Enterprise tax audit analysis system based on cloud computing

Abstract

The invention relates to the technical field of tax audit analysis, in particular to an enterprise tax audit analysis system based on cloud computing. The method comprises the following steps: acquiring enterprise tax data and preprocessing the enterprise tax data; performing text entity extraction on the preprocessed enterprise tax data through an improved long-short-term memory network model to obtain enterprise tax entities, analyzing mechanism relationships among various types of entities based on a dependency syntactic analysis method, and constructing an entity association graph based on the mechanism relationships; and clustering transaction numbers, commodity names, transaction amounts, transaction quantities, tax rates and tax amounts corresponding to various entities into an entity data pool based on the entity association graph, performing audit analysis on the entity data pool to obtain an audit analysis result, generating a risk monitoring index based on the audit analysis result, and sending the risk monitoring index to a server. And carrying out risk early warning on the enterprise tax based on the risk monitoring index. The enterprise tax auditing efficiency can be improved, the auditing cost is reduced, and the enterprise tax risk of an enterprise is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of tax audit analysis, and in particular to an enterprise tax audit analysis system based on cloud computing. Background Art

[0002] Traditional tax audits mainly rely on manual data processing, which presents the following challenges: low efficiency. Faced with huge financial data and complex tax rules, manual audits are inefficient and prone to errors; high costs: a large amount of human resources are required, increasing the audit costs of the company; high risks. Human factors may lead to inaccurate audit results, increasing the tax risks of the company.

[0003] Therefore, there is an urgent need for a cloud computing-based enterprise tax audit analysis system to solve the above problems. Summary of the invention

[0004] The purpose of the present invention is to provide a cloud computing-based enterprise tax audit analysis system to solve the technical problems of the existing solution that traditional tax audits mainly rely on manual data processing, resulting in low efficiency, high cost and high risk of enterprise tax audits.

[0005] The purpose of the present invention can be achieved through the following technical solutions:

[0006] The enterprise tax audit analysis system based on cloud computing includes a data acquisition module, an entity association graph construction module, a cloud computing audit analysis module and a risk monitoring module;

[0007] The data acquisition module is used to acquire enterprise tax data and pre-process the enterprise tax data, wherein the enterprise tax data includes financial data, tax data and business data;

[0008] The entity association graph construction module is used to extract text entities from the pre-processed corporate tax data through an improved long short-term memory network model to obtain corporate tax entities, classify corporate tax entities into financial entities, tax entities and business entities, analyze the mechanism relationship between entities of each category based on the dependency syntax analysis method, and construct an entity association graph based on the mechanism relationship;

[0009] The cloud computing audit analysis module is used to cluster the transaction numbers, commodity names, transaction amounts, transaction quantities, tax rates and tax amounts corresponding to various entities into entity data pools based on the entity association graph, and to perform audit analysis on the entity data pools to obtain audit analysis results, wherein each entity of each enterprise corresponds to an entity data pool;

[0010] The risk monitoring module is used to generate a risk monitoring index based on the audit analysis results, and to issue risk warnings for corporate taxation based on the risk monitoring index.

[0011] Furthermore, the pre-processing of corporate tax data specifically includes the following processes:

[0012] Data cleaning: Use the is.na function to identify missing values. If the number of missing values ​​is less than the preset number, directly delete the records or variables containing missing values. Use global constants, mean, mode, or regression model methods to fill in missing values. Use statistical analysis methods to identify outliers, correct or delete outliers, and use data deduplication technology to ensure that there are no duplicate records in the data set.

[0013] Data integration: extract the required tax data from various data sources, solve the problem of inconsistent data formats, units, and codes between different data sources, convert the data into a unified format, and load the converted data into a unified data storage device;

[0014] Data reduction: Summarize or aggregate the data to reduce the level of detail, compress the data using principal component analysis techniques, reduce the dimensionality of the data, and extract a representative sample from the original data set for subsequent analysis or modeling.

[0015] Furthermore, the improved long short-term memory network model is used to extract text entities from the preprocessed corporate tax data to obtain corporate tax entities. The corporate tax entities are classified into financial entities, tax entities and business entities, including the following processes:

[0016] The pre-processed corporate tax data is embedded into a vector sequence , Indicates The input vector of words, where word embedding is to randomly initialize the One-hot vector of a single character or word from the embedding matrix to a low-dimensional dense vector;

[0017] The vector sequence Input into the improved long short-term memory network model to obtain the text feature matrix P: the vector sequence Input into the improved long short-term memory network model to obtain the forward hidden layer sequence and the backward hidden layer sequence ,Bundle , Integrate by position to get the final complete hidden layer sequence ,Will The columns are connected to the linear layer, mapped from n-dimension to k-dimension, and the vector dimensions correspond to the length of the label set, respectively, and mapped into the text feature matrix P, where the value of k is the number of standard set labels;

[0018] The feature matrix P is classified and labeled through the CRF model to obtain financial entities, tax entities and business entities.

