Risk assessment and disposal method based on false opening behavior of taxpayer

Abstract

A taxpayer false opening behavior-based risk assessment and disposal method comprises the steps of S1, determining a to-be-assessed taxpayer list which comprises at least one taxpayer, and obtaining tax data and attribute data of the taxpayer according to the to-be-assessed taxpayer list; s2, cleaning and preprocessing the tax data and attribute data of the taxpayer, and then constructing a plurality of risk indexes of the taxpayer according to practical experience; s3, establishing an association relationship and a risk relationship between the taxpayers according to the upstream and downstream transaction relationship between the taxpayers; s4, a risk scoring system is constructed, the risk scoring system is provided with a risk scoring model, the risk scoring model is used for calculating a business risk score S1 and an attribute risk score S2 of the taxpayer, and then a comprehensive risk score S of the taxpayer is obtained through calculation; and S5, carrying out real-time tracking and data analysis on the business risk score S1 and the attribute risk score S2 of the taxpayer by the risk scoring model, and adjusting numerical values of weights a, b, c, e, f and g in real time.

Description

Technical Field

[0001] The present invention relates to the field of tax control technology, and in particular to a risk assessment and disposal method based on taxpayers' false invoice behavior. Background Art

[0002] With the continuous reform and digital transformation of the global tax collection and management system, tax authorities are increasingly relying on advanced information technology to improve the efficiency and quality of tax collection and management. In recent years, China has vigorously promoted the Golden Tax Project, especially the full launch of the Golden Tax Phase IV, marking that tax management has entered a new era of intelligence and data. The Golden Tax Phase IV is centered on tax big data, and through the construction of a comprehensive tax information platform and risk early warning system, it has achieved real-time monitoring and accurate analysis of taxpayers' behavior.

[0003] Currently, there are relatively few risk assessment patents in the field of false VAT invoices, and existing patents mainly focus on using simple data indicators to measure taxpayers' risks and conduct scoring and rating, which cannot comprehensively and accurately assess taxpayers' false invoice behavior. Summary of the invention

[0004] The present invention provides a risk assessment and disposal method based on taxpayers' false invoice behavior, so as to solve the technical problems existing in the above-mentioned prior art.

[0005] To achieve the above-mentioned purpose, the present invention provides a risk assessment and disposal method based on taxpayers' false invoice behavior, which includes:

[0006] S1: Determine a list of taxpayers to be assessed, the list of taxpayers to be assessed includes at least one taxpayer, and obtain the tax data and attribute data of the taxpayer according to the list of taxpayers to be assessed;

[0007] S2: Clean and pre-process the taxpayer’s tax data and attribute data, and then construct multiple taxpayer risk indicators based on practical experience;

[0008] S3: Establish the association relationship and risk relationship between taxpayers based on the upstream and downstream transaction relationships between taxpayers;

[0009] S4: Construct a risk scoring system. The risk scoring system has a risk scoring model. The risk scoring model is used to calculate the taxpayer's business risk score S1 and attribute risk score S2, and then calculate the taxpayer's comprehensive risk score S, where:

[0010] S=S1+S2,

[0011] S1=aD+bR+cH,

[0012] S2=e*region score+f*industry score+g*business scope score,

[0013] D is the taxpayer's risk index score, R is the taxpayer's risk relationship score, H is the historical risk assessment score, a, b, c, e, f, g are the corresponding weights,

[0014] The taxpayer's risk index score D is calculated as follows:

[0015] d i is the value of the taxpayer’s i-th risk indicator. The taxpayer has m risk indicators in total.

[0016] r j is the value of the jth risk relationship of the taxpayer. There are n risk relationships in total for the taxpayer.

[0017] H = 60% h1 + 30% h2 + 10% h3, where h1, h2, and h3 are the historical risk assessment scores of the last, the last, and the last respectively;

[0018] S5: The risk scoring model conducts real-time tracking and data analysis of the taxpayer’s business risk score S1 and attribute risk score S2, and adjusts the values ​​of weights a, b, c, e, f, and g in real time.

[0019] In one embodiment of the present invention, the taxpayer's tax data includes invoice data, deduction data, certification data and declaration data, and the taxpayer's attribute data includes the taxpayer's region, industry, business scope, bank transaction flow data and transaction statement data.

[0020] In one embodiment of the present invention, the taxpayer's tax data and attribute data include structured data, semi-structured data and unstructured data.

[0021] In one embodiment of the present invention, the cleaning and preprocessing of the taxpayer's tax data and attribute data includes removing incomplete, erroneous, and duplicate data and filling in missing values.

