Construction and query method of risk enterprise rule model based on interactive visual analysis

Through interactive visual analysis method, combined with hierarchical analysis method and SQL statement generation module, the problem of filtering rules-based risk assessment model in the big data environment and SQL statement generation is solved, and efficient screening and decision-making support for potential risk enterprises is achieved. .

CN117829600BActive Publication Date: 2024-11-08ZHEJIANG UNIV
View PDF 8 Cites 0 Cited by

Patent Information

Application Number
CN202311857292.6
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2023-12-29
Publication Date
2024-11-08
Estimated Expiration
2043-12-29

AI Technical Summary

Technical Problem

Rules-based legal risk assessment model is difficult to effectively screen key factors and build rule models under the problems of large data volume, high dimensions, information asymmetry and data reliability. The existing SQL automatic generation method lacks data construction for specific analysis tasks. Modal support, difficult to convert into standardized SQL query statements.

Method used

Using interactive visual analysis method, a preset hierarchical model is constructed through hierarchical analysis method, a criterion weight adaptive adjustment module and SQL statement generation module are built, and a visualization platform is provided, allowing analysts to adjust rule models and generate SQL statements, integrating data exploration and rules Model formulation and SQL statement generation.

Benefits of technology

It realizes efficient screening and SQL statement generation for potential risk enterprises, combines professional knowledge and computing intelligence to support more comprehensive decision-making tasks, simplifying the construction of rule models and the generation of SQL statements.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117829600B_ABST
    Figure CN117829600B_ABST
Patent Text Reader

Abstract

The present invention provides a risk enterprise rule model construction and query method based on interactive visual analysis, including the following steps: collecting enterprise-related data, pre-processing the data of each enterprise, and obtaining standardized data and criterion dimensions; based on the criterion dimensions, constructing a preset hierarchical model through hierarchical analysis method; constructing a hierarchical rule weight adaptive adjustment module and an SQL statement generation module; building a visual analysis platform integrating the criterion weight adaptive adjustment module and the SQL statement generation module, and providing an interactive interface; importing standardized data into the visual analysis platform, and inputting sample risk enterprises, adjusting the preset hierarchical model, and obtaining the adjusted risk enterprise hierarchical model and the SQL statement of the target enterprise set to be queried. The present invention can screen out a batch of potential risk enterprises of appropriate number based on a small number of enterprise lists with known risks through interaction with a visualization system, and simultaneously obtain SQL statements expressing the screening process.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of risk enterprise query management, and in particular to a risk enterprise rule model construction and query method based on interactive visual analysis. Background Art

[0002] Regional corporate risk assessment is a process of systematically analyzing and evaluating the various risks faced by companies in a specific region. Financial investment institutions, government agencies, companies or independent risk assessment agencies need to conduct corporate risk assessments to gain a deeper understanding of whether the company's activities comply with local, national and international regulations, and whether there is a possibility of facing fines, lawsuits or business cessation risks. Through this assessment, these institutions can more wisely formulate new investment plans or governance policies to reduce potential investment losses and help maintain social stability.

[0003] Data analysis has attracted much attention as a tool for assessing legal risks of regional enterprises. Rule-based risk assessment methods are favored by analysts. This method integrates domain expertise, which makes it perform well when sample data is scarce. At the same time, its model is easy to understand and helps institutions make subsequent decisions. However, considering the large amount of data, high data dimension, information asymmetry, data reliability and other issues, rule-based legal risk assessment models face a series of challenges in practice.

[0004] First, since the evaluation model is closely related to the analysis task, analysts need to work hard to mine common features from the many data dimensions of these samples to analyze the key factors that may lead to risks. In the case of large amounts of data and high data dimensions, analyzing the common features of samples and the differences between samples and the overall data is a time-consuming and complex task that requires analysts to invest a lot of time and energy. In addition, there may also be problems with the reliability and transparency of the data, which makes it more difficult to screen key factors and build rule models. This process relies on the analyst's domain knowledge and analysis experience, and automated data modeling methods are often difficult to completely replace.