[0019] Furthermore, the feature matrix P is classified and labeled through the CRF model to obtain financial entities, tax entities and business entities, including the following process:

[0020] Get entity meaning elements based on feature matrix P ,in, represents the i-th word in the j-th annotation;

[0021] By input and vector sequence The corresponding tag sequence , and combined with the state transfer matrix A of the CRF model, we get the vector sequence Scoring :

[0022] ;

[0023] in, Represents the adjacent state transfer matrix score value in the CRF model, Represents the output vector fraction value;

[0024] By dividing the score The probability formula of the model obtained after normalization is as follows:

[0025] ;

[0026] in, represents all possible labeling sequences;

[0027] Select the predictions labeled with the maximum likelihood estimation function to obtain financial entities, tax entities, and business entities, where each entity category corresponds to an output Interval value, the formula is as follows:

[0028] .

[0029] Furthermore, the mechanism relationship between entities of each category is analyzed based on the dependency syntactic analysis method, and the entity association graph is constructed based on the mechanism relationship, including the following processes:

[0030] Dependency extraction: Perform dependency syntactic analysis on sentences containing entities to extract dependency relationships between entities;

[0031] Relationship classification: According to the type of dependency and context information, the relationship between entities is classified into specific mechanism relationships, including causal relationships, parallel relationships, and inclusion relationships;

[0032] Relationship screening: remove noise and irrelevant relationships and retain only those relationships that are useful for building entity association graphs;

[0033] Entity node construction: The identified entities are used as nodes in the graph;

[0034] Relationship edge construction: Based on the extracted mechanism relationship, construct edges connecting entities in the graph and mark the edge types.

[0035] Furthermore, audit analysis is performed on the entity data pool to obtain the audit analysis results, which includes the following process:

[0036] Obtain the transaction number, commodity name, transaction amount, transaction quantity, tax rate and tax amount in the entity data pool, and the registered transaction number, registered commodity name, registered transaction amount, registered transaction quantity, registered tax rate and registered tax amount in the cloud service center;

[0037] Calculate the same rate JY between the transaction number and the registered transaction number, the same rate JE between the transaction amount and the registered transaction amount, the same rate SL between the transaction quantity and the registered transaction quantity, the same rate SW between the tax rate and the registered tax rate, and the same rate SEW between the tax amount and the registered tax amount respectively;

[0038] Calculate the proportion of product name types to registered product name types SP;

[0039] Record JY, JE, SL, SW, SEW and SP as the audit analysis results.

[0040] Furthermore, generating a risk monitoring index based on the audit analysis results and conducting risk warning for corporate taxation based on the risk monitoring index includes the following processes:

[0041] Calculate the corporate tax sameness index SWX based on JY, JE, SL, SW, and SEW:

[0042] ;

[0043] Substitute the corporate tax sameness index SWX and SP into the risk monitoring index calculation formula to obtain the risk monitoring index LMS, the calculation formula is as follows:

[0044] ;

[0045] in, , is the weight coefficient, set by the system;

[0046] When the risk monitoring index When it is greater than the preset threshold, a risk warning is generated and fed back to the enterprise.

[0047] Compared with the existing solutions, the present invention achieves the following beneficial effects:

[0048] The present invention obtains enterprise tax data and pre-processes the enterprise tax data, wherein the enterprise tax data includes financial data, tax data and business data; the pre-processed enterprise tax data is subjected to text entity extraction through an improved long short-term memory network model to obtain enterprise tax entities, and the enterprise tax entities are classified to obtain financial entities, tax entities and business entities; the mechanism relationship between entities of each category is analyzed based on the dependency syntax analysis method, and an entity association map is constructed based on the mechanism relationship; the transaction number, commodity name, transaction amount, transaction quantity, tax rate and tax amount corresponding to each entity are clustered into an entity data pool based on the entity association map, and the entity data pool is audited and analyzed to obtain an audit analysis result, wherein each entity of each enterprise corresponds to an entity data pool; a risk monitoring index is generated based on the audit analysis result, and a risk warning is given to enterprise tax based on the risk monitoring index. The efficiency of enterprise tax audit can be improved, the audit cost can be reduced, and the tax risk of the enterprise can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0050] Figure 1 The present invention is a system block diagram of a cloud computing-based enterprise tax audit analysis system.

[0051] Figure 2 The present invention provides a workflow diagram of a cloud computing-based enterprise tax audit analysis system. DETAILED DESCRIPTION

[0052] 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.