[0022] In one embodiment of the present invention, the risk indicators of taxpayers include abnormal growth rate of invoice amount, profit volatility, tax burden rate, difference between purchase and sales items, cross-regional transaction ratio, and ratio of voided and cancelled invoices.

[0023] In one embodiment of the present invention, the initial values ​​of weights a, b, c, e, f, and g are determined by industry experts based on experience or using a machine learning model.

[0024] In one embodiment of the present invention, each risk indicator of a taxpayer is set with a preset threshold. When the i-th risk indicator of a taxpayer is greater than the preset threshold, it is determined that there is a risk and the risk indicator is set according to the interval of the risk indicator. i Assignment.

[0025] In one embodiment of the present invention, the region score, industry score and business scope score are determined by pre-set scoring rules.

[0026] In one embodiment of the present invention, a risk level rule is preset to assess the taxpayer's risk according to the interval of the taxpayer's comprehensive risk score S. The risk level rule divides taxpayers into low risk, medium risk and high risk.

[0027] In one embodiment of the present invention, in step S3, a graph database and graph computing technology are used to construct the upstream and downstream transaction relationships between taxpayers, and an LSTM neural network model is used to establish the time dimension relationship, logical relationship, co-reference relationship and causal relationship between taxpayers.

[0028] The risk assessment and disposal method based on taxpayers' false invoice behavior provided by the present invention has the following beneficial effects:

[0029] (1) Improving the accuracy of risk assessment: By integrating multi-source data (including tax, industry and commerce, banks, etc.) and adopting advanced data processing technologies and algorithms (such as graph databases, graph computing, neural network models, etc.), the present invention can more accurately identify taxpayers' risk of false invoicing and reduce the possibility of misjudgment and omission.

[0030] (2) Enhance the timeliness of risk warning: The risk scoring system and real-time data analysis function in the solution enable the tax department to respond quickly to potential false invoicing, issue warnings in a timely manner, and take effective measures to curb the spread of risks.

[0031] (3) Optimizing the allocation of tax management resources: By conducting risk rating on taxpayers, the tax authorities can allocate management resources more reasonably, focusing limited resources on the supervision and review of high-risk taxpayers, and improving the efficiency and effectiveness of tax management.

[0032] (4) Improving the scientific nature of tax decisions: The present invention has powerful data support and risk analysis functions, providing a scientific basis for tax authorities in formulating tax policies and optimizing tax collection and management processes, which helps to improve the scientific nature and rationality of tax decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0034] Figure 1 A flow chart of a risk assessment and disposal method based on taxpayers' false invoice behavior according to an embodiment of the present invention;

[0035] Figure 2 is the risk indicator d i Assignment diagram. DETAILED DESCRIPTION

[0036] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0037] At present, there are relatively few risk assessment technologies in the field of VAT fraud, and the existing technologies mainly focus on using simple data indicators to measure taxpayers' risks and conduct scoring and rating. The present invention combines the characteristics of VAT business, introduces the basic attributes of taxpayers (such as region, industry, business scope, etc.) and transaction relationships (loop invoicing, abnormal aggregation, etc.) as risk measurement factors, and continuously optimizes the risk scoring model by adopting a weighting mechanism and a dynamic feedback mechanism, so as to be able to more comprehensively and accurately assess taxpayers' risks.

[0038] This invention relies on the rapid development of big data and artificial intelligence technology. Big data technology enables tax departments to collect, store, process and analyze massive amounts of tax data, thereby revealing the patterns and risks hidden behind the data. Artificial intelligence technology uses machine learning, deep learning and other algorithms to intelligently process big data, automatically identify and predict potential tax risks, and provide scientific decision-making support for tax departments.

[0039] The present invention monitors and evaluates taxpayers' tax behaviors in real time by setting risk indicators and evaluation models. Once a potential risk point is found, the early warning mechanism is immediately triggered, and risk evaluation reports and disposal suggestions are automatically generated. Based on these reports and suggestions, the tax department can quickly take corresponding regulatory measures to prevent the risk from further expanding.

[0040] The present invention provides a risk assessment and disposal method based on taxpayers' false invoice behavior, such as Figure 1 The flowchart of the risk assessment and disposal method based on taxpayers' false invoice behavior according to one embodiment of the present invention is shown. The method includes:

[0041] S1: Determine a list of taxpayers to be assessed, the list of taxpayers to be assessed includes at least one taxpayer, and obtain the tax data and attribute data of the taxpayer according to the list of taxpayers to be assessed;

[0042] In this embodiment, the taxpayer's tax data includes invoice data, deduction data, certification data and declaration data, and the taxpayer's attribute data includes the taxpayer's region, industry, business scope, bank transaction flow data and transaction statement data. Among them, the invoice data and declaration data are internal data of the tax department, and the taxpayer's region, industry, business scope, and bank transaction flow data and transaction statement data registered with the industrial and commercial department are external data. The data collected in this step can further include third-party data disclosed on the Internet. The present invention realizes the interconnection and interoperability of these data through data fusion technology, providing comprehensive and accurate information support for tax risk assessment and early warning.

