Finance and tax risk processing method based on large language model

Through a large language model, analyzing financial data and historical trends, combining expert suggestions, a fiscal and tax risk processing report is generated, which solves the problems of low efficiency and low intelligence in the existing technology, and realizes the quantitative and efficient processing of risk trends.

CN118898522AInactive Publication Date: 2024-11-05WEIHAI HAITONGGU ELECTRONIC TECHNOLOGY CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202410938097.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-12
Publication Date
2024-11-05
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing fiscal and tax risk treatment methods rely on manual analysis and are inefficient, difficult to identify potential risk trends, and are low in intelligence, so they cannot provide comprehensive risk situation analysis and targeted processing.

Method used

Using a financial and tax risk treatment method based on a large language model, we can collect and preprocess corporate financial data, generate risk warnings, analyze historical data based on risk warning types, calculate change trend index, integrate it with expert suggestions, generate final risk treatment reports, and improve Risk assessment and processing efficiency.

Benefits of technology

It has achieved the quantification of risk trends, helping enterprises consider potential risks in long-term strategic planning, improve adaptability and resilience, and provide high-quality and efficient consulting services by evaluating the professionalism and efficiency of experts, improving the efficiency of risk processing and quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118898522A_ABST
    Figure CN118898522A_ABST
Patent Text Reader

Abstract

The invention discloses a large language model-based finance and taxation risk processing method, and relates to the technical field of finance and taxation risk processing. According to the invention, after the enterprise generates the corresponding risk prompt, the historical data of the risk prompt type in the set time period is extracted and further analyzed to obtain the change trend index of the risk prompt type corresponding to the current time point of the enterprise, and the quantification of the risk trend can be realized according to the obtained change trend index. The method helps an enterprise to consider potential risks in long-term strategic planning, improves the adaptability and toughness of the enterprise, provides more comprehensive data for finance and tax risk processing, and solves the problem that in the prior art, corresponding risk processing opinions are output by inputting corresponding risk prompt keywords into an established large language model, so that the risk processing efficiency is improved. And the problem of one-sided treatment is solved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of financial and tax risk processing, and in particular to a financial and tax risk processing method based on a large language model. Background Art

[0002] With the rapid development of the economy, the financial and tax risks faced by enterprises are becoming increasingly complex; traditional financial and tax risk management methods rely on manual analysis, which is inefficient and difficult to fully identify potential risks; in recent years, the development of artificial intelligence technology, especially large language models, has provided new possibilities for the automated processing of financial and tax risks.

[0003] However, the existing financial and tax risk management methods still have the following defects: Most of them output corresponding risk handling opinions by inputting corresponding risk warning keywords into the established large language model. The processing is relatively one-sided. It cannot analyze the historical data of the risk warning type of the enterprise within the set time period based on the generation time point of the current risk warning, and evaluate the changing trend and potential risks of the enterprise for this risk warning type. The degree of intelligence is low, and it cannot help enterprises quickly understand the overall situation of risks and carry out targeted processing.

[0004] To this end, a financial and tax risk management method based on a large language model is introduced. Summary of the invention

[0005] In view of this, the present invention provides a financial and tax risk management method based on a large language model to solve the problems raised by the above background technology.

[0006] The purpose of the present invention can be achieved through the following technical solution: A method for handling financial and tax risks based on a large language model, comprising: Financial and tax risk analysis: collect relevant financial data of the enterprise and pre-process it; integrate the relevant financial data after pre-processing to construct the financial and tax data set of the enterprise at the current time point; set a series of financial normal indicators based on the actual situation of the current enterprise; extract the financial ratios in the financial data set of the enterprise at the current time point, and compare them with the corresponding financial normal indicators set; if the comparison result shows that the corresponding financial ratio is higher or lower than the corresponding set financial normal indicators, generate corresponding risk warnings based on the type of corresponding financial ratios; Input language model: Input the risk warning generated by the enterprise at the current time point into the constructed large language model. Based on the risk warning type generated by the enterprise at the current time point, extract the historical data of the financial ratio corresponding to the risk warning type from the database inside the large language model, and analyze it to obtain the change trend index of the risk warning type corresponding to the current time point. ; where j represents the number of the current risk warning type; the change trend index corresponding to the current time point according to the risk warning type , generate an initial report on the company’s financial and tax risks; Expert advice integration: Filter online experts corresponding to the risk warning type from the constructed expert network, select the expert with the largest processing evaluation index G as the consulting expert for the risk warning of the enterprise at the current time point, and send the generated initial report to the expert; Final report generation: Experts conduct an assessment based on the initial report generated for the enterprise's risk warning type at the current time point, and add the recommendations after the assessment to the initial report for feedback. The report after feedback is used as the risk treatment report for the enterprise's risk warning at the current time point.

