Financial risk control method based on enterprise finance and tax data

By receiving and analyzing financial and tax data, performing multi-dimensional feature derivatives and industry classification, combining chi-square binning and logistic regression models, a score card model is constructed, which solves the problems of large consumption and low accuracy of financial and tax data processing resources in the existing technology, and realizes the More accurate financial risk assessment and consistency between industry score comparisons.

CN118967345APending Publication Date: 2024-11-15NANJING BINGJIAN INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411143098.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-15

AI Technical Summary

Technical Problem

The prior art consumes a lot of resources and has low accuracy when processing large-scale fiscal and tax data, and fails to train machine learning models from multiple dimensions, resulting in incomplete risk assessment and no industry characteristics considerations, resulting in model accuracy. reduce.

Method used

Design a financial risk control method based on corporate financial and taxation data, receive structured data through the protocol interface, perform data analysis and feature derivation, perform feature analysis according to industry classification, and use chi-square binning and logistic regression models to build industry-specific scoring cards The model and combine business logic screening characteristics to ensure that the model has a relatively consistent rating in different industries.

Benefits of technology

It has achieved comprehensive exploration of financial risks in fiscal and taxation data from multiple dimensions, improved the accuracy of the model and resource utilization efficiency, ensured the score consistency of different industries, and more accurately identified enterprises with serious risks.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967345A_ABST
    Figure CN118967345A_ABST
Patent Text Reader

Abstract

According to the financial risk control method based on enterprise finance and taxation data, structured data related to tax and finance from an enterprise is received through a protocol interface, and then data analysis is performed on the structured data. In the aspect of model construction, the model is not trained according to industries and dimensions for different industry characteristics in the prior art, a sample is divided into 10 different industries, and financial risks in finance and tax data are comprehensively mined from 6 different dimensions. In the process of constructing the model, rule features and in-model features are screened out by combining an IV graph trend and business experience, and enterprises with serious risks are identified by the rule features in a mode of setting subtraction. And finally, the score level intervals of the score card models of different industries are pulled to the same level, so that final score comparison of different industries is facilitated, and financial risks existing in enterprise finance and tax data are comprehensively mined.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the field of financial and taxation control, and specifically relates to a financial risk control method based on corporate financial and taxation data. Background Art

[0002] Nowadays, the technology of obtaining financial statistics by identifying and extracting data from financial and tax report files has been very mature. However, this technology will consume a lot of modeling time and computer resources when facing large amounts of data, and the obtained statistical data will have certain errors compared to the real data. In addition, the data source of the existing corporate financial and tax risk control model is obtained by identifying and extracting tax invoice report files to obtain the declaration data for corporate taxation in the report files. This part of the data generally has to wait until the report is made public to obtain it, so there is a certain lag and the accuracy is low. The reason is that in the process of building the risk control model, features are not derived from multiple dimensions for training the machine learning model, so it is easy to lose features of certain dimensions when screening the features for modeling, and ultimately the model cannot assess risks from a more comprehensive perspective. In addition, the above technology does not take into account the differences in industry characteristics in the model. When all samples are evaluated with the same model, problems such as reduced model accuracy caused by industry differences will occur.

[0003] Therefore, a machine learning model that performs feature screening in multiple dimensions and considers the characteristics of different industries is needed to more comprehensively assess the financial risks in corporate financial and tax data. Summary of the invention

[0004] In order to solve the above technical problems, this application designs a financial risk control method based on corporate financial and tax data, in order to build a model that combines machine learning and business experience from the whole process, so as to comprehensively explore the financial risks existing in corporate financial and tax data.

