Enterprise tax risk monitoring and analyzing system based on big data

Abstract

The invention discloses an enterprise tax risk monitoring and analysis system based on big data, which belongs to the field of enterprise tax risk monitoring and analysis and comprises an enterprise data monitoring module, a big data search module, a business division and correction module and a tax risk analysis module. The enterprise data monitoring module is used for monitoring and collecting enterprise internal business data; the big data search module is used for searching tax information data of tax regulations and companying enterprises; according to the enterprise tax risk monitoring and analysis system based on the big data, the enterprise data monitoring module, the big data search module, the business division and correction module and the tax risk analysis module are arranged, so that business types can be judged by means of the big data, and tax errors caused by business confusion and regulation updating are prevented; meanwhile, analysis, early warning and prediction can be carried out on tax services, and enterprises are assisted to efficiently manage tax risks.

Description

Technical Field

[0001] The present invention belongs to the field of enterprise tax risk monitoring and analysis, and specifically relates to an enterprise tax risk monitoring and analysis system based on big data. Background Art

[0002] Traditional tax management models do not pay enough attention to non-tax business, which can easily lead to confusion in the division between tax and non-tax business, resulting in accounting deviations and reporting errors. At the same time, tax laws and regulations are updated rapidly, making it difficult for enterprises to detect in a timely manner, prone to omissions and misreports, increasing tax risks.

[0003] In addition, existing risk analysis methods also have deficiencies. The threshold settings for single-index warnings often lack full consideration of the market and the actual situation of the enterprise itself, resulting in low warning accuracy. For example, the thresholds are not reasonably set based on the actual data of enterprises in the same industry and with the same development status, making the warnings inaccurate. In addition, only analyzing risks through single indicators makes it difficult to discover potential tax risks of enterprises. Therefore, there is an urgent need for an enterprise tax risk monitoring and analysis system based on big data to achieve accurate identification, scientific analysis, timely warning, and effective monitoring of tax risks by leveraging the advantages of big data. Summary of the Invention

[0004] The present invention aims to solve at least one of the technical problems existing in the prior art. For this purpose, the present invention proposes an enterprise tax risk monitoring and analysis system based on big data, enabling more comprehensive monitoring of enterprise tax risks.

[0005] To achieve the above objective, the present invention adopts the following technical solutions:

[0006] An enterprise tax risk monitoring and analysis system based on big data, including an enterprise data monitoring module, a big data search module, a business division correction module, and a tax risk analysis module;

[0007] The enterprise data monitoring module is used to monitor and collect internal business data of the enterprise;

[0008] The big data search module is used to search for tax laws and regulations and tax information data of peer enterprises;

[0009] The business division correction module receives the searched tax law clauses, judges the internal business classification situation of the enterprise. When the classification is incorrect, it re-divides the business and then transmits the tax business and related data to the tax risk analysis module; when the classification is correct, it directly transmits the tax business-related data;

[0010] The tax risk analysis module is responsible for tax risk analysis, early warning and prediction. First, it divides individual indicators and conducts individual and comprehensive scoring. Second, it sets individual and comprehensive risk thresholds based on publicly available tax data of peers. When an individual indicator exceeds the standard, a red early warning is issued and the abnormal indicator is locked. When the comprehensive indicator exceeds the standard, an orange early warning is issued, and time series analysis is started to locate the risk source, predict the trend, and push the results to the enterprise data monitoring module for dynamic enhanced monitoring.

[0011] Further, the business division correction module receives the searched tax regulations and terms, judges the internal business classification of the enterprise. When the classification is incorrect, it re-divides the business and then transmits the tax business and related data to the tax risk analysis module. When the classification is correct, the process of directly transmitting the tax business-related data is as follows:

[0012] Analyze the received tax regulations and terms, extract the key information for defining business types, and compare the collected enterprise business data with the key information of the tax regulations and terms to judge whether the internal tax business and non-tax business of the enterprise are correctly divided. When the division is correct, transmit the relevant information of the tax business within the enterprise to the tax risk analysis module. When the division is incorrect, re-classify the internal tax business and non-tax business of the enterprise according to the key information of the tax regulations, correct the division scope of the tax business and non-tax business within the enterprise, and transmit the relevant information of the corrected tax business to the tax risk analysis module.

[0013] The regulations searched through big data can ensure the accurate division of tax and non-tax businesses. On the one hand, it can correct incorrect divisions in a timely manner, avoid tax processing errors caused by unclear business definitions, and reduce potential tax risks. On the other hand, it transmits the tax business information after correct division to the risk analysis module, providing a reliable data basis for subsequent in-depth analysis of tax risks.