[0007] In some embodiments, the change trend index of the risk warning type corresponding to the current time point is obtained. , specifically: Arrange the historical data of the financial ratio corresponding to the risk warning type according to the set time format; after the arrangement is completed, intercept the financial ratio data corresponding to the risk warning type within the set time period; further identify the data points that generate risk warnings corresponding to the financial ratio data within the set time period, and record them as risk values; Check the specific event marks of the risk values ​​within the set time period; specific event marks include market change marks, policy adjustment marks or internal enterprise change marks; preset the impact weight coefficients of different specific events; take the difference between each group of risk values ​​and the corresponding financial normal indicators, multiply each group of calculated differences by the corresponding impact weight coefficient, classify each group of multiplied values ​​based on the type of specific event, accumulate the values ​​belonging to the same category, and record the accumulated values ​​as market risk value, adjusted risk value and internal risk value respectively; The market risk value, adjusted risk value and internal risk value are used as the length, width and height of the three-dimensional rectangle respectively to construct the risk model of the enterprise within the set time period, and the surface area of ​​the risk model is calculated as the risk valuation of the enterprise corresponding to the risk warning type within the set time period. ; The number of risk values ​​within the set time period is recorded as the risk amount, and the proportion of the risk amount within the set time period to the total amount of financial ratio data is calculated. The calculated proportion is used as the risk proportion of the corresponding risk warning type within the set time period of the enterprise. ; Retain each group of risk values ​​within the set time period, and intercept the normal financial ratio data within the set time period, construct a line graph of the changes in each group of risk values, extract the difference between each group of risk values ​​and the corresponding normal financial indicators, and draw the numerical points corresponding to each group of differences in the line graph. The vertical axis is the specific value of the difference, and the horizontal axis is the time sequence. Connect adjacent numerical points to obtain the risk change line; use each group of numerical points in the line graph as the reference point, construct a vertical line between each group of reference points and the horizontal axis, count the length of each group of vertical lines and accumulate them, and use the accumulated length value as the risk length value of this risk warning type within the set time period. ; Further compare the lengths of two adjacent groups of vertical lines. When the length of the previous vertical line is less than the length of the subsequent vertical line, mark the angle value constructed by the two as angle change one; when the length of the previous vertical line is greater than the length of the subsequent vertical line, mark the constructed angle value as angle change two; extract all angle values ​​of angle change one and sum them to obtain a first total value; extract all angle values ​​of angle change two and sum them to obtain a second total value; divide the first total value by the second total value to obtain the risk change value of the risk warning type within the set time period. ; According to the formula The proportion of risk , Risk Long Value and risk variation Perform weighted calculation to obtain the risk valuation of the risk warning type within the set time period. ;in , as well as The corresponding risk proportions of the risk warning types , Risk Long Value and risk variation The weight factor of influence.

[0008] In some embodiments, the change trend index of the risk warning type corresponding to the current time point is obtained. , further comprising: Extract the risk valuation of the enterprise for this risk warning type within the set time period. and Risk Valuation II Substitute into the formula , perform weighted calculation to obtain the change trend index of the risk warning type corresponding to the current time point ;in and They respectively represent the maximum allowable risk valuation 1 and the maximum allowable risk valuation 2 for the risk warning type within the time period set by the enterprise; and The risk valuation corresponding to the risk warning type is and Risk Valuation II The weight factor of influence.