[0005] A financial risk control method based on corporate financial and tax data, comprising:

[0006] Step S1: receiving financial and tax structured data from an enterprise through a protocol interface; and performing data analysis on the financial and tax structured data;

[0007] Step S2: derive features from the parsed data to obtain feature data derived from the basic information dimension, derived feature data from the business scale dimension, derived feature data from the business stability dimension, derived feature data from the business growth dimension, derived feature data from the contract fulfillment history dimension, and derived feature data from the financial information dimension;

[0008] Step S3: According to the national economic industry classification, the sample data is divided into agriculture, forestry, animal husbandry, fishery, manufacturing & mining, electricity, heat, gas and water production and supply, transportation, warehousing and postal services, accommodation and catering, wholesale and retail, information transmission & scientific research, construction, real estate, and others, a total of 10 industries, and the financial and tax structured data of each industry is obtained as sample data, and feature analysis is performed by industry, and sample feature data is derived from six dimensions: basic information dimension, business scale dimension, business stability dimension, business growth dimension, performance history dimension, and financial information dimension; and the corresponding label y is spliced ​​to construct an industry-specific training set;

[0009] Step S4: group the derived sample characteristic data using the chi-square binning method, calculate the IV value, draw an IV graph, and calculate the bad debt rate corresponding to each bin;

[0010] The derived sample characteristic data with a small number of bins and large differences in bad debt rates corresponding to each bin are selected as the pre-rules of the logistic regression model. The threshold settings of these rules are determined according to the thresholds corresponding to the bins of each industry.

[0011] Step S5: Based on the derived sample feature data of the 10 industries, combined with the IV diagram and business logic, the input features are gradually screened from the remaining derived sample feature data, a logistic regression model is constructed, and converted into a scorecard model for each industry;

[0012] Step S6: Use the scoring card model of each industry to score the samples of each industry, and perform model evaluation and adjustment based on the model score to ensure that the average score of samples that hit the rule is lower than the average score of samples that do not hit the rule, and obtain the final model of each industry;

[0013] Step S7: Divide the final model evaluation scores of each industry into five levels: A, B, C, D, and E according to the same sample proportions, and align the score thresholds corresponding to the five levels by adjusting the scoring card models of each industry to facilitate the final score comparison of different industries.

[0014] Preferably, the financial and tax structured data include: taxpayer status, enterprise registration time, legal representative information, enterprise tax return form, tax arrears information form, enterprise balance sheet, income statement, cash flow statement and other financial and tax data.

[0015] Preferably, the label y includes positive examples and negative examples.

[0016] Preferably, the IV value is calculated as follows:

[0017]

[0018] Where yi is the number of positive examples in the current group, ys is the number of global positive examples, ni is the number of negative examples in the current group, and ns is the number of global negative examples.

[0019] Preferably, the bad debt rate br=n i / n s .

[0020] Preferably, the IV diagram is drawn by using the group number as the horizontal axis and the quantity and bad debt rate of each group as the double vertical axes.

[0021] Preferably, in step S5, the method for gradually screening the input features from the remaining derived sample feature data is: combining IV value and feature trend to perform feature screening in the binary classification model, and screening out the top 20 derived sample features with IV value greater than 0.2 and better feature trend as input features.

[0022] The advantages and effects of this application are as follows:

[0023] The present application designs a financial risk control method based on corporate fiscal and taxation data, which receives structured data related to taxation and finance from the enterprise through a protocol interface, and then performs data analysis on the structured data. In terms of model construction, the prior art does not train models by industry and dimension based on the characteristics of different industries. The present invention divides the samples into 10 different industries, and comprehensively mines the financial risks in fiscal and taxation data from 6 different dimensions. In the process of model construction, the present invention combines the IV graph trend with business experience to screen out rule features and model entry features. The rule features identify companies with serious risks by setting a deduction method. Finally, the score level intervals of the scoring card models of different industries are aligned to the same level, which is convenient for the final score comparison of different industries, thereby comprehensively mining the financial risks existing in corporate fiscal and taxation data.

[0024] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application so that it can be implemented in accordance with the contents of the specification, and to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the following is a detailed description of the preferred embodiments of the present application in conjunction with the accompanying drawings as follows.

[0025] Based on the detailed description of the specific embodiments of the present application in combination with the accompanying drawings below, those skilled in the art will become more aware of the above and other objects, advantages and features of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings required for use in the embodiments or the prior art descriptions are briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can also be obtained based on these drawings without creative work. In all drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, each element or part is not necessarily drawn according to the actual scale.