[0005] Secondly, given that the ultimate goal of analysts is to encourage decision makers to make informed decisions, further requirements are reflected in the analyst's expectation to convert the quantitative evaluation results of the rule model into more standardized SQL query statements that are convenient for subsequent business personnel to query, count, transmit and save. This process usually requires effectively mapping the conditions and rules used in the rule model to the conditions, filtering and calculations of the SQL query statement. Machine-based SQL automatic generation methods mainly focus on assisting users to build reasonable SQL statements based on user input (such as natural language), but they often lack data modeling support for specific analysis tasks, and do not allow users to deeply explore the laws and forms of data. Summary of the invention

[0006] Based on the above background, the purpose of the present invention is to provide a risk enterprise rule model construction and query method based on interactive visual analysis, so that analysts can adjust the rule model based on a small list of enterprises with known risks through interaction with the visualization system, screen out a suitable number of potential risk enterprises, and obtain SQL statements expressing the screening process.

[0007] The present invention achieves the above technical objectives based on the following technical solutions:

[0008] A risk enterprise rule model construction and query method based on interactive visual analysis includes the following steps:

[0009] Collect enterprise-related data, pre-process the data of each enterprise, and obtain standardized data and standard dimensions;

[0010] Based on the criteria dimensions, a preset hierarchical model is constructed through the hierarchical analysis method;

[0011] Construct hierarchical rule weight adaptive adjustment module and SQL statement generation module;

[0012] Building a visual analysis platform integrating the criterion weight adaptive adjustment module and the SQL statement generation module, and providing an interactive interface;

[0013] The standardized data is imported into the visual analysis platform, and a sample venture enterprise is input to adjust the preset hierarchical model, and an SQL statement of the adjusted venture enterprise hierarchical model and the target enterprise set to be queried is obtained.

[0014] Furthermore, the collection of enterprise-related data and preprocessing of each enterprise's data to obtain standardized data and criteria dimensions include:

[0015] Collect relevant data of batches of companies of interest through web crawlers;

[0016] Interpolate missing values ​​for continuous data and fill in missing values ​​of discrete data with 0;

[0017] Data involving text, which is labeled, cleaned, and converted into numerical form;

[0018] Standardize the data to obtain standardized data;

[0019] Select n data dimensions that need to be used as criteria for the rule model to obtain criterion dimensions.

[0020] Furthermore, the construction of the preset hierarchical model includes:

[0021] Based on the criterion dimension, a three-layer preset hierarchical model structure including the target layer, category layer and criterion layer is constructed;

[0022] Determine an initial judgment matrix, perform a consistency check, obtain the initialization weights from the criterion layer to the target layer in the three-layer preset hierarchical model structure, and obtain a preset hierarchical model;

[0023] According to the initialization weights, the n data dimensions in the standardized data are weighted to calculate the initialization enterprise risk score of each enterprise.

[0024] Furthermore, the said construction hierarchical rule weight adaptive adjustment module comprises:

[0025] All criteria of the criterion layer are regarded as a sequence, and are sorted according to their initial weights to form an initial criterion sequence;

[0026] Constructing an initial analytic hierarchy process rule model, and reordering the initial criterion sequence to obtain a new criterion sequence;

[0027] According to the order of the criteria in the new criterion sequence, pairwise comparisons are performed to calculate the relative importance ratios between the criteria:

[0028]

[0029] According to the importance ratio between the two criteria, a comparison matrix is ​​obtained;

[0030] The consistency ratio is used to perform a consistency check on the comparison matrix. If the comparison matrix passes the consistency check, each column of the comparison matrix is ​​normalized, and the average value of each normalized column is used as the new weight of the initial hierarchical analysis method rule model to form a hierarchical rule weight adaptive adjustment module.

[0031] Furthermore, the SQL statement generation module is used to:

[0032] Discretize the distribution of each dimension of the standardized data;

[0033] Receive the target enterprise set to be queried and set the target function;

[0034] Solve the target function and output the SQL statement that matches the target enterprise set to be queried.

[0035] Furthermore, the discretization of the distribution of each dimension of the standardized data includes:

[0036] Set the number of divisions to m, and determine whether the data distribution is discrete and whether the number of values ​​is less than m. If the distribution is discrete and the number of values ​​is less than m, do not discretize, otherwise discretize;

[0037] Determine whether the probability of 0 value exceeds 50%. If it exceeds, 0 value is taken as a separate interval, and then the remaining data is divided into m equal parts according to the value size. The numerical interval of each part is the result of discretization of data distribution;

[0038] After the discretization, each dimension of the data obtains m basic query screening conditions, and the basic query conditions are subjected to Boolean operations to form composite query screening conditions.

