of the method for predicting the risk of abnormal fundraising of enterprises based on government data is standardized, including noise removal and missing value processing in source data, unified data format and coding standards, association and fusion of data from different sources, formulation of indicator strategies and processing based on standardized data relationships, so as to construct a complete panoramic portrait of the enterprise.

[0050] Furthermore, when the rule mining module of the method for predicting the risk of abnormal fundraising of enterprises based on government data performs outlier detection, based on massive data, the mean, variance, distribution analysis methods in statistical methods or the isolation forest algorithm and local anomaly factor algorithm in machine learning algorithms are used to discover abnormal information points that are significantly different from other enterprises.

[0051] The benefits of the present invention are:

[0052] (1) Integration of multi-source government data: Integrate multi-source data including market entity information, basic information of legal persons, basic information of directors, supervisors and senior managers, basic information of affiliated enterprises, paid-in and subscribed capital information of investors, social security participation information of enterprises, basic information of annual reports of enterprises, basic information of shareholders of enterprises, basic information of enterprise investments, information on abnormal business operations, administrative penalty information, persons subject to execution for dishonesty, information on restrictions on high consumption, information on finalized cases, enterprise tax payment information, online public opinion information, etc., to build a complete enterprise portrait and present enterprise information in an all-round way.

[0053] (2) Application of knowledge graph technology: With the help of knowledge graph technology, based on certain enterprise association dimensions, such as investment relationships, guarantee relationships, equity relationships, etc., multi-source heterogeneous data is converted into the form of knowledge graphs to more accurately identify enterprises' illegal fundraising activities.

[0054] (3) Risk feature rule mining: a. Correlation comparison learning: By comparing known illegal fundraising companies with normal companies, we can find the risk features that distinguish illegal fundraising companies from normal companies. For example, illegal fundraising companies generally engage in false advertising and inflated registered capital. b. Anomaly detection: Based on a large amount of data, we can find abnormal information points that are significantly different from other companies. For example, if a company frequently changes its legal person, investors, senior executives, etc., or if its registered capital is increased significantly multiple times in a short period of time, there may be potential risks of abnormal business behavior.

[0055] (4) Functional Overview: This invention integrates multi-source government data to construct a panoramic portrait of the enterprise, and uses knowledge graph technology and risk feature rule mining methods to predict and score the risk of illegal fundraising by enterprises, providing a basis for decision-making for regulatory authorities, corporate partners, and investors. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0057] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments are not intended to limit the present invention.

[0058] The present invention provides a method for predicting the risk of abnormal enterprise fundraising based on government affairs data, comprising:

[0059] Step 1: Data collection and integration:

[0060] Obtain multi-source data from various government departments through the data application platform.

[0061] Conduct data integration: standardize the collected multi-source data and build a panoramic portrait of the enterprise.

[0062] When collecting data, multi-source data is obtained from various government departments through the data application platform. The multi-source data includes market entity information, basic information of legal persons, basic information of directors, supervisors and senior managers, basic information of affiliated enterprises, paid-in and subscribed capital information of investors, social security participation information of enterprises, basic information of annual reports of enterprises, basic information of shareholders of enterprises, basic information of enterprise investments, information on the list of abnormal operations, administrative penalty information, persons subject to execution for dishonesty, restrictions on high consumption, information on finalized cases, enterprise tax information, and online public opinion information.

[0063] Carry out standardized governance, including noise removal and missing value processing in source data, unify data formats and coding standards, associate and integrate data from different sources, formulate indicator strategies and process them based on standardized data relationships, so as to build a complete panoramic portrait of the enterprise.

[0064] Step 2: Knowledge graph construction:

[0065] Entity identification: Identify entities from the enterprise's panoramic portrait, including enterprises, legal persons, investors, directors, supervisors, and senior managers, and affiliated companies.

[0066] Perform relationship extraction: extract the relationship between entities, including investment relationship, guarantee relationship, equity relationship,

[0067] Generate knowledge graph: Represent entities and relationships in the form of a graph, use a graph database such as NebulaGraph to store the knowledge graph data, and visualize it.

[0068] Step 3: Conduct risk feature rule mining:

[0069] Conducting correlation comparison learning: Based on the machine learning decision tree algorithm, the data of known abnormal fundraising enterprises and normal enterprises in the existing data are compared to find the characteristic points of abnormal fundraising enterprises, classify them, and analyze the characteristic differences of different categories.

[0070] When detecting outliers in step 3, based on massive data, we use statistical methods such as mean, variance, and distribution analysis, or machine learning algorithms such as the isolation forest algorithm and local anomaly factor algorithm to find outlier information points that are significantly different from those of other companies.