[0014] Further, the process of the tax risk analysis module being responsible for tax risk analysis, early warning and prediction, first dividing individual indicators and conducting individual and comprehensive scoring, second setting individual and comprehensive risk thresholds based on publicly available tax data of peers, issuing a red early warning and locking abnormal indicators when an individual indicator exceeds the standard, issuing an orange early warning when the comprehensive indicator exceeds the standard, starting time series analysis to locate the risk source, predict the trend, and pushing the results to the enterprise data monitoring module for dynamic enhanced monitoring is as follows:

[0015] Divide tax indicators, including tax burden rate, gross profit margin, asset-liability ratio, invoice issuance rate, and sales volume change rate, and analyze the correlation between the enterprise's past tax risk events and each indicator, assign high weights to indicators with high correlation, and assign low weights to indicators with low correlation;

[0016] Receive relevant data of peer enterprises, judge the status of peer enterprises, including the initial stage, the rising stage, the stable stage, and the recession stage, select the type that suits this enterprise, calculate the average value of the single thresholds of enterprises of the same type, and use the average value as the final threshold. At the same time, set the comprehensive threshold according to the current weights of each indicator, and analyze the relationship between the comprehensive threshold and the weights of each indicator; calculate the score value of the indicator and the comprehensive risk value of each indicator, and at the same time calculate the hedging degree between the single indicator and the comprehensive score. When the hedging degree is abnormal, adjust the weights of each indicator to make the hedging degree within the normal range. Finally, modify the comprehensive threshold according to the adjusted weights.

[0017] When the value of a single indicator exceeds the single threshold, a red warning is triggered. When the comprehensive risk value triggers the comprehensive score, an orange warning is triggered. When both the red warning and the orange warning are triggered, directly analyze the single indicator that triggers the warning. When the red warning is not triggered and the orange warning is triggered, use time series analysis to analyze each single indicator, locate the specific anomaly, and conduct risk prediction. Finally, transmit the risk prediction result to the enterprise data monitoring module and implement dynamic intensive monitoring of the relevant indicators of the enterprise.

[0018] Furthermore, the tax indicators are divided. The indicators include tax burden rate, gross profit margin, asset - liability, invoice issuance rate, and sales volume change rate, and analyze the past tax risk events of the enterprise. Specifically, the process of analyzing the indicators with high correlation with the events, assigning high weights to the indicators with high correlation, and low weights to the indicators with low correlation is as follows:

[0019] Divide the tax indicators within the enterprise into tax burden rate, gross profit margin, asset - liability, invoice issuance rate, and sales volume change rate, calculate the correlation coefficient between each indicator and the occurrence frequency of the past tax risk events of the enterprise to measure the linear correlation degree, and finally normalize the absolute value of the correlation coefficient into weights to ensure that the sum of the weights is 1. The specific formula is:

[0020]

[0021] where, w i is the calculated indicator weight, and r i ∣ is the absolute value of the correlation coefficient of indicator i;

[0022] Use the calculated weights as the weights of specific indicators. The indicators with high correlation are assigned high weights, and the indicators with low correlation are assigned low weights;

[0023] Furthermore, the process of receiving relevant data of peer enterprises, judging the status of peer enterprises, including the initial stage, the rising stage, the stable stage, and the decline stage, selecting the type that suits this enterprise, calculating the average value of the single thresholds of enterprises of the same type, using the average value as the final threshold, setting the comprehensive threshold according to the current weights of each index, and analyzing the relationship between the comprehensive threshold and the weights of each index; calculating the score values of the indexes and the comprehensive risk values of each index, calculating the hedging degree between the single index and the comprehensive score at the same time, when the hedging degree is abnormal, adjusting the weights of each index to make the hedging degree within the normal range, and finally modifying the comprehensive threshold according to the adjusted weights is as follows:

[0024] According to the data of the received peer enterprises, judge the types of peer enterprises and this enterprise; when the sales volume change rate > 20% and the gross profit margin < the industry average, it is classified as the initial stage; when the sales volume change rate > 10% and the asset - liability ratio < 60%, it is classified as the rising stage; when the sales volume change rate is ±5% and the tax burden rate is stable, it is classified as the rising stage; when the sales volume change rate < - 5% and the asset - liability ratio > 80%, it is classified as the decline stage;

[0025] After classifying peer enterprises and this enterprise, select enterprises of the same category as this enterprise, calculate the single - index thresholds of enterprises of the same category, and use the calculated single thresholds as the warning limits of this enterprise. At the same time, set the comprehensive threshold in combination with the historical data of this enterprise, and analyze the relationship between the comprehensive threshold and the proportion of the single - index weights;

[0026] Determine the reasonable range of each index according to the data of peer enterprises. When a single index is within the reasonable range, this index obtains the basic score. For every 1% that a single index exceeds the reasonable range, a deduction is made once, and finally the specific values of each single index are obtained;

[0027] Calculate the comprehensive risk value S according to the obtained values of the single indexes a , and the specific formula is:

[0028] w1·s1 + w2·s2 + w3·s3 + w4·s4 + w5·s5 = S a ;

[0029] Among them, w1, w2, w3, w4, w5 are the weights of each index, and s1, s2, s3, s4, s5 are the specific score values of each index;

[0030] Set a preset hedging degree threshold, calculate the correlation coefficient R between the score of each individual indicator and the comprehensive score, and calculate the hedging degree H based on the obtained correlation coefficient. When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. When the hedging degree is abnormal, further adjust the weights of each indicator. After the adjustment is completed, calculate the hedging degree between the individual indicator and the comprehensive score again until the hedging degree H is within the preset hedging degree threshold. Finally, adjust the comprehensive threshold according to the indicators with adjusted weights; when the hedging degree is normal, further analyze the tax risk and issue a warning.