[0039] Furthermore, the receiving of the target enterprise set to be queried and setting of the target function include:

[0040] Receive the target enterprise set to be queried, denoted as N o ;

[0041] Based on the Jaccard coefficient, the objective function L is set as:

[0042]

[0043] Wherein, o represents the query condition of the SQL statement for the target enterprise set to be queried. represents an approximation of o by minimizing get; Indicates that through The set of enterprises queried by the constructed SQL statement.

[0044] Furthermore, the solving of the objective function and outputting a SQL statement matching the target enterprise set to be queried include:

[0045] The basic query screening condition is stored in the query condition list, and the minimum objective function value is initialized to positive infinity;

[0046] The query filter conditions in the query filter list are combined in pairs through AND, OR, and DIF logical operations to obtain new composite query filter conditions;

[0047] According to the new composite query screening conditions, the corresponding enterprise set is found in the database, combined with the target enterprise set, the objective function value is calculated, and the objective function value is updated;

[0048] Add the first three compound query filter conditions with the smallest objective function value in this iteration to the query condition list;

[0049] The above steps are repeated until the minimum objective function value reaches a preset threshold, and the optimal screening condition of the target enterprise set is obtained. Based on the screening condition, an SQL statement matching the target enterprise set to be queried is constructed.

[0050] Furthermore, the construction of a visual analysis platform integrating the criterion weight adaptive adjustment module and the SQL statement generation module, and providing an interactive interface includes:

[0051] Integrate the criterion weight adaptive adjustment module and the SQL statement generation module on the backend server, and provide an interactive interface between the criterion weight adaptive adjustment module and the SQL statement generation module;

[0052] Establishing a visualization interface at the front end, and drawing an enterprise global visualization view, a hierarchical analysis method rule view, and an SQL query condition view on the visualization interface;

[0053] Establish a communication connection between the front-end and the back-end server to interact with data instructions.

[0054] Furthermore, the drawing of a global visualization view of the enterprise includes:

[0055] Use data dimensionality reduction methods to reduce n data dimensions to 2 dimensions;

[0056] Use enterprise scatter plots to provide an overview of enterprise data. The scatter plots show the results of dimensionality reduction of data for all enterprises, and provide circle selection tools on the scatter plots.

[0057] Based on the color of the scatter plot, the enterprise risk score calculated according to the weight of the AHP rule model and the change of the enterprise risk score after the weight adjustment are encoded;

[0058] The detailed information of the circled enterprise points is displayed through the enterprise details list, which provides a sample risk enterprise input box and can set the circled enterprise as a target enterprise set to be queried by the SQL statement generation module.

[0059] Furthermore, the drawing of the AHP rule view includes:

[0060] Provides visualization of the AHP rule model through the AHP rule view, showing the target layer, category layer, and criterion layer of the hierarchy;

[0061] Display the distribution diagram of each dimension of the data corresponding to the criterion layer in the AHP rule view, and based on the input sample risk enterprises, display the specific value of the sample in each dimension in the distribution diagram, and superimpose the distribution of the values ​​of the enterprises circled in the enterprise scatter diagram in the distribution diagram of each dimension;

[0062] A button for modifying the order of criteria in a preset hierarchical model is provided in the AHP rule view, and the new criterion sequence is transmitted to the criterion weight adaptive adjustment module in real time to adaptively increase or decrease the weight value.

[0063] Furthermore, the drawing of the SQL query condition view includes:

[0064] Call the interactive interface of the SQL statement generation module to generate corresponding SQL statements according to the set target enterprise set to be queried;

[0065] Visualize SQL statements, display basic query filter conditions in SQL, and the compound query filter conditions they constitute, and display the enterprises queried by the compound query filter conditions;

[0066] The SQL statement is fully displayed in the SQL query condition view for copying, running or saving.