[0043] Taxpayers' tax data and attribute data include structured data, semi-structured data and unstructured data.

[0044] S2: Clean and pre-process the taxpayer’s tax data and attribute data, and then construct multiple taxpayer risk indicators based on practical experience;

[0045] Cleaning and preprocessing taxpayers' tax data and attribute data include removing incomplete, erroneous, and duplicate data and filling in missing values ​​to ensure the accuracy and completeness of the data.

[0046] Risk indicators are constructed based on the rich experience of audit experts. The risk indicators of taxpayers include abnormal growth rate of invoice amount, profit volatility, tax burden rate, difference between purchase and sales goods, cross-regional transaction ratio, and proportion of invalidated and cancelled invoices. The above risk indicators can reflect the operating conditions and invoicing behavior of taxpayers.

[0047] S3: Establish the association relationship and risk relationship between taxpayers based on the upstream and downstream transaction relationships between taxpayers;

[0048] In the field of value-added tax, gangs committing fraudulent VAT invoices have always been a headache for the tax authorities. The offenders set up multiple shell taxpayers or multiple taxpayers jointly commit crimes. There is a strict upstream and downstream division of labor between taxpayers. They fabricate product acquisition and processing business chains, and then falsely issue VAT invoices to falsely deduct VAT. In this type of phenomenon, there are complex transaction relationships between taxpayers. These complex transaction relationships cannot be measured by risk indicators. Therefore, it is necessary to establish the taxpayer's association relationship through upstream and downstream transaction relationships. The establishment of risk relationships provides a basis for the subsequent risk relationship scoring.

[0049] In step S3, the graph database and graph computing technology are used to build the upstream and downstream transaction relationships between taxpayers, and the LSTM neural network model is used to establish the time dimension relationship, logical relationship, co-reference relationship and causal relationship between taxpayers.

[0050] In view of the common gang crime phenomenon in the field of value-added tax, this step uses graph database and graph computing technology to establish a complex transaction relationship network between taxpayers. For tax data, the taxpayers are used as nodes, transactions between taxpayers are used as edges, transaction flows are used as the direction of the edges, and transaction amounts are used as the weights of the edges to realize the construction of the graph network. At the same time, neural network models such as LSTM are used to inject time dimensions and logical relationships of things, such as co-reference relationships and causal relationships, to make the relationship map more three-dimensional and accurate. Finally, algorithms such as SVM, K-means clustering or LPA community discovery are used to dig out risk relationships such as abnormal aggregation, blacklist communities, invoicing, and loop invoicing, and they are intuitively presented in the form of knowledge graphs to facilitate tax personnel to track and trace.

[0051] S4: Construct a risk scoring system. The risk scoring system has a risk scoring model. The risk scoring model is used to calculate the taxpayer's business risk score S1 and attribute risk score S2, and then calculate the taxpayer's comprehensive risk score S, where:

[0052] S=S1+S2,

[0053] S1=aD+bR+cH,

[0054] S2=e*region score+f*industry score+g*business scope score,

[0055] D is the taxpayer's risk index score, R is the taxpayer's risk relationship score, H is the historical risk assessment score, a, b, c, e, f, g are the corresponding weights,

[0056] The taxpayer's risk index score D is calculated as follows:

[0057] d iis the value of the taxpayer’s i-th risk indicator. The taxpayer has m risk indicators in total.

[0058] r j is the value of the j-th risk relationship of the taxpayer. There are n risk relationships of the taxpayer in total. The value of the j-th risk relationship is read from the risk relationship constructed in step S3. This is a commonly used technical means in this field and will not be repeated here.