[0031] Further, for the above-mentioned preset hedging degree threshold, calculate the correlation coefficient R between the score of each individual indicator and the comprehensive score, and calculate the hedging degree H based on the obtained correlation coefficient. When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. When the hedging degree is abnormal, further adjust the weights of each indicator. After the adjustment is completed, calculate the hedging degree between the individual indicator and the comprehensive score again until the hedging degree H is within the preset hedging degree threshold. The processing procedure is as follows:

[0032] Calculate the correlation coefficient R between the score of each individual indicator and the comprehensive score according to the correlation coefficient. The specific formula is: H = ∣R·Si - wi·S a ∣, where Si is the specific score of a single indicator, and wi is the specific weight of a single indicator;

[0033] When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. When the hedging degree is abnormal and the correlation coefficient R > 0, adjust the single weight, wi new = wi + α, where wi new is the adjusted weight, and α is the adjustment coefficient. Then adjust the other weights (j ≠ i), where wj is the weight of other indicators, and wj new is the adjusted weight of other indicators; when the hedging degree is abnormal and the correlation coefficient R < 0, adjust the single weight, wi new = wi - α, and then adjust the other weights (j ≠ i);

[0034] Use the adjusted weights to recalculate the comprehensive score, and calculate the hedging degree between the comprehensive score and the individual indicator again. When the hedging degree still exceeds the threshold range, repeat the adjustment of the indicator weights until the hedging degree is within the threshold range.

[0035] Further, when the value of a single indicator exceeds the single threshold, a red warning is triggered. When the comprehensive risk value triggers the comprehensive score, an orange warning is triggered. When both the red warning and the orange warning are triggered, directly analyze the single indicators that trigger the warning. When the red warning is not triggered and the orange warning is triggered, use time series analysis to analyze each single indicator, locate specific anomalies, and conduct risk prediction. Finally, transmit the risk prediction results to the enterprise data monitoring module and implement dynamic enhanced monitoring on the relevant enterprise indicators. The specific process is as follows:

[0036] Compare the single indicator with the single threshold. When the single indicator exceeds the single threshold, a red warning is triggered. When the red warning is triggered, there is no need to calculate the comprehensive score, and the orange warning is directly triggered. Further analyze this indicator and formulate countermeasures;

[0037] Compare the single indicator with the single threshold. When the single indicator does not exceed the single threshold, calculate the comprehensive score. When the comprehensive score does not exceed the comprehensive threshold, continuously monitor the enterprise data; when the comprehensive score exceeds the threshold, use time series analysis to analyze each single indicator, locate specific anomalies, and conduct risk prediction. Finally, transmit the risk prediction results to the enterprise data monitoring module and implement dynamic enhanced monitoring on the relevant enterprise indicators.

[0038] Further, when the comprehensive score exceeds the threshold, use time series analysis to analyze each single indicator, locate specific anomalies, and conduct risk prediction. Finally, transmit the risk prediction results to the enterprise data monitoring module and implement dynamic enhanced monitoring on the relevant enterprise indicators. The specific process is as follows:

[0039] Collect historical data of each single indicator, and decompose the time series X of each single indicator t to separate the trend term T t , seasonal term S t and residual term R t . The specific formula is: X t =T t +S t +R t . Set the residual threshold as μR±3σR, where μR is the historical residual. When the residual term R t exceeds the confidence interval ±3σ, it is marked as an abnormal point. Further calculate the seasonal intensity Q corresponding to the abnormal point to determine whether the current fluctuation conforms to the historical seasonal pattern. The specific formula is: When Q>0.6, it indicates a strong seasonal influence. When Q<0.3, it indicates random fluctuations, and filter the abnormal points accordingly;

[0040] Conduct risk prediction based on the filtered abnormal points and transmit the prediction results to the enterprise data monitoring module for dynamic enhanced monitoring.

[0041] Furthermore, the risk prediction is carried out based on the screened abnormal points, and the prediction results are transmitted to the enterprise data monitoring module for dynamic enhanced monitoring. The specific process is as follows:

[0042] Extract the slope of the trend term, the abnormal frequency and amplitude of the residuals. At the same time, calculate the industry deviation degree according to the data of peer enterprises, and construct the eigenvector by combining these data;

[0043] Use the SARIMAX model for modeling and prediction to predict the values of indicators in the future time. At the same time, output the risk probability through the loss function and obtain the risk time window, and finally integrate and output the specific risks.

[0044] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0045] The business division correction module set in the present invention can judge the internal business types of the enterprise before risk analysis. When directly analyzing the internal tax business of the enterprise, there may be a problem of confusion between tax business and non-tax business, resulting in incorrect business division. Therefore, judging the internal business categories of the enterprise in advance can effectively avoid tax calculation deviation caused by business confusion and reduce tax declaration errors caused by incorrect division. In addition, this module also combines the tax regulations and terms searched by big data to avoid the problems of missed reports and false reports caused by the enterprise's failure to detect the clause update in time, thereby effectively monitoring and reducing tax risks;

[0046] The tax risk analysis module set in the present invention can not only conduct in-depth tax risk analysis, but also conduct tax risk early warning and prediction. First, divide tax indicators, and judge the enterprises in the same state as this enterprise. Use the average of the single-index thresholds of multiple enterprises in the same state as the threshold of this enterprise, so that the setting of the single threshold can closely fit the market and its own actual situation, thereby increasing the accuracy of single-index early warning;