[0009] In some embodiments, an initial report of the enterprise's financial and tax risks is generated, specifically: Extract the preset index value ranges of the risk warning type from the database, set each index value range to correspond to a risk assessment level, and calculate the change trend index of the risk warning type at the current time point. Match with the value range of each preset index to obtain the risk assessment level of the enterprise corresponding to the current time point; extract the pre-stored report template of this risk warning type from the database, and input the enterprise's risk warning type, generation time point, risk assessment level, processing suggestions output by the language model, and historical processing records of this risk warning type into the corresponding pre-stored report template to obtain an initial report on the enterprise's financial and tax risks.

[0010] In some embodiments, the processing evaluation index G of the corresponding online expert is obtained, specifically: Obtain the personal information of the online expert corresponding to the risk warning type; the personal information includes working hours and number of consultations; convert the working hours and number of consultations of the corresponding online expert into corresponding scoring values; extract the value ranges of each group of working hours and the value ranges of each group of consultations corresponding to the risk warning type from the database; set each group of working hours value ranges and each group of consultations value ranges to correspond to an experience score; match the working hours and number of consultations of the corresponding online expert with the corresponding value ranges, and use the matching results as the working valuation and consultation valuation of the corresponding online expert; Based on the established feedback mechanism, after each consultation with the corresponding online expert, the expert is evaluated according to the actual effect of the implementation, and the evaluation result is quantified through a scoring system; the score value corresponding to each consultation number of the corresponding online expert is obtained, and the average of each group of score values ​​is taken as the professional valuation of the corresponding online expert; the weight coefficients of the expert's professional valuation, consulting valuation and professional valuation corresponding to the risk warning type are extracted from the database; the professional valuation, consulting valuation and professional valuation of the online expert corresponding to the risk warning type are multiplied by the corresponding weight coefficients respectively, and then the sum is obtained to obtain the professional evaluation index of the corresponding online expert .

[0011] In some embodiments, obtaining the processing evaluation index G corresponding to the online expert further includes: Obtain the duration of each consultation of the corresponding online expert, mark the time point when the initial report is sent to the expert as the start time, mark the time point when the expert completes the feedback and processing suggestions as the end time, calculate the time difference between the start time and the end time, obtain the duration of the consultation, calculate the mean of the duration used for each group, use the calculated mean as the average consultation duration of the corresponding online expert, extract the value range of the average consultation duration of each group corresponding to the risk warning type from the database; set the value range of the average consultation duration of each group to correspond to an efficiency score; match the average consultation duration of the corresponding online expert with the value range of the average consultation duration of each group, and use the matching result as the efficiency evaluation index of the corresponding online expert ; Will correspond to the professional evaluation index of online experts Efficiency Evaluation Index Substitute into the formula Perform weighted calculation to obtain the online expert's processing evaluation index G corresponding to the risk warning type; where e is a preset constant; b1 and b2 are professional evaluation indexes respectively. Efficiency Evaluation Index The weight factor of influence.

[0012] Compared with the prior art, the present invention has the following beneficial effects: After the enterprise generates the corresponding risk warning, the present invention extracts the historical data of the risk warning type within a set time period and further analyzes it to obtain the change trend index of the risk warning type corresponding to the current time point of the enterprise. According to the obtained change trend index, the risk trend can be quantified, which helps the enterprise consider potential risks in long-term strategic planning, improves the adaptability and resilience of the enterprise, and provides more comprehensive data for the processing of financial and tax risks; The present invention obtains the professional evaluation index and efficiency evaluation index of the corresponding online expert by extracting relevant parameters of the online expert of the corresponding risk warning type and analyzing them, comprehensively evaluates the ability of the expert, and further performs weighted calculation on the professional evaluation index and the efficiency evaluation index to obtain the processing evaluation index of the online expert corresponding to the risk warning type. Based on the calculated processing evaluation index, it can help enterprises screen out the most suitable experts to obtain high-quality and efficient consulting services, thereby improving the efficiency and quality of risk processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Further details, features and advantages of the present application are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which: Figure 1 is a flow chart of the present invention; Figure 2 It is a line graph of changes in the present invention. DETAILED DESCRIPTION

[0014] Several embodiments of the present application will be described in more detail below with reference to the accompanying drawings so that those skilled in the art can implement the present application. The present application can be embodied in many different forms and purposes and should not be limited to the embodiments described herein. These embodiments are provided to make the present application comprehensive and complete, and to fully convey the scope of the present application to those skilled in the art. The embodiments do not limit the present application.