[0059] H = 60% h1 + 30% h2 + 10% h3, where h1, h2, and h3 are the historical risk assessment scores of the last, the last, and the last respectively;

[0060] like Figure 2 The following is a schematic diagram of risk indicator assignment. Each taxpayer’s risk indicator has a preset threshold, such as Figure 2 The first row 1~i is a total of i risk indicators, the 1st to 3rd risk indicators are u, v, w respectively, and the U, V, W in the third row are the thresholds of the 1st to 3rd risk indicators respectively. When the i-th risk indicator of the taxpayer is greater than the preset threshold, it is determined that there is a risk and the risk indicator is adjusted according to the interval of the risk indicator. i Assignment, such as Figure 2 As shown, taking the first risk indicator as an example, when the first risk indicator u is greater than U, it is determined that there is a risk, and the intervals greater than U are divided into U1, U2 and U3. If u falls within the U2 interval, the value of d1 is set to the corresponding value. Figure 2 Other risk indicators in the same way are assigned corresponding values. It should be noted that the "risk indicator" and "value of risk indicator" mentioned above are not the same thing. The risk indicator refers to the actual value collected or calculated for a certain risk indicator, while the value of risk indicator is the value obtained by comparing the risk indicator with the interval determined by the threshold and confirming the interval where the risk indicator is located. It is an artificial value.

[0061] For example, when the risk indicator is the proportion of the amount of red-invoices, if the amount of red-invoices in this month accounts for more than 50% of the total invoice amount, it is considered to be risky. Therefore, the threshold of this risk indicator is 50%. The intervals greater than 50% are divided into (50%, 60%], (60%, 70%], (70%, 80%] and (80%, 90%] (the case greater than 90% is rare and therefore ignored). In the interval (50%, 60%], the value of the risk indicator is recorded as 10 points; in the interval (60%, 70%], the value of the risk indicator is recorded as 20 points; in the interval (70%, 80%], the value of the risk indicator is recorded as 30 points; in the interval (80%, 90%], the value of the risk indicator is recorded as 40 points; in the interval (80%, 90%], the value of the risk indicator is recorded as 50 points. For example, in July 2023, the proportion of the amount of invoices offset by taxpayer A is 75%, then the value of this risk indicator of taxpayer A is 30 points.

[0062] Since taxpayers who issue false VAT invoices often show clustering characteristics in terms of region, industry and business scope, the present invention incorporates basic attributes such as the taxpayer's region, industry and business scope into the scoring system. S2 = e*region score + f*industry score + g*business scope score. The region score, industry score and business scope score are determined by pre-set scoring rules. For the regional score, for example, the risk score of Shanxi Province is 10, the risk score of Hubei Province is 5, and the risk score of Beijing is 0.

[0063] S5: The risk scoring model conducts real-time tracking and data analysis of the taxpayer’s business risk score S1 and attribute risk score S2, and adjusts the values ​​of weights a, b, c, e, f, and g in real time.

[0064] In this embodiment, the initial values ​​of weights a, b, c, e, f, and g are determined by industry experts based on experience or using a machine learning model.

[0065] In this embodiment, a risk level rule is pre-set to assess the taxpayer's risk based on the interval of the taxpayer's comprehensive risk score S. The risk level rule divides taxpayers into low risk, medium risk and high risk. For example, the intervals corresponding to low risk, medium risk and high risk are (0, 40], (40, 60] and (60, 80] respectively. The comprehensive risk score S of taxpayer A is 56.8 points, which is within the interval of (40, 60]. Therefore, the risk level corresponding to taxpayer A is medium risk. In other embodiments, there can be more levels of risk levels, that is, taxpayers can be divided more finely according to their risks.

[0066] After assessing the risk level of the taxpayer, the tax bureau staff can refer to the risk level and score details to handle the taxpayer's risks, and give different treatment methods to the taxpayer based on the expert experience of the tax bureau staff. For example, taxpayers with ultra-high risks will be pushed to the audit team as a basis for case selection; high-risk taxpayers will be directly issued warning notices; higher-risk taxpayers will be pushed to the competent tax authorities, and their competent tax authorities will continue to review their financial data and declaration data; the rest of the lower and low risks will not be handled or will continue to follow up.

[0067] After the tax bureau personnel issue risk disposal measures, the attribute risk scoring parameters will be adjusted according to the risk disposal results. This is achieved here through a pre-set online learning function, that is, after the above-mentioned scoring and rating model is deployed, the online learning function will be turned on, and the reward mechanism of reinforcement learning will be set so that the model can adjust the weight according to the actual disposal results. For example, if it is found that the disposal rate of taxpayers in a certain area has decreased, the model will automatically reduce the corresponding weight of the risk score of the area to improve the accuracy and effectiveness of the evaluation model. Through continuous feedback and automatic adjustment mechanisms, it is ensured that the present invention can keep up with the new situation and new requirements of tax management.

[0068] The risk assessment and disposal method based on taxpayers' false invoice behavior provided by the present invention has the following beneficial effects:

[0069] (1) Improving the accuracy of risk assessment: By integrating multi-source data (including tax, industry and commerce, banks, etc.) and adopting advanced data processing technologies and algorithms (such as graph databases, graph computing, neural network models, etc.), the present invention can more accurately identify taxpayers' risk of false invoicing and reduce the possibility of misjudgment and omission.