[0067] Furthermore, the SQL statement for importing the standardized data into the visual analysis platform, inputting the sample venture enterprises, and obtaining the adjusted venture enterprise hierarchical model and the target enterprise set to be queried includes:

[0068] Import standardized data, reduce the data dimension through the visual analysis platform, and display the results in the enterprise scatter plot. The platform loads the preset hierarchical model;

[0069] Input sample venture enterprises, observe their clustering in the distribution bar chart, and decide whether to adjust the order of corresponding criteria in the preset hierarchical model;

[0070] Based on the color of the scatter plot, the change of the enterprise risk score after the adjustment of the AHP rule model is obtained, and the information of the enterprise individuals circled around the input sample point is compared in the enterprise details list to see whether it is similar to the input sample;

[0071] Based on the distribution curve of the selected enterprises in the bar chart, and according to the difference between the value distribution of the circled enterprises and the overall distribution, decide whether to adjust the order of the corresponding criteria in the preset hierarchical model again;

[0072] Repeat the above steps, based on the real-time changing weights, determine whether all important criteria have been set to weights greater than the preset values, and whether risky enterprises to be queried that meet the conditions have been set;

[0073] Construct the SQL statement for the risk enterprise to be queried, determine whether the query screening condition is appropriate based on the SQL query condition view, then output the query results and save the SQL statement.

[0074] Compared with the prior art, the present invention has the following beneficial effects:

[0075] The present invention provides a new interactive visual method for building a risk enterprise rule model and generating risk enterprise SQL query statements, integrating the three processes of data exploration, rule model formulation, and SQL statement generation. The method of the present invention allows professionals to explore enterprise data patterns, and combines the professional's domain knowledge with computational intelligence, adjusts the rule model for different analysis tasks, and efficiently screens out a batch of potential risk enterprises of appropriate number, while also obtaining SQL statements that express the screening process, providing more comprehensive support for subsequent decision-making tasks. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 It is a flow chart of an embodiment of the risk enterprise rule model construction and query method of the present invention.

[0077] Figure 2 It is a schematic diagram of a visual analysis platform established in an embodiment of the present invention.

[0078] Figure 3 It is a sunburst diagram of enterprise details and a bar diagram of dimensional distribution drawn in an embodiment of the present invention. Specific implementation plan

[0079] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0080] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0081] The present invention is based on a risk enterprise rule model construction and query method based on interactive visual analysis, comprising the following steps:

[0082] S1. Collect enterprise-related data, pre-process the data of each enterprise, and obtain standardized data and standard dimensions.

[0083] As a preferred embodiment, step S1 specifically includes:

[0084] S11. Collect relevant data of batches of interested enterprises through web crawlers;

[0085] S12, interpolate and supplement the missing values ​​of continuous data, and fill the missing values ​​of discrete data with 0;

[0086] S13. Data involving text, such as company descriptions, industries, and public opinion news, need to be labeled, cleaned, and converted into numerical form. The labeling process includes counting the number of times sensitive words appear in the text through sensitive word library matching. Public opinion news uses a pre-trained deep natural language model to divide the text emotions and then digitize them separately.

[0087] S14. Use the Quantile Transform method to map the data distribution to a standard normal distribution with a mean of 0 and a standard deviation of 1, so as to standardize the data and obtain standardized data;

[0088] S15. Select 25 data dimensions that need to be used as criteria for the rule model to obtain criterion dimensions.

[0089] S2. Based on the criteria dimension, a preset hierarchical model is constructed through the hierarchical analysis method.

[0090] As a preferred embodiment, step S2 specifically includes:

[0091] S21. Based on the criterion dimension, a three-layer preset hierarchical model structure including a target layer, a category layer and a criterion layer is constructed;

[0092] S22, determining an initial judgment matrix, and performing a consistency check, obtaining initialization weights from the criterion layer to the target layer in the three-layer preset hierarchical model structure, and obtaining a preset hierarchical model;

[0093] S23. According to the initialization weights, the 25 data dimensions in the standardized data are weighted to calculate the initialization enterprise risk score of each enterprise.

[0094] S3. Construct a hierarchical rule weight adaptive adjustment module and a SQL statement generation module.

[0095] As a preferred implementation scheme, in step S3, constructing a hierarchical rule weight adaptive adjustment module specifically includes:

[0096] S31, treating all criteria of the criterion layer as a sequence, and sorting them according to their initial weights to form an initial criterion sequence;

[0097] S32, constructing an initial analytic hierarchy process rule model, and reordering the initial criterion sequence to obtain a new criterion sequence;

[0098] S33. According to the order of the criteria in the new criterion sequence, perform pairwise comparisons and calculate the relative importance ratios between the criteria:

[0099]

[0100] S4. According to the importance ratio between the two criteria, a comparison matrix is ​​obtained. For example, the order of criteria 1, criteria 2, and criteria 3 is [0, 1, 2], then the comparison matrix is:

[0101]

[0102] S5. Use the consistency ratio to perform a consistency check on the comparison matrix. If the comparison matrix passes the consistency check, each column of the comparison matrix is ​​normalized, and the average value of each normalized column is used as the new weight of the initial hierarchical analysis method rule model (i.e., the AHP rule model) to form a hierarchical rule weight adaptive adjustment module.