[0047] When the value of a single index does not exceed the threshold, it means that no specific risk has been identified yet. Therefore, by setting a comprehensive score, potential abnormal points can be found, and at the same time, tax risk prediction can be carried out according to the potential abnormal points; transmitting the predicted specific data to the enterprise data monitoring module can also effectively strengthen the monitoring of tax risks;

[0048] When calculating the comprehensive score, the weights of each index will be set according to the correlation degree between the internal tax events and the index of the enterprise. The indexes with high correlation degree are given high weights, and the indexes with low correlation degree are given low weights, so that the score can more accurately reflect the actual situation of tax risks;

[0049] By calculating the hedging degree between the comprehensive score and the individual scores, it is possible to determine whether the weight division of each indicator is reasonable. When the hedging degree is unreasonable, the weights of individual indicators are further adjusted, so that the calculated comprehensive score is more accurate. After the comprehensive score is adjusted, the comprehensive threshold is adjusted, which can also improve the accuracy of early warning;

[0050] When the score of an individual indicator triggers a red early warning, there is no need to calculate the comprehensive score, and the red and orange early warnings are directly triggered. This can simplify the analysis steps and facilitate the direct analysis of risk indicators. When the score of an individual indicator does not trigger a red early warning, but the comprehensive score triggers an orange early warning, the risk points of specific indicators will be analyzed through time series analysis, and the reasons for seasonal fluctuations will be excluded. Finally, risk prediction will be carried out based on the risk points, and the predicted data will be transmitted to the enterprise data monitoring module for dynamic enhanced monitoring to further strengthen tax risk monitoring. Description of the Drawings

[0051] Figure 1 It is a block diagram of an enterprise tax risk monitoring and analysis system based on big data according to the present invention. Detailed Embodiments

[0052] Next, the technical solutions of the present invention will be clearly and completely described in conjunction with the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] As Figure 1 shown, an enterprise tax risk monitoring and analysis system based on big data includes an enterprise data monitoring module, a big data search module, a business division correction module, and a tax risk analysis module;

[0054] The enterprise data monitoring module is used to monitor and collect the internal business data of the enterprise;

[0055] The big data search module is used to search tax regulations and tax information data of peer enterprises;

[0056] The business division correction module receives the searched tax regulation clauses, judges the internal business classification situation of the enterprise. When the classification is incorrect, the business is re-divided, and then the tax business and related data are transmitted to the tax risk analysis module; when the classification is correct, the tax business-related data is directly transmitted;

[0057] In this embodiment, the process of the business division correction module receiving the searched tax regulation clauses, judging the internal business classification situation of the enterprise, re-dividing the business when the classification is incorrect, and then transmitting the tax business and related data to the tax risk analysis module; directly transmitting the tax business-related data when the classification is correct is as follows:

[0058] Parse the received tax regulation clauses, extract the key information for defining business types, and compare the collected enterprise business data with the key information of the tax regulation clauses to determine whether the tax and non-tax operations within the enterprise are correctly classified. When the classification is correct, transmit the relevant information of the tax operations within the enterprise to the tax risk analysis module. When the classification is incorrect, reclassify the tax and non-tax operations within the enterprise according to the key information of the tax regulations, correct the classification scope of the tax and non-tax operations within the enterprise, and transmit the relevant information of the corrected tax operations to the tax risk analysis module.

[0059] It should be noted that due to the update of tax regulations, the updated tax regulations may change the scope of tax and non-tax operations. By searching for the latest tax regulations through big data and extracting keywords, it is possible to determine whether the business classification within the enterprise is correct, avoid the risk of incorrect payment of non-tax operations and missed payment of tax operations, and achieve the role of monitoring the tax risks of the enterprise;

[0060] When the tax regulations are not updated and the existing tax and non-tax operations have been compared and classified according to the regulations searched by big data, directly transmit the relevant data of the tax operations to the tax risk analysis module;

[0061] When there are new operations within the enterprise, it is also necessary to judge the categories of all operations according to the currently searched tax regulations to avoid incorrect business category classification;

[0062] The tax risk analysis module is responsible for tax risk analysis, early warning and prediction. First, divide single indicators and conduct single and comprehensive scoring. Second, set single indicator and comprehensive risk thresholds based on the publicly available tax data of the same industry. When a single indicator exceeds the standard, issue a red early warning and lock the abnormal indicator. When the comprehensive indicator exceeds the standard, issue an orange early warning, start time series analysis to locate the risk source, predict the trend, and push the results to the enterprise data monitoring module for dynamic enhanced monitoring;

[0063] In this embodiment, tax indicators are divided, and the indicators include tax burden rate, gross profit margin, asset-liability ratio, invoice issuance rate, and sales volume change rate. Analyze the past tax risk events of the enterprise. Specifically, analyze the indicators with high correlation with the events, and assign high weights to the indicators with high correlation and low weights to the indicators with low correlation. The processing process is as follows:

[0064] Divide the tax indicators within the enterprise into tax burden rate, gross profit margin, asset-liability ratio, invoice issuance rate, and sales volume change rate, calculate the correlation coefficient between each indicator and the occurrence frequency of the past tax risk events of the enterprise, measure the linear correlation degree, and finally normalize the absolute value of the correlation coefficient to weights to ensure that the sum of the weights is 1. The specific formula is:

[0065]

[0066] Among them, w i is the calculated index weight, and r i ∣ is the absolute value of the correlation coefficient of index i;

[0067] The calculated weight is used as the weight of the specific index. The index with a high degree of correlation is given a high weight, and the index with a low degree of correlation is given a low weight.