[0027] Figure 1 A flowchart of a financial risk control method based on corporate financial and tax data designed for this application. DETAILED DESCRIPTION

[0028] To make the purpose, technical scheme and advantages of the embodiment of the present application clearer, the technical scheme in the embodiment of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiment of the present application. Obviously, the described embodiment is a part of the embodiment of the present application, rather than all of the embodiments. In the following description, specific details such as specific configuration and components are provided only to help fully understand the embodiments of the present application. Therefore, it should be clear to those skilled in the art that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the present application. In addition, for clarity and brevity, the description of known functions and structures is omitted in the embodiment.

[0029] It should be understood that the references to "one embodiment" or "this embodiment" throughout the specification mean that the specific features, structures, or characteristics associated with the embodiment are included in at least one embodiment of the present application. Therefore, the references to "one embodiment" or "this embodiment" appearing throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

[0030] In addition, the present application may repeat reference numerals and / or letters in different examples. This repetition is for the purpose of simplicity and clarity, and does not in itself indicate the relationship between the various embodiments and / or settings discussed.

[0031] The term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, B exists alone, and A and B exist at the same time. The term " / and" in this article describes another type of association object relationship, indicating that there can be two relationships. For example, A / and B can mean: A exists alone, and A and B exist alone. In addition, the character " / " in this article generally indicates that the previous and next associated objects are in an "or" relationship.

[0032] The term "at least one" in this article is merely a description of the association relationship of associated objects, indicating that there may be three relationships. For example, at least one of A and B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0033] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusions.

[0034] Example

[0035] This embodiment mainly introduces the specific design of a financial risk control method based on corporate financial and tax data.

[0036] Please refer to Figure 1 , a financial risk control method based on corporate financial and tax data, including:

[0037] Step S1, receiving financial and tax structured data from the enterprise through the protocol interface; performing data analysis on the financial and tax structured data; analyzing the financial and tax data of the enterprise in the protocol interface will obtain basic information such as taxpayer status, enterprise registration time, legal representative information, etc., and will also obtain specific financial and tax data such as enterprise tax return form, tax arrears information form, enterprise balance sheet, income statement, cash flow statement, etc. From these different report information, the derivative logic of each feature can be determined from the perspectives of the number of changes in financial and tax data over the years, growth rate, ratio, etc., such as: the sales revenue growth rate of value-added tax declaration in the past year, the coefficient of variation of sales revenue of value-added tax declaration in the past year, the number of zero declarations of value-added tax taxable sales revenue in the past year...etc.

[0038] Step S2, then use Python programming language to write derivative logic code, and derive features from six dimensions of basic information, business scale, business stability, business growth, contract performance history, and financial information for the parsed data. These six dimensions can fully cover the information dimensions existing in the fiscal and taxation data, and then obtain derived feature data from the basic information dimension, derived feature data from the business scale dimension, derived feature data from the business stability dimension, derived feature data from the business growth dimension, derived feature data from the contract performance history dimension, and derived feature data from the financial information dimension;

[0039] Step S3: According to the national economic industry classification, the sample data is divided into agriculture, forestry, animal husbandry, fishery, manufacturing & mining, electricity, heat, gas and water production and supply, transportation, warehousing and postal services, accommodation and catering, wholesale and retail, information transmission & scientific research, construction, real estate, and others, a total of 10 industries, and the financial and tax structured data of each industry is obtained as sample data, and feature analysis is performed by industry, and sample feature data is derived from six dimensions: basic information dimension, business scale dimension, business stability dimension, business growth dimension, performance history dimension, and financial information dimension; and the corresponding label y is spliced ​​to construct an industry-specific training set;

[0040] Step S4: group the derived sample characteristic data using the chi-square binning method, calculate the IV value, draw an IV graph, and calculate the bad debt rate corresponding to each bin;

[0041] Furthermore, the chi-square binning is a commonly used feature discretization method used to convert continuous variables into ordered discrete variables. It is based on the idea of ​​the chi-square test and determines the optimal binning boundary by comparing the difference between the observed frequency and the expected frequency. This method can effectively reduce the impact of outliers on the model.