[0071] Generate risk signature rules: Generate risk signature rules based on the results of association comparison learning and anomaly detection. For example, if a company frequently changes registration items such as the legal person, investors, directors, supervisors, and senior managers, and registered capital, and if the registered capital is significantly increased multiple times in a short period of time, it will be marked as a potential illegal fundraising risk company.

[0072] Step 4: Risk prediction and scoring:

[0073] Extract features: Extract features related to the risk of abnormal fundraising from the company's panoramic portrait and knowledge graph. Relevant features include corporate identity information, business behavior, ethnic relations, negative public opinion, etc.

[0074] Training model: Based on interpretability, the extracted features are trained using the logistic regression algorithm to build a risk prediction model.

[0075] Risk prediction: Input the characteristic data of the enterprise to be predicted into the trained model to predict whether the enterprise has abnormal fundraising risks.

[0076] Risk Scoring: Companies are assessed for risk based on risk profile rules and model predictions. For example, a comprehensive assessment of factors such as a company's group risk, business practices, inter-ethnic relations, and negative public opinion can be used to determine a risk score.

[0077] Step 5: Apply the results:

[0078] Applied to risk warning: Real-time warning for high-risk enterprises, providing decision-making basis for regulatory authorities,

[0079] Application to investment reference: Provide enterprise partners and investors with risk assessment results of abnormal fundraising for investment decision-making.

[0080] Example 2

[0081] The present invention also provides a device for predicting the risk of abnormal corporate fundraising based on government data, including a collection and integration module, a knowledge graph construction module, a rule mining module, a risk prediction and scoring module, and an application module.

[0082] The collection and integration module collects and integrates data:

[0083] Obtain multi-source data from various government departments through the data application platform.

[0084] Conduct data integration: standardize the collected multi-source data and build a panoramic portrait of the enterprise.

[0085] The knowledge graph construction module constructs the knowledge graph:

[0086] Entity identification: Identify entities from the enterprise's panoramic portrait, including enterprises, legal persons, investors, directors, supervisors, and senior managers, and affiliated companies.

[0087] Perform relationship extraction: extract the relationship between entities, including investment relationship, guarantee relationship, equity relationship,

[0088] Generate knowledge graph: Represent entities and relationships in the form of a graph, use a graph database to store the knowledge graph data generated, and perform visual display.

[0089] The rule mining module mines risk feature rules:

[0090] Conducting correlation comparison learning: Based on the machine learning decision tree algorithm, the data of known abnormal fundraising enterprises and normal enterprises in the existing data are compared to find the characteristic points of abnormal fundraising enterprises, classify them, and analyze the characteristic differences of different categories.

[0091] Detect anomalies: Based on massive amounts of data, use statistical methods or machine learning algorithms to find abnormal information points that are significantly different from other companies.

[0092] Generate risk feature rules: Generate risk feature rules based on the results of association comparison learning and anomaly detection.

[0093] Risk Prediction and Scoring Module Risk Prediction and Scoring:

[0094] Extract features: Extract features related to the risk of abnormal fundraising from the company's panoramic portrait and knowledge graph. The relevant features include corporate identity information, business behavior, ethnic relations, and negative public opinion.

[0095] Training model: Based on interpretability, the extracted features are trained using the logistic regression algorithm to build a risk prediction model.

[0096] Risk prediction: Input the characteristic data of the enterprise to be predicted into the trained model to predict whether the enterprise has abnormal fundraising risks.

[0097] Risk scoring: Based on risk feature rules and model prediction results, the enterprise is scored for risk.

[0098] Apply the module to apply the results:

[0099] Applied to risk warning: Real-time warning for high-risk enterprises, providing decision-making basis for regulatory authorities,

[0100] Application to investment reference: Provide enterprise partners and investors with risk assessment results of abnormal fundraising for investment decision-making.

[0101] Since the information interaction, execution process and other contents between the modules in the above-mentioned device are based on the same concept as the embodiment of the method of the present invention, the specific contents can be found in the description of the embodiment of the method of the present invention and will not be repeated here.

[0102] Likewise, the device of the present invention is beneficial in that:

[0103] (1) Integration of multi-source government data: Integrate multi-source data including market entity information, basic information of legal persons, basic information of directors, supervisors and senior managers, basic information of affiliated enterprises, paid-in and subscribed capital information of investors, social security participation information of enterprises, basic information of annual reports of enterprises, basic information of shareholders of enterprises, basic information of enterprise investments, information on abnormal business operations, administrative penalty information, persons subject to execution for dishonesty, information on restrictions on high consumption, information on finalized cases, enterprise tax payment information, online public opinion information, etc., to build a complete enterprise portrait and present enterprise information in an all-round way.