[0068] In this embodiment, relevant data of peer enterprises are received, the status of peer enterprises is judged, including the initial stage, the rising stage, the stable stage, and the recession stage. The type that suits this enterprise is selected, the average value of the single thresholds of enterprises of the same type is calculated, and the average value is used as the final threshold. At the same time, a comprehensive threshold is set according to the current weights of each index, and the relationship between the comprehensive threshold and the weights of each index is analyzed; the score value of the index and the comprehensive risk value of each index are calculated. At the same time, the hedging degree between the single index and the comprehensive score is calculated. When the hedging degree is abnormal, the weights of each index are adjusted to make the hedging degree within the normal range. Finally, the comprehensive threshold is modified according to the adjusted weights. The process is as follows:

[0069] According to the data of the received peer enterprises, judge the types of peer enterprises and this enterprise; and the sales growth rate > 20%, and the gross profit margin < industry average are classified as the initial stage; the sales growth rate > 10%, and the asset - liability ratio < 60% are classified as the rising stage; the sales growth rate ± 5%, and the tax burden rate is stable are classified as the rising stage; the sales growth rate < - 5%, and the asset - liability ratio > 80% are classified as the recession stage;

[0070] After classifying peer enterprises and this enterprise, select enterprises of the same category as this enterprise, calculate the single - index thresholds of enterprises of the same category, and use the calculated single thresholds as the warning boundaries of this enterprise. At the same time, set a comprehensive threshold in combination with the historical data of this enterprise, and analyze the relationship between the comprehensive threshold and the proportion of the single - index weights;

[0071] Determine the reasonable range of each index according to the data of peer enterprises. When a single index is within the reasonable range, this index obtains the basic score. For every 1% that a single index exceeds the reasonable range, a deduction is made once. Finally, the specific value of each single index is obtained;

[0072] Calculate the comprehensive risk value S a , and the specific formula is:

[0073] w1·s1 + w2·s2 + w3·s3 + w4·s4 + w5·s5 = S a ;

[0074] Among them, w1, w2, w3, w4, and w5 are the weights of each index, and s1, s2, s3, s4, and s5 are the specific scores of each index;

[0075] Set a preset hedging degree threshold, calculate the correlation coefficient R between the score of each single index and the comprehensive score, and calculate the hedging degree H based on the obtained correlation coefficient. When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. When the hedging degree is abnormal, further adjust the weights of each index. After the adjustment is completed, calculate the hedging degree between the single index and the comprehensive score again until the hedging degree H is within the preset hedging degree threshold. Finally, adjust the comprehensive threshold according to the index with the adjusted weight; when the hedging degree is normal, further analyze the tax risk and issue a warning.

[0076] It should be noted that by judging the development status of the enterprise and peer enterprises, screening out peer enterprises with the same status as the enterprise, calculating the average of the single thresholds of these same-type enterprises and using it as the final threshold, the reference standard can be closely tailored to the actual situation of the enterprise, thereby enhancing the accuracy of tax warning.

[0077] In this embodiment, a preset hedging degree threshold is set, the correlation coefficient R between the score of each single index and the comprehensive score is calculated, and the hedging degree H is calculated based on the obtained correlation coefficient. When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. When the hedging degree is abnormal, further adjust the weights of each index. After the adjustment is completed, calculate the hedging degree between the single index and the comprehensive score again until the hedging degree H is within the preset hedging degree threshold. The processing process is as follows:

[0078] Calculate the correlation coefficient R between the score of each single index and the comprehensive score according to the correlation coefficient. The specific formula is: H = ∣R·Si - wi·S a ∣, where Si is the specific score of a single index and wi is the specific weight of a single index;

[0079] When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. When the hedging degree is abnormal and the correlation coefficient R > 0, adjust the single weight, wi new = wi + α, where wi new is the adjusted weight and α is the adjustment coefficient. Then adjust the other weights (j ≠ i), where wj is the weight of other indexes and wj new is the adjusted weight of other indexes; when the hedging degree is abnormal and the correlation coefficient R < 0, adjust the single weight, wi new = wi - α, and then adjust the other weights (j ≠ i);

[0080] Using the adjusted weights, recalculate the comprehensive score, and calculate the hedging degree between the comprehensive score and the individual indicators again. When the hedging degree still exceeds the threshold range, repeat the adjustment of the indicator weights until the hedging degree is within the threshold range.

[0081] It should be noted that by calculating the hedging degree to dynamically adjust the weights, it can more accurately reflect the actual role of each indicator in different scenarios, making the comprehensive evaluation result highly consistent with the actual situation. At the same time, it can also timely identify those indicators that are abnormally hedged with the comprehensive situation. Adjusting their weights can reduce the evaluation deviation caused by abnormal fluctuations of individual indicators and enhance the early warning and response capabilities to potential risks.