[0103] As a preferred embodiment, in step S3, the constructed SQL statement generation module is used to:

[0104] Discretize the distribution of each dimension of the standardized data;

[0105] Receive the target enterprise set to be queried and set the target function;

[0106] Solve the target function and output the SQL statement that matches the target enterprise set to be queried.

[0107] In one illustrative example, discretizing the distribution of each dimension of the normalized data includes:

[0108] Set the number of divisions to m, and determine whether the data distribution is discrete and whether the number of values ​​is less than 5. If the distribution is discrete and the number of values ​​is less than 5, do not discretize, otherwise discretize;

[0109] Determine whether the probability of 0 value exceeds 50%. If it exceeds, 0 value is taken as a separate interval, and then the remaining data is divided into 5 equal parts according to the value size. The numerical interval of each part is the result of discretization of data distribution;

[0110] After discretization, each dimension of the data obtains a series of basic query filter conditions, such as 2<CompanyAge<5. These basic query conditions can be processed through Boolean operations to form composite query filter conditions. For example:

[0111] 2<CompanyAge<5AND NumberOfAdministrativePenalties>10,

[0112] Regardless of whether the query filter conditions are basic or complex, a set of qualified enterprises or an empty set can be found in the database through the SELECT statement.

[0113] In an example, receiving a target enterprise set to be queried, and setting the target function specifically includes:

[0114] Receive the target enterprise set to be queried, denoted as N o ;

[0115] Based on the Jaccard coefficient, the objective function L is set as:

[0116]

[0117] Wherein, o represents the query condition of the SQL statement for the target enterprise set to be queried. represents an approximation of o by minimizing get; Indicates that through The set of enterprises queried by the constructed SQL statement.

[0118] In an illustrated example, solving the target function and outputting an SQL statement that matches the target enterprise set to be queried includes:

[0119] Store the basic query screening conditions into the query condition list, and initialize the minimum objective function value to positive infinity;

[0120] The query filter conditions in the query condition list are combined in pairs through AND, OR, and SUB logical operations to obtain new composite query filter conditions;

[0121] According to the new composite query screening conditions, the corresponding enterprise set is found in the database, combined with the target enterprise set, the objective function value is calculated, and the objective function value is updated;

[0122] Add the first three compound query filter conditions with the smallest objective function value in this iteration to the query condition list;

[0123] The above steps are repeated until the minimum objective function value reaches a preset threshold, and the optimal screening condition of the target enterprise set is obtained. Based on the screening condition, an SQL statement matching the target enterprise set to be queried is constructed.

[0124] S4. Build a visual analysis platform that integrates the criterion weight adaptive adjustment module and the SQL statement generation module, and provide an interactive interface.

[0125] As a preferred embodiment, step S4 specifically includes:

[0126] S41, integrating the criterion weight adaptive adjustment module and the SQL statement generation module on the background server, and providing an interactive interface between the criterion weight adaptive adjustment module and the SQL statement generation module;

[0127] S42, establishing a visualization interface at the front end, and drawing an enterprise global visualization view, a hierarchy analysis method rule view (i.e., an AHP view), and an SQL query condition view on the visualization interface;

[0128] S43. Establish a communication connection between the front-end and the back-end server to interact with data instructions.