[0104] (2) Application of knowledge graph technology: With the help of knowledge graph technology, based on certain enterprise association dimensions, such as investment relationships, guarantee relationships, equity relationships, etc., multi-source heterogeneous data is converted into the form of knowledge graphs to more accurately identify enterprises' illegal fundraising activities.

[0105] (3) Risk feature rule mining: a. Correlation comparison learning: By comparing known illegal fundraising companies with normal companies, we can find the risk features that distinguish illegal fundraising companies from normal companies. For example, illegal fundraising companies generally engage in false advertising and inflated registered capital. b. Anomaly detection: Based on a large amount of data, we can find abnormal information points that are significantly different from other companies. For example, if a company frequently changes its legal person, investors, senior executives, etc., or if its registered capital is increased significantly multiple times in a short period of time, there may be potential risks of abnormal business behavior.

[0106] (4) Functional Overview: This invention integrates multi-source government data to construct a panoramic portrait of the enterprise, and uses knowledge graph technology and risk feature rule mining methods to predict and score the risk of illegal fundraising by enterprises, providing a basis for decision-making for regulatory authorities, corporate partners, and investors.

[0107] It should be noted that not all steps and modules in the above-mentioned processes and device structures are required, and certain steps or modules can be omitted according to actual needs. The execution order of each step is not fixed and can be adjusted as needed. The system structure described in the above-mentioned embodiments can be a physical structure or a logical structure, that is, some modules may be implemented by the same physical entity, or some modules may be implemented by multiple physical entities, or may be implemented by certain components in multiple independent devices.

[0108] The above embodiments are merely preferred embodiments for the purpose of fully illustrating the present invention, and the scope of protection of the present invention is not limited thereto. Equivalent substitutions or modifications made by those skilled in the art based on the present invention are within the scope of protection of the present invention. The scope of protection of the present invention shall be subject to the claims.

Claims

1. A method for predicting the risk of abnormal corporate fundraising based on government data, characterized by: include: Step 1: Data collection and integration: Obtain multi-source data from various government departments through the data application platform. Conduct data integration: standardize the collected multi-source data and build a panoramic portrait of the enterprise. Step 2: Knowledge graph construction: Entity identification: Identify entities from the enterprise's panoramic portrait, including enterprises, legal persons, investors, directors, supervisors, and senior managers, and affiliated companies. Perform relationship extraction: extract the relationship between entities, including investment relationship, guarantee relationship, equity relationship, Generate knowledge graph: Represent entities and relationships in the form of a graph, use a graph database to store the knowledge graph data generated, and perform visual display. Step 3: Conduct risk feature rule mining: Conducting correlation comparison learning: Based on the machine learning decision tree algorithm, the data of known abnormal fundraising enterprises and normal enterprises in the existing data are compared to find the characteristic points of abnormal fundraising enterprises, classify them, and analyze the characteristic differences of different categories. Detect anomalies: Based on massive amounts of data, use statistical methods or machine learning algorithms to find abnormal information points that are significantly different from other companies. Generate risk feature rules: Generate risk feature rules based on the results of association comparison learning and anomaly detection. Step 4: Risk prediction and scoring: Extract features: Extract features related to the risk of abnormal fundraising from the company's panoramic portrait and knowledge graph. The relevant features include corporate identity information, business behavior, ethnic relations, and negative public opinion. Training model: Based on interpretability, the extracted features are trained using the logistic regression algorithm to build a risk prediction model. Risk prediction: Input the characteristic data of the enterprise to be predicted into the trained model to predict whether the enterprise has abnormal fundraising risks. Risk scoring: Based on risk feature rules and model prediction results, the enterprise is scored for risk. Step 5: Apply the results: Applied to risk warning: Real-time warning for high-risk enterprises, providing decision-making basis for regulatory authorities, Application to investment reference: Provide enterprise partners and investors with risk assessment results of abnormal fundraising for investment decision-making.

2. A method for predicting the risk of abnormal fundraising of enterprises based on government data according to claim 1, characterized in that when data collection is performed in step 1, multi-source data is obtained from various government departments through a data application platform, and the multi-source data includes market entity information, basic information of legal persons, basic information of directors, supervisors and senior managers, basic information of affiliated enterprises, paid-in / subscribed capital information of contributors / investors, social security participation information of enterprises, basic information of annual reports of enterprises, basic information of shareholders of enterprises, basic information of enterprise investments, list of abnormal operations, administrative penalty information, dishonest debtors, restrictions on high consumption, finalized case information, enterprise tax information, and online public opinion information.