[0015] Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those of ordinary skill in the art to which this application belongs. It will be further understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the relevant art and / or the context of this specification, and will not be interpreted in an idealized or overly formal sense unless explicitly defined as such herein.

[0016] See also Figure 1-Figure 2 As shown, a method for handling financial and tax risks based on a large language model includes: Financial and tax risk analysis: collect relevant financial data of the enterprise and pre-process it; automatic data collection can be achieved by developing interfaces with the enterprise's financial and tax systems; relevant financial data include but are not limited to balance sheets, income statements, cash flow statements, tax returns and bank statements; pre-processing includes data cleaning, format unification and data verification; integrate the relevant financial data after pre-processing to construct the financial and tax data set of the enterprise at the current time point; set a series of financial normal indicators based on the actual situation of the current enterprise; the actual situation includes but is not limited to the industry characteristics, enterprise scale and business strategy of the enterprise; extract the financial ratios in the financial data set of the enterprise at the current time point and compare them with the corresponding set financial normal indicators; financial ratios include but are not limited to net profit margin, debt ratio and cash flow ratio; if the comparison results show that the corresponding financial ratio is higher or lower than the corresponding set financial normal indicator, a corresponding risk warning is generated based on the type of the corresponding financial ratio; risk warnings include but are not limited to the risk of declining profitability, excessive debt risk and cash flow shortage risk; Input language model: Input the risk warning generated by the enterprise at the current time point into the constructed large language model. Based on the risk warning type generated by the enterprise at the current time point, extract the historical data of the financial ratio corresponding to the risk warning type from the database inside the large language model, and analyze it to obtain the change trend index of the risk warning type corresponding to the current time point. ; where j represents the number of the current risk warning type; the change trend index corresponding to the current time point according to the risk warning type , generate an initial report on the company’s financial and tax risks; Get the change trend index of the risk warning type corresponding to the current time point , specifically: Arrange the historical data of the financial ratio corresponding to the risk warning type according to the set time format; the set time format can be monthly, quarterly or annual; after the arrangement is completed, intercept the financial ratio data corresponding to the risk warning type within the set time period; for example, the historical data of the past three months, the past quarter or the past year can be intercepted; further identify the data points that generate risk warnings corresponding to the financial ratio data within the set time period, and record them as risk values; Check the specific event marks of risk values ​​within the set time period; specific event marks include market change marks, policy adjustment marks or internal enterprise change marks; preset impact weight coefficients of different specific events; the preset impact weight coefficients can be adjusted according to actual conditions; take the difference between each group of risk values ​​and the corresponding normal financial indicators, multiply each group of calculated differences by the corresponding impact weight coefficient, classify each group of multiplied values ​​based on the type of specific event, accumulate the values ​​belonging to the same category, and record the accumulated values ​​as market risk value, adjusted risk value and internal risk value respectively; The market risk value, adjusted risk value and internal risk value are used as the length, width and height of the three-dimensional rectangle respectively to construct the risk model of the enterprise within the set time period, and the surface area of ​​the risk model is calculated as the risk valuation of the enterprise corresponding to the risk warning type within the set time period. ; It should be noted that quantifying the impact of specific events on financial ratios in a numerical way makes risk assessment more objective and accurate.