[0042] Furthermore, the chi-square binning includes the following key steps:

[0043] 1. Divide the continuous variable into initial binning intervals according to its value range;

[0044] 2. Count the number of samples in each bin interval to form the initial observation frequency matrix;

[0045] 3. According to the distribution of the overall sample, calculate the expected frequency of each bin interval for subsequent chi-square test;

[0046] 4. Use the chi-square test to measure the difference between the observed frequency and the expected frequency, calculate the chi-square statistic, and measure the goodness of fit of the overall model;

[0047] 5. Consider adjacent bins one by one from left to right, and calculate the chi-square statistic after merging. If the merger can significantly reduce the chi-square statistic, merge the two bins into a new bin;

[0048] 6. Repeat the merging operation of adjacent bins until the chi-square statistic cannot be further reduced;

[0049] 7. Determine the conditions for stopping merging based on business needs and model performance, such as the number of bins, minimum number of samples, etc.

[0050] Furthermore, the final binning boundary is obtained according to the merging result to form the discretized features.

[0051] IV value is used to evaluate the encoding and prediction ability of features. In each bin, the ratio of the current positive example yi to the global positive example ys and the ratio of the current group negative example ni to the global negative example ns are calculated, and then the bad debt rate br = n i / n s It refers to the ratio of bad samples (such as default, arrears, etc.) observed in each bin to the total number of samples, that is, the ratio of the current group counterexamples ni to the global counterexamples ns.

[0052]

[0053] Where yi is the number of positive examples in the current group, ys is the number of global positive examples, ni is the number of negative examples in the current group, and ns is the number of global negative examples.

[0054] Furthermore, the IV diagram is drawn in the following way: the group number is used as the horizontal axis, and the number and bad debt rate of each group are used as the double vertical axes. The IV diagram can intuitively represent the trend of the proportion of counterexamples corresponding to the change of the feature value. The feature trend here refers to the degree of correlation between the change trend and direction between the feature and the prediction target. In order to measure this degree of correlation, the correlation coefficient can be used as a statistical indicator. If the change trend of the feature is consistent with the change trend of the prediction target, and the correlation coefficient is high (close to 1), it indicates that the trend of the feature is good. If the change trend of the feature is opposite to the change trend of the prediction target, and the correlation coefficient is low (close to -1), it also indicates that the trend is good. However, if the change trend of the prediction target is consistent with the change trend of the feature in some feature value intervals, and opposite in other intervals, the correlation is low and the trend is poor.

[0055] Then, the derived sample feature data with a small number of bins and large differences in bad debt rates corresponding to each bin are selected as the pre-rules of the logistic regression model. The threshold settings of these rules are determined according to the thresholds corresponding to the bins of each industry.

[0056] After obtaining the binning results and IV diagrams of all features, the features with fewer bins and larger differences in bad debt rates between bins are selected as the rule features of the model. The number of bins for each rule feature is adjusted and merged to be within 3 (including null values), and the corresponding scores are all deduction items, so that the final scores of enterprises with greater financial risks fall into the lower-middle range. In the process of setting the rule deduction items, the Pearson correlation coefficient is first used to measure the correlation between features. The Pearson correlation coefficient is a statistical indicator used to measure the strength and direction of the linear relationship between two continuous variables, and its value range is between -1 and 1. When the coefficient is 1, it indicates a complete positive correlation; when the coefficient is -1, it indicates a complete negative correlation; when the coefficient is close to 0, it indicates that there is no linear relationship between the two variables. We screen out the features with an absolute value of the correlation coefficient between two features greater than 0.5 and classify them into a rule feature group. The features with an absolute value of the correlation coefficient less than 0.5 are each a rule feature group. The deduction weight of each rule feature group is the same, and the deduction weight within the same feature group is evenly distributed. This will ensure that during the rule-based scoring process, companies that hit the same type of rules will not lose too many points, thereby reducing the accuracy of the model.