[0082] Compare the individual indicators with the individual thresholds. When the individual indicators do not exceed the individual thresholds, calculate the comprehensive score. When the comprehensive score does not exceed the comprehensive threshold, continuously monitor the enterprise data; when the comprehensive score exceeds the threshold, use time series analysis to analyze each individual indicator, locate the specific anomalies, and conduct risk prediction. Finally, transmit the risk prediction results to the enterprise data monitoring module and implement dynamic enhanced monitoring of the relevant enterprise indicators;

[0083] In this embodiment, when the value of an individual indicator exceeds the individual threshold, a red warning is triggered. When the comprehensive risk value triggers the comprehensive score, an orange warning is triggered. When both the red warning and the orange warning are triggered, directly analyze the individual indicator that triggers the warning. When the red warning is not triggered and the orange warning is triggered, use time series analysis to analyze each individual indicator, locate the specific anomalies, and conduct risk prediction. Finally, transmit the risk prediction results to the enterprise data monitoring module and implement dynamic enhanced monitoring of the relevant enterprise indicators. The processing process is as follows:

[0084] Compare the individual indicators with the individual thresholds. When an individual indicator exceeds the individual threshold, a red warning is triggered. When the red warning is triggered, there is no need to calculate the comprehensive score, and the orange warning is directly triggered. Further analyze the indicator and formulate countermeasures;

[0085] It should be noted that when a red warning is triggered as soon as an individual indicator exceeds the threshold, and the orange warning is directly triggered by skipping the calculation of the comprehensive score when the red warning is triggered, it can effectively simplify the processing process, thus facilitating direct focus on the problem indicators for in-depth analysis and formulation of countermeasures;

[0086] In this embodiment, when the comprehensive score exceeds the threshold, use time series analysis to analyze each individual indicator, locate the specific anomalies, and conduct risk prediction. Finally, transmit the risk prediction results to the enterprise data monitoring module and implement dynamic enhanced monitoring of the relevant enterprise indicators. The processing process is as follows:

[0087] Collect the historical data of each individual indicator, and for the time series X of each individual indicator tDecompose it to isolate the trend term T t , the seasonal term S t and the residual term R t , and the specific formula is: X t = T t + S t + R t . Set the residual threshold as μR ± 3σR, where μR is the historical residual. When the residual term R t exceeds the confidence interval ±3σ, it is marked as an outlier. Further calculate the seasonal intensity Q corresponding to the outlier to determine whether the current fluctuation conforms to the historical seasonal pattern. The specific formula is: When Q > 0.6, it indicates a strong seasonal influence. When Q < 0.3, it indicates random fluctuations, and the outliers are screened accordingly;

[0088] Conduct risk prediction based on the screened outliers and transmit the prediction results to the enterprise data monitoring module for dynamic enhanced monitoring.

[0089] In this embodiment, the process of conducting risk prediction based on the screened outliers and transmitting the prediction results to the enterprise data monitoring module for dynamic enhanced monitoring is as follows:

[0090] Extract the slope of the trend term, the abnormal frequency and amplitude of the residuals, and at the same time calculate the industry deviation based on the data of peer enterprises. Combine these data to construct the eigenvector;

[0091] Use the SARIMAX model for modeling and prediction to predict the values of the indicators in the future time. At the same time, output the risk probability through the loss function and obtain the risk time window, and finally integrate and output the specific risks.

[0092] In the embodiments provided by the present invention, it should be understood that the disclosed devices, apparatuses and methods can be implemented in other ways. For example, the device embodiments described above are only illustrative. For example, the division of modules is only a logical function division, and there may be other division methods in actual implementation; the modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the method of this embodiment.

[0093] The above embodiments are only used to illustrate the technical methods of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical methods of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An enterprise tax risk monitoring and analysis system based on big data, characterized in that: It includes an enterprise data monitoring module, a big data search module, a business division correction module, and a tax risk analysis module; The enterprise data monitoring module is used to monitor and collect the internal business data of the enterprise; The big data search module is used to search for tax regulations and tax information data of peer enterprises; The business division correction module receives the searched tax regulation clauses, judges the internal business classification situation of the enterprise. When the classification is incorrect, it re-divides the business and then transmits the tax business and related data to the tax risk analysis module; When the classification is correct, it directly transmits the tax business-related data; The tax risk analysis module is responsible for tax risk analysis, early warning and prediction. First, it divides single indicators and conducts single and comprehensive scoring. Secondly, it sets single-indicator and comprehensive risk thresholds based on the publicly available tax data of peer enterprises. When a single indicator exceeds the standard, it issues a red early warning and locks the abnormal indicator. When the comprehensive indicator exceeds the standard, it issues an orange early warning, starts time-series analysis to locate the risk source, predicts the trend, and pushes the results to the enterprise data monitoring module to implement dynamic enhanced monitoring.

2. The enterprise tax risk monitoring and analysis system based on big data according to claim 1, characterized in that: The process of the business division correction module receiving the searched tax regulation clauses, judging the internal business classification situation of the enterprise. When the classification is incorrect, it re-divides the business and then transmits the tax business and related data to the tax risk analysis module; when the classification is correct, it directly transmits the tax business-related data is as follows: Parse the received tax regulation clauses, extract the key information for defining the business type, and compare the collected enterprise business data with the key information of the tax regulation clauses to judge whether the tax business and non-tax business within the enterprise are correctly divided. When the division is correct, transmit the relevant information of the tax business within the enterprise to the tax risk analysis module. When the division is incorrect, re-classify the tax business and non-tax business within the enterprise according to the key information of the tax regulations, correct the division scope of the tax business and non-tax business within the enterprise, and transmit the relevant information of the corrected tax business to the tax risk analysis module.