3. According to claim 1, a method for predicting the risk of abnormal fundraising of enterprises based on government data is characterized in that standardized management is carried out in step 1, including noise removal and missing value processing in source data, unified data format and coding standards, association and fusion of data from different sources, formulation of indicator strategies and processing based on standardized data relationships, so as to construct a complete panoramic portrait of the enterprise.

4. A method for predicting the risk of abnormal fundraising of enterprises based on government data according to claim 1, characterized in that when performing outlier detection in step 3, based on massive data, the mean, variance, distribution analysis method in statistical methods or the isolation forest algorithm and local anomaly factor algorithm in machine learning algorithms are used to discover abnormal information points that are significantly different from other enterprises.

5. A device for predicting the risk of abnormal corporate fundraising based on government data, characterized by: It includes collection and integration module, knowledge graph construction module, rule mining module, risk prediction and scoring module and application module. The collection and integration module collects and integrates data: Obtain multi-source data from various government departments through the data application platform. Conduct data integration: standardize the collected multi-source data and build a panoramic portrait of the enterprise. The knowledge graph construction module constructs the knowledge graph: Entity identification: Identify entities from the enterprise's panoramic portrait, including enterprises, legal persons, investors, directors, supervisors, and senior managers, and affiliated companies. Perform relationship extraction: extract the relationship between entities, including investment relationship, guarantee relationship, equity relationship, Generate knowledge graph: Represent entities and relationships in the form of a graph, use a graph database to store the knowledge graph data generated, and perform visual display. The rule mining module mines risk feature rules: Conducting correlation comparison learning: Based on the machine learning decision tree algorithm, the data of known abnormal fundraising enterprises and normal enterprises in the existing data are compared to find the characteristic points of abnormal fundraising enterprises, classify them, and analyze the characteristic differences of different categories. Detect anomalies: Based on massive amounts of data, use statistical methods or machine learning algorithms to find abnormal information points that are significantly different from other companies. Generate risk feature rules: Generate risk feature rules based on the results of association comparison learning and anomaly detection. Risk Prediction and Scoring Module Risk Prediction and Scoring: Extract features: Extract features related to the risk of abnormal fundraising from the company's panoramic portrait and knowledge graph. The relevant features include corporate identity information, business behavior, ethnic relations, and negative public opinion. Training model: Based on interpretability, the extracted features are trained using the logistic regression algorithm to build a risk prediction model. Risk prediction: Input the characteristic data of the enterprise to be predicted into the trained model to predict whether the enterprise has abnormal fundraising risks. Risk scoring: Based on risk feature rules and model prediction results, the enterprise is scored for risk. Apply the module to apply the results: Applied to risk warning: Real-time warning for high-risk enterprises, providing decision-making basis for regulatory authorities, Application to investment reference: Provide enterprise partners and investors with risk assessment results of abnormal fundraising for investment decision-making.

6. The device for predicting the risk of abnormal fund-raising of enterprises based on government data according to claim 5 is characterized by: When the collection and integration module collects data, it obtains multi-source data from various government departments through the data application platform. The multi-source data includes market entity information, basic information of legal persons, basic information of directors, supervisors and senior managers, basic information of affiliated enterprises, paid-in / subscribed capital information of investors, social security participation information of enterprises, basic information of annual reports of enterprises, basic information of shareholders of enterprises, basic information of enterprise investments, information on the list of abnormal operations, administrative penalty information, persons subject to execution for dishonesty, restrictions on high consumption, information on finalized cases, enterprise tax information, and online public opinion information.

7. The device for predicting the risk of abnormal fund-raising of enterprises based on government data according to claim 5 is characterized by: The collection and integration module performs standardized governance, including noise removal and missing value processing in source data, unifying data formats and coding standards, associating and fusing data from different sources, formulating indicator strategies and processing based on standardized data relationships, and thus building a complete panoramic portrait of the enterprise.

8. The device for predicting the risk of abnormal fund-raising of enterprises based on government data according to claim 5 is characterized by: When the rule mining module detects anomalies, it uses statistical methods such as mean, variance, and distribution analysis, or the isolation forest algorithm and local anomaly factor algorithm in machine learning algorithms based on massive data to discover abnormal information points that are significantly different from other companies.

Images