[0070] (2) Enhance the timeliness of risk warning: The risk scoring system and real-time data analysis function in the solution enable the tax department to respond quickly to potential false invoicing, issue warnings in a timely manner, and take effective measures to curb the spread of risks.

[0071] (3) Optimizing the allocation of tax management resources: By conducting risk rating on taxpayers, the tax authorities can allocate management resources more reasonably, focusing limited resources on the supervision and review of high-risk taxpayers, and improving the efficiency and effectiveness of tax management.

[0072] (4) Improving the scientific nature of tax decisions: The present invention has powerful data support and risk analysis functions, providing a scientific basis for tax authorities in formulating tax policies and optimizing tax collection and management processes, which helps to improve the scientific nature and rationality of tax decisions.

[0073] Those skilled in the art can understand that the accompanying drawings are only schematic diagrams of an embodiment, and the modules or processes in the accompanying drawings are not necessarily required to implement the present invention.

[0074] Those skilled in the art can understand that the modules in the device in the embodiment can be distributed in the device in the embodiment according to the description of the embodiment, or can be changed accordingly and located in one or more devices different from the embodiment. The modules in the above embodiment can be combined into one module, or can be further divided into multiple sub-modules.

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A risk assessment and disposal method based on taxpayers' false invoice behavior, characterized in that: include: S1: Determine a list of taxpayers to be assessed, the list of taxpayers to be assessed includes at least one taxpayer, and obtain the tax data and attribute data of the taxpayer according to the list of taxpayers to be assessed; S2: Clean and pre-process the taxpayer’s tax data and attribute data, and then construct multiple taxpayer risk indicators based on practical experience; S3: Establish the association relationship and risk relationship between taxpayers based on the upstream and downstream transaction relationships between taxpayers; S4: Construct a risk scoring system. The risk scoring system has a risk scoring model. The risk scoring model is used to calculate the taxpayer's business risk score S1 and attribute risk score S2, and then calculate the taxpayer's comprehensive risk score S, where: S=S1+S2, S1=aD+bR+cH, S2=e*region score+f*industry score+g*business scope score, D is the taxpayer's risk index score, R is the taxpayer's risk relationship score, H is the historical risk assessment score, a, b, c, e, f, g are the corresponding weights, The taxpayer's risk index score D is calculated as follows: d i is the value of the taxpayer’s i-th risk indicator. The taxpayer has m risk indicators in total. r j is the value of the jth risk relationship of the taxpayer. There are n risk relationships in total for the taxpayer. H = 60% h1 + 30% h2 + 10% h3, where h1, h2, and h3 are the historical risk assessment scores of the last, the last, and the last respectively; S5: The risk scoring model conducts real-time tracking and data analysis of the taxpayer’s business risk score S1 and attribute risk score S2, and adjusts the values ​​of weights a, b, c, e, f, and g in real time.

2. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: The taxpayer's tax data includes invoice data, deduction data, certification data and declaration data; the taxpayer's attribute data includes the taxpayer's region, industry, business scope, bank transaction flow data and transaction statement data.

3. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: Taxpayers' tax data and attribute data include structured data, semi-structured data and unstructured data.

4. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: Cleaning and preprocessing of taxpayers' tax data and attribute data includes removing incomplete, erroneous, and duplicate data and filling in missing values.

5. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: Taxpayers' risk indicators include abnormal growth rate of invoice amount, profit volatility, tax burden rate, difference between purchase and sales items, cross-regional transaction ratio, and proportion of invalidated and cancelled invoices.

6. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: The initial values ​​of weights a, b, c, e, f, and g are determined by industry experts based on experience or using a machine learning model.

7. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: Each taxpayer's risk indicator has a preset threshold. When the taxpayer's i-th risk indicator is greater than the preset threshold, it is determined that there is a risk and the risk indicator is divided into d i Assignment.

8. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: The regional score, industry score and business scope score are determined by pre-set scoring rules.

9. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: A risk level rule is set in advance to assess the taxpayer's risk based on the interval of the taxpayer's comprehensive risk score S. The risk level rule divides taxpayers into low risk, medium risk and high risk.

10. The risk assessment and disposal method based on taxpayers' false invoice behavior according to claim 1 is characterized in that: In step S3, the graph database and graph computing technology are used to build the upstream and downstream transaction relationships between taxpayers, and the LSTM neural network model is used to establish the time dimension relationship, logical relationship, co-reference relationship and causal relationship between taxpayers.

Images