[0017] The number of risk values ​​within the set time period is recorded as the risk amount, and the proportion of the risk amount within the set time period to the total amount of financial ratio data is calculated. The calculated proportion is used as the risk proportion of the corresponding risk warning type within the set time period of the enterprise. ; Keep each group of risk values ​​within the set time period, and intercept the normal financial ratio data within the set time period, build a line chart of the changes in each group of risk values, extract the difference between each group of risk values ​​and the corresponding normal financial indicators, and draw the value points corresponding to each group of differences in the line chart. The vertical axis is the specific value of the difference, and the horizontal axis is the time sequence. Connect adjacent value points to get the risk change line; Take each group of numerical points in the line chart as the reference point, construct the vertical line between each group of reference points and the horizontal axis, count the length of each group of vertical lines and accumulate them, and use the accumulated length value as the risk length value of the risk warning type within the set time period ; Further compare the lengths of two adjacent groups of vertical lines. When the length of the previous vertical line is less than the length of the subsequent vertical line, mark the angle value constructed by the two as angle change one; when the length of the previous vertical line is greater than the length of the subsequent vertical line, mark the constructed angle value as angle change two; extract all angle values ​​of angle change one and sum them to obtain a first total value; extract all angle values ​​of angle change two and sum them to obtain a second total value; divide the first total value by the second total value to obtain the risk change value of the risk warning type within the set time period. ; According to the formula The proportion of risk , Risk Long Value and risk variation Perform weighted calculation to obtain the risk valuation of the risk warning type within the set time period. ;in , as well as The corresponding risk proportions of the risk warning types , Risk Long Value and risk variation The impact weight factor is set based on the specific risk warning type; It should be noted that the risk valuation 2 calculated through weighted calculation provides a numerical value for quantifying the risk. The larger the risk valuation 2, the greater the triggering frequency and change range of this risk warning type for the enterprise within the set time period.

[0018] Further, extract the risk valuation of the enterprise for this risk warning type within the set time period. and Risk Valuation II Substitute into the formula , perform weighted calculation to obtain the change trend index of the risk warning type corresponding to the current time point ;in and They respectively represent the maximum allowable risk valuation 1 and the maximum allowable risk valuation 2 for the risk warning type within the time period set by the enterprise; and The risk valuation corresponding to the risk warning type is and Risk Valuation II The impact weight factor is set based on the specific risk warning type; It should be noted that after the enterprise generates the corresponding risk warning, the historical data of the risk warning type within the set time period is extracted and further analyzed to obtain the change trend index of the risk warning type corresponding to the current time point of the enterprise. Based on the obtained change trend index, the risk trend can be quantified, which helps the enterprise consider potential risks in long-term strategic planning, improves the adaptability and resilience of the enterprise, and provides more comprehensive data for the handling of financial and tax risks.

[0019] Extract the preset index value ranges of the risk warning type from the database, set each index value range to correspond to a risk assessment level, and calculate the change trend index of the risk warning type at the current time point. Match with the preset index value range to obtain the enterprise's risk assessment level at the current time point; the preset index value range can be adjusted later according to actual conditions; Extract the pre-stored report template of the risk warning type from the database, input the enterprise's risk warning type at the current time point, generation time point, risk assessment level, processing suggestions output by the language model, and historical processing records of the risk warning type into the corresponding pre-stored report template, and obtain an initial report on the enterprise's financial and tax risks; Integration of expert advice: Filter online experts corresponding to the risk warning type from the constructed expert network; the expert network covers experts in different fields; select the expert with the largest processing evaluation index G as the consulting expert for the risk warning of the enterprise at the current time point, and send the generated initial report to the expert; Get the processing evaluation index G of the corresponding online expert: Obtain the personal information of the online expert corresponding to the risk warning type; the personal information includes working hours and number of consultations; convert the working hours and number of consultations of the corresponding online expert into corresponding scoring values; extract the value ranges of each group of working hours and the value ranges of each group of consultations corresponding to the risk warning type from the database; set each group of working hours value ranges and each group of consultations value ranges to correspond to an experience score; the experience score range can be set to 1-10, and can be adjusted later according to actual conditions; match the working hours and number of consultations of the corresponding online expert with the corresponding value ranges, and use the matching results as the working valuation and consultation valuation of the corresponding online expert; It should be noted that, for example, if the current risk warning type is "risk of declining profitability", online experts with professional knowledge in the fields of profitability analysis, financial strategy and market trends are screened from the constructed expert network, and the working hours and consultation times of these experts are collected and converted into scoring values; the corresponding time value ranges and number of times value ranges of each group of the risk type of "risk of declining profitability" are extracted from the database; Experts with more than 10 years of experience can be given 10 points, those with 5-10 years of experience can be given 7 points, and those with less than 5 years of experience can be given 4 points; experts with more than 100 consultations can be given 10 points, those with 50-100 consultations can be given 7 points, and those with less than 50 consultations can be given 4 points.