[0057] Step S5: Based on the derived sample feature data of the 10 industries, combined with the IV diagram and business logic, the input features are gradually screened from the remaining derived sample feature data, a logistic regression model is constructed, and converted into a scorecard model for each industry;

[0058] Furthermore, the remaining features of these 10 industries are gradually screened for modeling features, mainly combining IV values ​​and feature trends to screen features in the binary classification model, and the top 20 features with IV values ​​greater than 0.2 and good feature trends are selected as modeling features. Use these modeling features to build a logistic regression model, and then convert the coefficients in the logistic regression model into scores in the scorecard, usually using WOE (Weight of Evidence) and IV (Information Value) for conversion. WOE conversion: WOE is used to measure the degree of influence of feature values ​​on the target variable. First, the continuous features are discretized into a set of different bins, each bin representing a specific value interval. Then, the good-bad ratio of the target variable in each bin is calculated (the ratio of the number of good customers to the number of bad customers). Then, the WOE of each bin is calculated, that is, the difference of the good-bad ratio of the natural logarithm. WOE represents the degree of influence of feature values ​​on the target variable. A positive value indicates a positive impact on good customers, and a negative value indicates a positive impact on bad customers. IV conversion: IV is a measure of the amount of information of a feature in a prediction model. By weighting the WOE values ​​of all bins, the information value (IV) of the feature can be calculated. The higher the IV, the stronger the predictive power of the feature for the target variable.

[0059] Step S6: Use the scoring card model of each industry to score the samples of each industry, and perform model evaluation and adjustment based on the model score to ensure that the average score of samples that hit the rule is lower than the average score of samples that do not hit the rule, and obtain the final model of each industry;

[0060] Furthermore, the score of each feature is calculated based on the converted coefficients and feature values. Generally speaking, the scorecard model uses a linear weighted method to calculate the score. For each feature, the WOE corresponding to its value is multiplied by the corresponding coefficient, and then the intercept term of the model is added to finally obtain the score of each feature. This is the first version of the scorecard model. We then make appropriate adjustments to the binning thresholds and scores of the scorecard based on business experience to make the model more consistent with the risk assessment logic of the credit risk binary classification scenario.

[0061] Step S7: Divide the final model evaluation scores of each industry into five levels: A, B, C, D, and E according to the same sample proportions, and align the score thresholds corresponding to the five levels by adjusting the scoring card models of each industry to facilitate the final score comparison of different industries.

[0062] In step S5, the method for gradually screening the input features from the remaining derived sample feature data is: combining IV value and feature trend to perform feature screening in the binary classification model, and screening out the top 20 derived sample features with IV value greater than 0.2 and good feature trend as input features.

[0063] Furthermore, the scorecard models and pre-rules of the 10 industries are combined to score and evaluate their test sets, and the final test set scores of each industry are divided into five levels: A, B, C, D, and E. The number of samples of each level is divided according to a certain proportion of the total number of samples in the test set. For example, the A-level samples account for 10%, the B-level samples account for 25%, the C-level samples account for 30%, the D-level samples account for 25%, and the E-level samples account for 10%. In this way, the score interval corresponding to each level can be obtained. Finally, the binning thresholds corresponding to the scorecards of different industries at the same level in the test set scores are compared, and the scorecards of each industry are manually adjusted to make the score intervals corresponding to the same level close. This is to bring the scorecard evaluation capabilities of different industries to the same level, which enables the scorecard models of different industries to be compared under the same standard to ensure that enterprises of the same level in different industries have similar credit risk levels, so as to carry out unified risk management and decision-making.

[0064] The present invention receives the actual financial and tax data from the enterprise through the access protocol interface. The construction of the entire model is based on more accurate and reliable data, so the entire model is more convincing.