3. The enterprise tax risk monitoring and analysis system based on big data according to claim 1, characterized in that: The process of the tax risk analysis module being responsible for tax risk analysis, early warning and prediction. First, it divides single indicators and conducts single and comprehensive scoring. Secondly, it sets single-indicator and comprehensive risk thresholds based on the publicly available tax data of peer enterprises. When a single indicator exceeds the standard, it issues a red early warning and locks the abnormal indicator. When the comprehensive indicator exceeds the standard, it issues an orange early warning, starts time-series analysis to locate the risk source, predicts the trend, and pushes the results to the enterprise data monitoring module to implement dynamic enhanced monitoring is as follows: Divide tax indicators, including tax burden rate, gross profit margin, asset-liability ratio, invoice issuance rate, and sales volume change rate, and analyze the correlation between the enterprise's past tax risk events and each indicator, assign high weights to the indicators with high correlation, and assign low weights to the indicators with low correlation; Receive the relevant data of peer enterprises, judge the status of peer enterprises, including the initial stage, the rising stage, the stable stage, and the decline stage, select the type that suits the enterprise, calculate the average of the single thresholds of the same type of enterprises, use the average as the final threshold, and at the same time set the comprehensive threshold according to the current weights of each indicator, and analyze the relationship between the comprehensive threshold and the weights of each indicator; Calculate the score values of the indicators and the comprehensive risk values of each indicator. At the same time, calculate the hedging degree between the individual indicator and the comprehensive score. When the hedging degree is abnormal, adjust the weights of each indicator to keep the hedging degree within the normal range. Finally, modify the comprehensive threshold according to the adjusted weights. When the value of an individual indicator exceeds the individual threshold, a red warning is triggered. When the comprehensive risk value triggers the comprehensive score, an orange warning is triggered. When both the red warning and the orange warning are triggered, directly analyze the individual indicators that trigger the warning. When the red warning is not triggered and the orange warning is triggered, use time series analysis to analyze each individual indicator, locate the specific anomalies, and conduct risk prediction. Finally, transmit the risk prediction results to the enterprise data monitoring module and implement dynamic enhanced monitoring of the enterprise's relevant indicators.

4. The enterprise tax risk monitoring and analysis system based on big data according to claim 3, characterized in that The process of dividing tax indicators, including tax burden rate, gross profit margin, asset - liability ratio, invoice issuance rate, and sales volume change rate, and analyzing the past tax risk events of the enterprise, specifically analyzing the indicators with high correlation with the events, and assigning high weights to the indicators with high correlation and low weights to the indicators with low correlation is as follows: Divide the tax indicators within the enterprise into tax burden rate, gross profit margin, asset - liability ratio, invoice issuance rate, and sales volume change rate. Calculate the correlation coefficient between each indicator and the occurrence frequency of the past tax risk events of the enterprise to measure the linear correlation degree. Finally, normalize the absolute value of the correlation coefficient into weights to ensure that the sum of the weights is 1. The specific formula is: Among them, w i is the calculated index weight, r i ∣ is the absolute value of the correlation coefficient of index i; Use the calculated weights as the weights of the specific indicators. The indicators with high correlation are assigned high weights, and the indicators with low correlation are assigned low weights.

5. The enterprise tax risk monitoring and analysis system based on big data according to claim 3, characterized in that, Receive the relevant data of peer enterprises, judge the status of peer enterprises, including the initial stage, growth stage, stable stage, and decline stage. Select the type that suits the enterprise, calculate the average value of the individual thresholds of enterprises of the same type, and use the average value as the final threshold. At the same time, set the comprehensive threshold according to the current weights of each indicator, and analyze the relationship between the comprehensive threshold and the weights of each indicator. The process of calculating the score values of the indicators and the comprehensive risk values of each indicator, calculating the hedging degree between the individual indicator and the comprehensive score, adjusting the weights of each indicator when the hedging degree is abnormal to keep the hedging degree within the normal range, and finally modifying the comprehensive threshold according to the adjusted weights is as follows: Based on the data of peer enterprises received, judge the types of peer enterprises and the enterprise itself. And when the sales volume change rate > 20%, and the gross profit margin < industry average, it is classified as the initial stage; when the sales volume change rate > 10%, and the asset - liability ratio < 60%, it is classified as the growth stage; when the sales volume change rate is ±5%, and the tax burden rate is stable, it is classified as the growth stage; when the sales volume change rate < - 5%, and the asset - liability ratio > 80%, it is classified as the decline stage. After classifying peer enterprises and the enterprise itself, select the enterprises of the same category as the enterprise, calculate the individual indicator thresholds of the enterprises of the same category, and use the calculated individual thresholds as the warning boundaries of the enterprise. At the same time, set the comprehensive threshold in combination with the historical data of the enterprise, and analyze the relationship between the comprehensive threshold and the proportion of the weights of individual indicators. Determine the reasonable range of each indicator based on the data of peer enterprises. When a single indicator is within the reasonable range, the indicator obtains the basic score. For every 1% that a single indicator exceeds the reasonable range, a deduction is made. Finally, the specific value of each single indicator is obtained; Calculate the comprehensive risk value S based on the obtained values of individual indicators a , and the specific formula is as follows: w1·s1 + w2·s2 + w3·s3 + w4·s4 + w5·s5 = S a ; Among them, w1, w2, w3, w4, and w5 are the weights of each indicator, and s1, s2, s3, s4, and s5 are the specific scores of each indicator; Preset the hedging degree threshold, calculate the correlation coefficient R between the score of each single indicator and the comprehensive score, and calculate the hedging degree H based on the obtained correlation coefficient. When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. When the hedging degree is abnormal, further adjust the weights of each indicator. After the adjustment is completed, calculate the hedging degree between the single indicator and the comprehensive score again until the hedging degree H is within the preset hedging degree threshold. Finally, adjust the comprehensive threshold according to the indicators with adjusted weights; when the hedging degree is normal, further analyze the tax risk and issue a warning.