[0020] Through the above, it is possible to quantify expert experience and improve the professionalism of financial and tax risk management.

[0021] At the same time, based on the established feedback mechanism, after each consultation with the corresponding online expert, the expert is evaluated according to the actual effect of the implementation, and the evaluation result is quantified through a scoring system; the scoring range can be set from 1 to 10, and the operator or relevant management personnel will provide feedback and scoring; the scoring value corresponding to each consultation number of the corresponding online expert is obtained, and the average of each group of scoring values ​​is taken as the professional valuation of the corresponding online expert; Extract the weight coefficients of the expert valuation, consulting valuation and professional valuation corresponding to the risk warning type from the database; if the weight coefficients of the expert valuation, consulting valuation and professional valuation are marked as s1, s2 and s3 respectively, then s3>s2>s1; The professional valuation, consulting valuation and professional valuation of the online expert corresponding to the risk warning type are multiplied by the corresponding weight coefficient respectively, and then the sum is obtained to obtain the professional evaluation index of the corresponding online expert. ; It should be noted that experts selected through professional assessment indexes can provide more professional and in-depth financial and tax risk management advice.

[0022] Furthermore, the duration of each consultation of the corresponding online expert is obtained, the time point when the initial report is sent to the expert is marked as the start time, the time point when the expert completes the feedback and processing suggestions is marked as the end time, the time difference between the start time and the end time is calculated to obtain the duration of the consultation, the mean of the duration used for each group is calculated, the calculated mean is used as the average consultation duration of the corresponding online expert, and the value range of the average consultation duration of each group corresponding to the risk warning type is extracted from the database; the value range of the average consultation duration of each group is set to correspond to an efficiency score; the shorter the average consultation duration, the higher the matching efficiency score; the efficiency score range can be set from 1 to 10, and can be adjusted later according to actual conditions; the average consultation duration of the corresponding online expert is matched with the value range of the average consultation duration of each group, and the matching result is used as the efficiency evaluation index of the corresponding online expert. ; It should be noted that the efficiency score corresponding to the range of average consultation duration for each group is set. For example: less than 2 hours: 10 points; 2-4 hours: 7 points; 4-6 hours: 4 points; more than 6 hours: 1 point.

[0023] Will correspond to the professional evaluation index of online experts Efficiency Evaluation Index Substitute into the formula Perform weighted calculation to obtain the online expert processing evaluation index G corresponding to the risk warning type; e is a preset constant, and its value is set to 1.592; b1 and b2 are professional evaluation indexes respectively. Efficiency Evaluation Index The impact weight factor is set based on the specific risk warning type; It should be noted that the calculated processing assessment index can help enterprises screen out the most suitable experts to obtain high-quality and efficient consulting services, thereby improving the efficiency and quality of risk management.

[0024] Final report generation: Experts conduct an assessment based on the initial report generated for the enterprise's risk warning type at the current time point, and add the recommendations after the assessment to the initial report for feedback. The report after feedback is used as the risk treatment report for the enterprise's risk warning at the current time point.

[0025] It should be noted that the risk management reports in the above process comply with the requirements of relevant laws and regulations, and are continuously optimized according to changes in laws and regulations and updates to regulatory requirements.