[0065] The present invention derives features from six subdivision dimensions respectively, so as to ensure the comprehensiveness of the dimensions in the process of feature screening, thereby being able to comprehensively mine the financial risks existing in the fiscal and taxation data.

[0066] The above description is only the preferred embodiment of the present invention, which does not limit the protection scope of the present invention. For those skilled in the art, the present invention can be modified and varied in various ways. Within the spirit and principle of the present invention, any change, modification, replacement, integration and parameter change of these embodiments by conventional substitution or capable of achieving the same function without departing from the principle and spirit of the present invention shall fall within the protection scope of the present invention.

Claims

1. A financial risk control method based on corporate financial and tax data, characterized in that: include: Step S1: receiving financial and tax structured data from the enterprise through a protocol interface; Performing data analysis on the financial and tax structured data; Step S2: derive features from the parsed data to obtain feature data derived from the basic information dimension, derived feature data from the business scale dimension, derived feature data from the business stability dimension, derived feature data from the business growth dimension, derived feature data from the contract fulfillment history dimension, and derived feature data from the financial information dimension; Step S3: According to the national economic industry classification, the sample data is divided into agriculture, forestry, animal husbandry, fishery, manufacturing & mining, electricity, heat, gas and water production and supply, transportation, warehousing and postal services, accommodation and catering, wholesale and retail, information transmission & scientific research, construction, real estate, and others, a total of 10 industries, and the financial and tax structured data of each industry is obtained as sample data, and feature analysis is performed by industry, and sample feature data is derived from six dimensions: basic information dimension, business scale dimension, business stability dimension, business growth dimension, performance history dimension, and financial information dimension; and the corresponding label y is spliced ​​to construct an industry-specific training set; Step S4: group the derived sample characteristic data using the chi-square binning method, calculate the IV value, draw an IV graph, and calculate the bad debt rate corresponding to each bin; The derived sample characteristic data with a small number of bins and large differences in bad debt rates corresponding to each bin are selected as the pre-rules of the logistic regression model. The threshold settings of these rules are determined according to the thresholds corresponding to the bins of each industry. Step S5: Based on the derived sample feature data of the 10 industries, combined with the IV diagram and business logic, the input features are gradually screened from the remaining derived sample feature data, a logistic regression model is constructed, and converted into a scorecard model for each industry; Step S6: Use the scoring card model of each industry to score the samples of each industry, and perform model evaluation and adjustment based on the model score to ensure that the average score of samples that hit the rule is lower than the average score of samples that do not hit the rule, and obtain the final model of each industry; Step S7: Divide the final model evaluation scores of each industry into five levels: A, B, C, D, and E according to the same sample proportions, and align the score thresholds corresponding to the five levels by adjusting the scoring card models of each industry to facilitate the final score comparison of different industries.

2. A financial risk control method based on corporate tax data according to claim 1, characterized in that: The financial and tax structured data include: taxpayer status, enterprise registration time, legal representative information, enterprise tax return form, tax arrears information form, enterprise balance sheet, income statement, cash flow statement and other financial and tax data.

3. A financial risk control method based on corporate tax data according to claim 1, characterized in that: The label y includes positive examples and negative examples.

4. A financial risk control method based on corporate tax data according to claim 1, characterized in that: The IV value is calculated as follows: Where yi is the number of positive examples in the current group, ys is the number of global positive examples, ni is the number of negative examples in the current group, and ns is the number of global negative examples.

5. The financial risk control method based on corporate tax data according to claim 1 is characterized in that: The bad debt rate br = n i / n s .

6. A financial risk control method based on corporate tax data according to claim 1, characterized in that: The IV diagram is drawn in the following way: the group number is used as the horizontal axis, and the number and bad debt rate of each group are used as the double vertical axes.

7. The financial risk control method based on corporate tax data according to claim 1 is characterized in that: In step S5, the method for gradually screening the input features from the remaining derived sample feature data is: combining IV value and feature trend to perform feature screening in the binary classification model, and screening out the top 20 derived sample features with IV value greater than 0.2 and good feature trend as input features.