[0129] In one illustrative example, drawing a global visualization of an enterprise includes:

[0130] Use UMAP to reduce the 25 data dimensions to 2 dimensions;

[0131] The enterprise scatter plot is used to overview the enterprise data. The scatter plot shows the results of dimensionality reduction of the data of all enterprises. Red indicates positive growth and blue indicates negative growth. Figure 2 As shown in the middle circle A; and a circle selection tool is provided in the scatter plot;

[0132] Based on the color of the scatter plot, the enterprise risk score calculated according to the weight of the AHP rule model and the change of the enterprise risk score after the weight adjustment are encoded;

[0133] The detailed information of the selected enterprise points is displayed through the enterprise details list, such as Figure 2 As shown in the middle circles B and C. The list uses a sunburst chart to display enterprise information. Each sunburst chart represents an enterprise, showing the basic information of the enterprise and the data information of each dimension. The sunburst chart contains three layers of rings. Each sector of the outermost ring represents a dimension of the enterprise. The color depth of the sector indicates the size of the dimension value. The size of the angle of the sector encodes the size of the criterion weight of the hierarchical analysis method rule model; the second layer of rings indicates the risk category to which each criterion belongs; the innermost ring encodes the industry category of the enterprise's industrial and commercial registration information through color. The center of the sunburst chart provides the enterprise number, establishment time and registered capital, such as Figure 3 The enterprise details list provides a sample risk enterprise input box, and the circled enterprise can be set as a target enterprise set to be queried by the SQL statement generation module.

[0134] In one illustrative example, drawing an AHP rule view includes:

[0135] The AHP rule view provides visualization of the AHP rule model, showing the target layer, category layer, and criterion layer of the hierarchy, such as Figure 2 As shown in the middle circle D1;

[0136] The distribution details of each dimension of the data corresponding to the criterion layer are displayed in the hierarchical analysis rule view, and the criterion layer is connected to the corresponding dimension distribution through a solid line. The thickness of the solid line encodes the weight of the dimension, as shown in circle D2 in the figure. The distribution of each dimension is displayed through a bar chart. The x-axis of each bar chart represents the value of each dimension, and the y-axis represents the frequency of each value. After entering the sample venture enterprise, each bar chart will draw a red straight line perpendicular to the x-axis, indicating the specific value of the sample in this dimension. The distribution of the circled enterprises in each dimension from the enterprise scatter plot is then calculated using the kernel density estimation method to form a smooth curve and superimposed on the bar chart, as shown in the figure. Figure 3 As shown;

[0137] A button for modifying the order of criteria in a preset hierarchical model is provided in the AHP rule view, and the new criterion sequence is transmitted to the criterion weight adaptive adjustment module in real time to adaptively increase or decrease the weight value.

[0138] In one illustrative example, drawing a SQL query condition view includes:

[0139] Call the interactive interface of the SQL statement generation module to generate corresponding SQL statements according to the set target enterprise set to be queried;

[0140] Visualize SQL statements through tree structures, such as Figure 2 As shown in the middle circle E1. Leaf nodes represent basic query filter conditions, while non-leaf nodes represent compound query filter conditions. The three filter condition combinations AND / OR / SUB are coded in red, yellow, and blue and displayed on non-leaf nodes. Clicking a leaf node can highlight the corresponding dimension and filter value range in the distribution bar chart of the AHP rule view. Clicking a non-leaf node will display the companies found by the compound query filter conditions in the scatter plot.

[0141] In the SQL query condition view, the SQL statement is fully displayed for copying, running or saving. Figure 2 Shown in the middle circle E3.

[0142] S5. Importing standardized data into the visual analysis platform, inputting sample venture enterprises, adjusting the preset hierarchical model, and obtaining an SQL statement of the adjusted venture enterprise hierarchical model and the target enterprise set to be queried.

[0143] As a preferred embodiment, step S5 specifically includes:

[0144] S51, import standardized data, reduce the dimension of the data through the visual analysis platform, and display the results in the enterprise scatter plot. The platform loads the preset hierarchical model;

[0145] S52. Input sample venture enterprises, observe their aggregation in the distribution bar chart, and decide whether to adjust the order of corresponding criteria in the preset hierarchical model;

[0146] S53, based on the color of the scatter plot, obtain the change in the enterprise risk score after the adjustment of the AHP rule model, and compare in the enterprise details list whether the information of the enterprise individuals circled around the input sample point is similar to the input sample;

[0147] S54, based on the distribution curve of the selected enterprise in the bar chart, and according to the difference between the value distribution of the circled enterprise and the overall distribution, decide whether to adjust the order of the corresponding criteria in the preset hierarchical model again;

[0148] S54, repeating the above steps S52-S54, based on the real-time changing weights, determining whether all important criteria have been set to weights greater than the preset values, and whether risky enterprises to be queried that meet the conditions have been set;

[0149] S55. Click the Construct button to construct the SQL statement for the risk enterprise to be queried, determine whether the query filter condition is appropriate based on the SQL query condition view, and then output the query result and save the SQL statement.