[0026] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to only specific implementation methods. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for handling financial and tax risks based on a large language model, characterized in that: include: Financial and tax risk analysis: collect relevant financial data of the enterprise and conduct pre-processing; Integrate the pre-processed relevant financial data to construct the financial and tax data set of the enterprise at the current time point; Based on the actual situation of the current enterprise, a series of financial normal indicators are set; the financial ratios in the enterprise financial data set at the current time point are extracted and compared with the corresponding set financial normal indicators; if the comparison result shows that the corresponding financial ratio is higher or lower than the corresponding set financial normal indicator, a corresponding risk warning is generated based on the type of the corresponding financial ratio; Input language model: Input the risk warning generated by the enterprise at the current time point into the constructed large language model. Based on the risk warning type generated by the enterprise at the current time point, extract the historical data of the financial ratio corresponding to the risk warning type from the database inside the large language model, and analyze it to obtain the change trend index of the risk warning type corresponding to the current time point. ; where j represents the number of the current risk warning type; the change trend index corresponding to the current time point according to the risk warning type , generate an initial report on the company’s financial and tax risks; Expert advice integration: Filter online experts corresponding to the risk warning type from the constructed expert network, select the expert with the largest processing evaluation index G as the consulting expert for the risk warning of the enterprise at the current time point, and send the generated initial report to the expert; Final report generation: Experts conduct an assessment based on the initial report generated for the enterprise's risk warning type at the current time point, and add the recommendations after the assessment to the initial report for feedback. The report after feedback is used as a risk management report for the enterprise's risk warning at the current time point and is sent to the operator.

2. According to claim 1, a method for handling financial and tax risks based on a large language model is characterized in that: Get the change trend index of the risk warning type corresponding to the current time point , specifically: Arrange the historical data of the financial ratio corresponding to the risk warning type according to the set time format; after the arrangement is completed, intercept the financial ratio data corresponding to the risk warning type within the set time period; further identify the data points that generate risk warnings corresponding to the financial ratio data within the set time period, and record them as risk values; View specific event markers of risk values ​​within a set time period; specific event markers include market change markers, policy adjustment markers, or internal enterprise change markers; Preset the impact weight coefficients of different specific events; take the difference between each group of risk values ​​and the corresponding financial normal indicators, multiply each group of calculated differences by the corresponding impact weight coefficient, classify each group of multiplied values ​​based on the type of specific events, accumulate the values ​​belonging to the same classification, and record the accumulated values ​​as market risk value, adjusted risk value and internal risk value respectively; The market risk value, adjusted risk value and internal risk value are used as the length, width and height of the three-dimensional rectangle respectively to construct the risk model of the enterprise within the set time period, and the surface area of ​​the risk model is calculated as the risk valuation of the enterprise corresponding to the risk warning type within the set time period. ; The number of risk values ​​within the set time period is recorded as the risk amount, and the proportion of the risk amount within the set time period to the total amount of financial ratio data is calculated. The calculated proportion is used as the risk proportion of the corresponding risk warning type within the set time period of the enterprise. ; Retain each group of risk values ​​within the set time period, and intercept the normal financial ratio data within the set time period, construct a line graph of the changes in each group of risk values, extract the difference between each group of risk values ​​and the corresponding normal financial indicators, and draw the numerical points corresponding to each group of differences in the line graph. The vertical axis is the specific value of the difference, and the horizontal axis is the time sequence. Connect adjacent numerical points to obtain the risk change line; use each group of numerical points in the line graph as the reference point, construct a vertical line between each group of reference points and the horizontal axis, count the length of each group of vertical lines and accumulate them, and use the accumulated length value as the risk length value of this risk warning type within the set time period. ; Further compare the lengths of two adjacent groups of vertical lines. When the length of the previous vertical line is less than the length of the subsequent vertical line, mark the angle value constructed by the two as angle change one; when the length of the previous vertical line is greater than the length of the subsequent vertical line, mark the constructed angle value as angle change two; extract all angle values ​​of angle change one and sum them to obtain a first total value; extract all angle values ​​of angle change two and sum them to obtain a second total value; divide the first total value by the second total value to obtain the risk change value of the risk warning type within the set time period. ; According to the formula The proportion of risk , Risk Long Value and risk variation Perform weighted calculation to obtain the risk valuation of the risk warning type within the set time period. ;in , as well as The corresponding risk proportions of the risk warning types , Risk Long Value and risk variation The weight factor of influence.