6. The enterprise tax risk monitoring and analysis system based on big data according to claim 5, characterized in that For the preset hedging degree threshold, calculate the correlation coefficient R between the score of each single indicator and the comprehensive score, and calculate the hedging degree H based on the obtained correlation coefficient. When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. When the hedging degree is abnormal, further adjust the weights of each indicator. After the adjustment is completed, calculate the hedging degree between the single indicator and the comprehensive score again until the hedging degree H is within the preset hedging degree threshold. The processing process is as follows: Calculate the correlation coefficient R between the score of each individual index and the comprehensive score according to the correlation coefficient. The specific formula is: H = ∣R·Si - wi·S a ∣, where Si is the specific score of a single index and wi is the specific weight of a single index; When the hedging degree H > the preset hedging degree threshold, it indicates that the hedging degree is abnormal. Is the hedging degree abnormal and the correlation coefficient R > 0? Adjust the individual weight, wi new = wi + α, where wi new is the adjusted weight, α is the adjustment coefficient, and then adjust other weights (j ≠ i), where wj is the weight of other indicators, wj new is the adjusted weight of other indicators; Is the hedge degree abnormal and the correlation coefficient R < 0? Adjust the individual weight, wi new = wi - α, and then adjust the other weights (j ≠ i); Use the adjusted weights to recalculate the comprehensive score, and calculate the hedging degree between the comprehensive score and the single indicator again. When the hedging degree still exceeds the threshold range, repeat the adjustment of the indicator weights until the hedging degree is within the threshold range.

7. A big data-based enterprise tax risk monitoring and analysis system according to claim 3, characterized in that, When the value of a single indicator exceeds the single threshold, a red warning is triggered. When the comprehensive risk value triggers the comprehensive score, an orange warning is triggered. When both the red warning and the orange warning are triggered, directly analyze the single indicator that triggers the warning. When the red warning is not triggered and the orange warning is triggered, use time series analysis to analyze each single indicator, locate the specific abnormality, and conduct risk prediction. Finally, transmit the risk prediction result to the enterprise data monitoring module and implement dynamic enhanced monitoring on the relevant indicators of the enterprise. The processing process is as follows: Compare the single indicator with the single threshold. When the single indicator exceeds the single threshold, a red warning is triggered. When the red warning is triggered, there is no need to calculate the comprehensive score, and the orange warning is directly triggered. Further analyze the indicator and formulate countermeasures; Compare the single indicator with the single threshold. When the single indicator does not exceed the single threshold, calculate the comprehensive score. When the comprehensive score does not exceed the comprehensive threshold, continuously monitor the enterprise data; when the comprehensive score exceeds the threshold, use time series analysis to analyze each single indicator, locate the specific abnormality, and conduct risk prediction. Finally, transmit the risk prediction result to the enterprise data monitoring module and implement dynamic enhanced monitoring on the relevant indicators of the enterprise.

8. An enterprise tax risk monitoring and analysis system based on big data according to claim 7, characterized in that, When the comprehensive score exceeds the threshold, time series analysis is used to analyze each individual indicator, locate specific anomalies, and conduct risk prediction. Finally, the risk prediction results are transmitted to the enterprise data monitoring module, and the dynamic enhanced monitoring process for relevant enterprise indicators is as follows: Collect historical data for each individual indicator, and for the time series X of each individual indicator t perform decomposition to isolate the trend term T t , the seasonal term S t and the residual term R t . The specific formula is: X t = T t + S t + R t . Set the residual threshold to μR ± 3σR, where μR is the historical residual. When the residual term R t exceeds the confidence interval of ±3σ, it is marked as an outlier. Further calculate the seasonal intensity Q corresponding to the outlier to determine whether the current fluctuation conforms to the historical seasonal pattern. The specific formula is: When Q > 0.6, it indicates a strong seasonal influence. When Q < 0.3, it indicates random fluctuations, and the outliers are screened accordingly; Risk prediction is carried out based on the screened abnormal points, and the prediction results are transmitted to the enterprise data monitoring module for dynamic enhanced monitoring.

9. The enterprise tax risk monitoring and analysis system based on big data according to claim 8, wherein, The process of conducting risk prediction based on the screened abnormal points and transmitting the prediction results to the enterprise data monitoring module for dynamic enhanced monitoring is as follows: Extract the slope of the trend term, the abnormal frequency and amplitude of the residuals, and at the same time calculate the industry deviation degree according to the data of peer enterprises, and construct eigenvectors by combining these data; Use the SARIMAX model for modeling and prediction to predict the values of indicators in the future time. At the same time, output the risk probability through the loss function and obtain the risk time window, and finally integrate and output the specific risks.

Images