[0150] The above embodiments are only used to help understand the method and core idea of ​​the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

Claims

1. A risk enterprise rule model construction and query method based on interactive visual analysis, characterized in that: The steps include: Collect enterprise-related data, pre-process the data of each enterprise, and obtain standardized data and standard dimensions; Based on the criteria dimensions, a preset hierarchical model is constructed through the hierarchical analysis method; Construct a criterion weight adaptive adjustment module and a SQL statement generation module, wherein the SQL statement generation module is used to: Discretize the distribution of each dimension of the standardized data; Receive the target enterprise set to be queried and set the target function, including: Receive the target enterprise set to be queried, denoted as N o ; Based on the Jaccard coefficient, the objective function L is set as: Wherein, o represents the query condition of the SQL statement for the target enterprise set to be queried. represents an approximation of o by minimizing get; Indicates that The set of enterprises queried by the constructed SQL statement; Solve the objective function and output SQL statements that match the target enterprise set to be queried, including: Store the basic query screening conditions into the query condition list, and initialize the minimum objective function value to positive infinity; The basic query filter conditions in the query condition list are combined in pairs through AND, OR, and DIF logical operations to obtain new composite query filter conditions; According to the new composite query screening conditions, the corresponding enterprise set is found in the database, combined with the target enterprise set, the objective function value is calculated, and the objective function value is updated; Add the first three compound query filter conditions with the smallest objective function value in this iteration to the query condition list; Repeat the above steps until the minimum objective function value reaches a preset threshold, and obtain the optimal screening condition of the target enterprise set, and form an SQL statement matching the target enterprise set to be queried based on the screening condition; Build a visual analysis platform that integrates the criterion weight adaptive adjustment module and the SQL statement generation module, and provide an interactive interface, specifically including: Integrate the criterion weight adaptive adjustment module and the SQL statement generation module on the backend server, and provide an interactive interface between the criterion weight adaptive adjustment module and the SQL statement generation module; A visualization interface is established at the front end, and a global visualization view of the enterprise, a hierarchical analysis method rule view, and an SQL query condition view are drawn on the visualization interface; wherein the drawing of the global visualization view of the enterprise includes: Use data dimensionality reduction methods to reduce n data dimensions to 2 dimensions; Use enterprise scatter plots to overview enterprise data. The enterprise scatter plots show the results of dimensionality reduction of data for all enterprises, and provide circle selection tools on the enterprise scatter plots. Based on the color of the enterprise scatter plot, the enterprise risk score calculated according to the weight of the AHP rule model and the change of the enterprise risk score after the weight adjustment are coded; The detailed information of the selected enterprise scatter points is displayed through the enterprise details list, the enterprise details list provides a sample risk enterprise input box, and the selected enterprise can be set as a target enterprise set to be queried by the SQL statement generation module; The drawing of the AHP rule view comprises: Provides visualization of the AHP rule model through the AHP rule view, showing the target layer, category layer, and criterion layer of the hierarchy; Display the distribution bar graphs of each dimension of the data corresponding to the criterion layer in the AHP rule view, and based on the input sample risk enterprises, display the specific values ​​of the sample in each dimension in the distribution bar graph, and superimpose the distribution of the values ​​of the enterprises circled in the enterprise scatter diagram in the distribution bar graph of each dimension; A button is provided in the AHP rule view to modify the order of criteria in the preset hierarchical model, and the new criterion sequence is transmitted to the criterion weight adaptive adjustment module in real time to adaptively increase or decrease the weight value; Establish a communication connection between the front-end and the back-end server to interact with data instructions; The standardized data is imported into the visual analysis platform, and a sample venture enterprise is input to adjust the preset hierarchical model to obtain an adjusted venture enterprise rule model and an SQL statement of a target enterprise set to be queried, including: Import standardized data, reduce the data dimension through the visual analysis platform, and display the results in the enterprise scatter plot. The platform loads the preset hierarchical model; Input sample venture enterprises, observe their clustering in the distribution bar chart, and decide whether to adjust the order of corresponding criteria in the preset hierarchical model; Based on the color of the enterprise scatter plot, the change of enterprise risk score after the adjustment of the AHP rule model is obtained, and the information of the enterprise individuals circled around the input sample point is compared in the enterprise details list to see whether it is similar to the input sample; Based on the distribution curve of the selected enterprise in the distribution bar chart, and according to the difference between the value distribution of the circled enterprise and the overall distribution, decide whether to adjust the order of the corresponding criteria in the preset hierarchical model again; Repeat the above steps, based on the real-time changing weights, determine whether all important criteria have been set to weights greater than the preset values, and whether risky enterprises to be queried that meet the conditions have been set; Construct the SQL statement for the risk enterprise to be queried, determine whether the query screening condition is appropriate based on the SQL query condition view, then output the query results and save the SQL statement.