3. According to claim 2, a method for handling financial and tax risks based on a large language model is characterized in that: Get the change trend index of the risk warning type corresponding to the current time point , further comprising: Extract the risk valuation of the enterprise for this risk warning type within the set time period. and Risk Valuation II Substitute into the formula , perform weighted calculation to obtain the change trend index of the risk warning type corresponding to the current time point ;in and They respectively represent the maximum allowable risk valuation 1 and the maximum allowable risk valuation 2 for the risk warning type within the time period set by the enterprise; and The risk valuation corresponding to the risk warning type is and Risk Valuation II The weight factor of influence.

4. According to claim 3, a method for handling financial and tax risks based on a large language model is characterized in that: Generate an initial report on the company's fiscal and tax risks, specifically: Extract the preset index value ranges of the risk warning type from the database, set each index value range to correspond to a risk assessment level, and calculate the change trend index of the risk warning type at the current time point. Match with the value range of each preset index to obtain the risk assessment level of the enterprise corresponding to the current time point; extract the pre-stored report template of this risk warning type from the database, and input the enterprise's risk warning type, generation time point, risk assessment level, processing suggestions output by the language model, and historical processing records of this risk warning type into the corresponding pre-stored report template to obtain an initial report on the enterprise's financial and tax risks.

5. According to claim 4, a method for handling financial and tax risks based on a large language model is characterized in that: The processing evaluation index G of the corresponding online expert is obtained, specifically: Obtain the personal information of the online expert corresponding to the risk warning type; the personal information includes the length of service and number of consultations; convert the length of service and number of consultations of the corresponding online expert into a corresponding score value; Extract from the database the value ranges of each group of working hours and each group of consultation times corresponding to the risk warning type; Set a value range of each group's working hours and a value range of each group's number of consultations to correspond to an experience score; Match the working hours and consultation times of the corresponding online experts with the corresponding value ranges, and use the matching results as the working valuation and consultation valuation of the corresponding online experts; Based on the established feedback mechanism, after each consultation with the corresponding online expert, the expert is evaluated according to the actual effect of the implementation, and the evaluation result is quantified through a scoring system; the score value corresponding to each consultation number of the corresponding online expert is obtained, and the average of each group of score values ​​is taken as the professional valuation of the corresponding online expert; the weight coefficients of the expert's professional valuation, consulting valuation and professional valuation corresponding to the risk warning type are extracted from the database; the professional valuation, consulting valuation and professional valuation of the online expert corresponding to the risk warning type are multiplied by the corresponding weight coefficients respectively, and then the sum is obtained to obtain the professional evaluation index of the corresponding online expert .

6. A method for handling financial and tax risks based on a large language model according to claim 5, characterized in that: Obtaining the processing evaluation index G of the corresponding online expert further includes: Obtain the duration of each consultation of the corresponding online expert, mark the time point when the initial report is sent to the expert as the start time, mark the time point when the expert completes the feedback and processing suggestions as the end time, calculate the time difference between the start time and the end time, obtain the duration of the consultation, calculate the mean of the duration used for each group, use the calculated mean as the average consultation duration of the corresponding online expert, extract the value range of the average consultation duration of each group corresponding to the risk warning type from the database; set the value range of the average consultation duration of each group to correspond to an efficiency score; match the average consultation duration of the corresponding online expert with the value range of the average consultation duration of each group, and use the matching result as the efficiency evaluation index of the corresponding online expert ; Will correspond to the professional evaluation index of online experts Efficiency Evaluation Index Substitute into the formula Perform weighted calculation to obtain the online expert's processing evaluation index G corresponding to the risk warning type; where e is a preset constant; b1 and b2 are professional evaluation indexes respectively. Efficiency Evaluation Index The weight factor of influence.

Citation Information

Cited By

  • Finance and tax data processing method and processing system

    CN119443825A