2. The risk enterprise rule model construction and query method based on interactive visual analysis according to claim 1 is characterized in that: The aforementioned collection of enterprise-related data, pre-processing of each enterprise's data, and obtaining standardized data and criteria dimensions include: Collect relevant data of batches of companies of interest through web crawlers; Interpolate missing values ​​for continuous data and fill in missing values ​​of discrete data with 0; Data involving text, which is labeled, cleaned, and converted into numerical form; Standardize the data to obtain standardized data; Select n data dimensions that need to be used as criteria for the rule model to obtain criterion dimensions.

3. The risk enterprise rule model construction and query method based on interactive visual analysis according to claim 2 is characterized in that: The construction of the preset hierarchical model includes: Based on the criterion dimension, a three-layer preset hierarchical model structure including the target layer, category layer and criterion layer is constructed; Determine an initial judgment matrix, perform a consistency check, obtain the initialization weights from the criterion layer to the target layer in the three-layer preset hierarchical model structure, and obtain a preset hierarchical model; According to the initialization weights, the n data dimensions in the standardized data are weighted to calculate the initialization enterprise risk score of each enterprise.

4. The risk enterprise rule model construction and query method based on interactive visual analysis according to claim 3 is characterized in that: The construction criterion weight adaptive adjustment module includes: All criteria of the criterion layer are regarded as a sequence, and are sorted according to their initial weights to form an initial criterion sequence; Constructing an initial analytic hierarchy process rule model, and reordering the initial criterion sequence to obtain a new criterion sequence; According to the order of the criteria in the new criterion sequence, pairwise comparisons are performed to calculate the relative importance ratios between the criteria: According to the importance ratio between the two criteria, a comparison matrix is ​​obtained; The consistency ratio is used to perform a consistency check on the comparison matrix. If the comparison matrix passes the consistency check, each column of the comparison matrix is ​​normalized, and the average value of each normalized column is used as a new weight of the initial hierarchical analysis method rule model to form a criterion weight adaptive adjustment module.

5. The risk enterprise rule model construction and query method based on interactive visual analysis according to claim 1 is characterized in that: The discretization of the distribution of each dimension of the standardized data includes: Set the number of divisions to m, and determine whether the data distribution is discrete and whether the number of values ​​is less than m. If the distribution is discrete and the number of values ​​is less than m, do not discretize, otherwise discretize; Determine whether the probability of 0 value exceeds 50%. If it exceeds, 0 value is taken as a separate interval, and then the remaining data is divided into m equal parts according to the value size. The numerical interval of each part is the result of discretization of data distribution; After the discretization, each dimension of the data obtains m basic query screening conditions, and the basic query screening conditions are subjected to Boolean operations to form composite query screening conditions.

6. The risk enterprise rule model construction and query method based on interactive visual analysis according to claim 1 is characterized in that: The drawing of the SQL query condition view includes: Call the interactive interface of the SQL statement generation module to generate corresponding SQL statements according to the set target enterprise set to be queried; Visualize the SQL statement, display the basic query screening conditions in the SQL, and the compound query screening conditions formed by the basic query screening conditions, and display the enterprises queried by the compound query screening conditions; The SQL statement is fully displayed in the SQL query condition view for copying, running or saving.

Citation Information

Patent Citations

  • Financial online supervision system and method based on big data

    CN110400207A

  • Method for automatically generating SQL statements based on dimension model

    CN112100200A

  • Enterprise fund risk assessment method based on analytic hierarchy process and cloud model

    CN114881485A

  • Enterprise risk assessment method based on nonlinear dimension reduction, computer equipment and storage medium

    CN115131039A

  • Traditional Chinese medicine ancient book visual analysis method and system based on knowledge graph

    CN116340544A