[0053] In addition, the described features, structures or characteristics may be combined in one or more example embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the example embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, steps, etc. may be adopted. In other cases, well-known structures, methods, implementations or operations are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0054] This embodiment provides a cloud computing-based enterprise tax audit analysis system. Figure 1 is a system block diagram of a cloud computing-based enterprise tax audit analysis system according to an embodiment of the present invention. Figure 1 As shown, the system includes a data acquisition module, an entity association graph construction module, a cloud computing audit analysis module and a risk monitoring module;

[0055] The data acquisition module is used to acquire enterprise tax data and pre-process the enterprise tax data, wherein the enterprise tax data includes financial data, tax data and business data;

[0056] The entity association graph construction module is used to extract text entities from the pre-processed corporate tax data through an improved long short-term memory network model to obtain corporate tax entities, classify corporate tax entities into financial entities, tax entities and business entities, analyze the mechanism relationship between entities of each category based on the dependency syntax analysis method, and construct an entity association graph based on the mechanism relationship;

[0057] The cloud computing audit analysis module is used to cluster the transaction numbers, commodity names, transaction amounts, transaction quantities, tax rates and tax amounts corresponding to various entities into entity data pools based on the entity association graph, and to perform audit analysis on the entity data pools to obtain audit analysis results, wherein each entity of each enterprise corresponds to an entity data pool;

[0058] The risk monitoring module is used to generate a risk monitoring index based on the audit analysis results, and to issue risk warnings for corporate taxation based on the risk monitoring index.

[0059] In summary, the present invention obtains enterprise tax data and preprocesses the enterprise tax data, wherein the enterprise tax data includes financial data, tax data and business data; the improved long short-term memory network model is used to extract text entities from the preprocessed enterprise tax data to obtain enterprise tax entities, and the enterprise tax entities are classified to obtain financial entities, tax entities and business entities; the mechanism relationship between entities of each category is analyzed based on the dependency syntax analysis method, and an entity association graph is constructed based on the mechanism relationship; based on the entity association graph, the transaction number, commodity name, transaction amount, transaction quantity, tax rate and tax amount corresponding to each type of entity are clustered into an entity data pool, and the entity data pool is audited and analyzed to obtain the audit analysis result, wherein each entity of each enterprise corresponds to an entity data pool; a risk monitoring index is generated based on the audit analysis result, and a risk warning is issued for enterprise tax based on the risk monitoring index. The efficiency of enterprise tax audit can be improved, the audit cost can be reduced, and the tax risk of the enterprise can be reduced.

[0060] In some embodiments, preprocessing the enterprise tax data specifically includes the following process:

[0061] Data cleaning: Use the is.na function to identify missing values. If the number of missing values ​​is less than the preset number, directly delete the records or variables containing missing values. Use global constants, mean, mode, or regression model methods to fill in missing values. Use statistical analysis methods to identify outliers, correct or delete outliers, and use data deduplication technology to ensure that there are no duplicate records in the data set.

[0062] Data integration: extract the required tax data from various data sources, solve the inconsistency of data format, unit, coding, etc. between different data sources, convert the data into a unified format, and load the converted data into a unified data storage device;

[0063] Data reduction: Summarize or aggregate the data to reduce the level of detail, compress the data using principal component analysis techniques, reduce the dimensionality of the data, and extract a representative sample from the original data set for subsequent analysis or modeling.

[0064] In some embodiments, Figure 2 is a workflow diagram of a cloud computing-based enterprise tax audit analysis system according to an embodiment of the present invention. Figure 2 As shown, the improved long short-term memory network model is used to extract text entities from the preprocessed corporate tax data to obtain corporate tax entities. Classifying corporate tax entities to obtain financial entities, tax entities, and business entities includes the following steps:

[0065] Step S201: construct a vector sequence for the pre-processed corporate tax data through word embedding , Indicates The input vector of words, where word embedding is to randomly initialize the One-hot vector of a single character or word from the embedding matrix to a low-dimensional dense vector;

[0066] Step S202: vector sequence Input into the improved long short-term memory network model to obtain the text feature matrix P: the vector sequence Input into the improved long short-term memory network model to obtain the forward hidden layer sequence and the backward hidden layer sequence ,Bundle , Integrate by position to get the final complete hidden layer sequence ,Will The columns are connected to the linear layer, mapped from n-dimension to k-dimension, and the vector dimensions correspond to the length of the label set, respectively, and mapped into the text feature matrix P, where the value of k is the number of standard set labels;

[0067] Step S203: classify and label the feature matrix P through the CRF model to obtain financial entities, tax entities and business entities.

[0068] Among them, the entity corresponding to the corporate tax data of the present invention can be the taxpayer name, business registration number, balance sheet, income statement, cash flow statement, etc., which are all pre-set by the system.

[0069] It is worth noting that the CRF model (Conditional Random Field) is a discriminative probability model used to predict the corresponding output label sequence for a given input sequence. It combines the characteristics of the maximum entropy model and the hidden Markov model, and captures the contextual information of the sequence data by considering the dependency between labels.

[0070] Furthermore, the feature matrix P is classified and labeled through the CRF model to obtain financial entities, tax entities and business entities, including the following process:

[0071] Get entity meaning elements based on feature matrix P ,in, represents the i-th word in the j-th annotation;

[0072] By input and vector sequence The corresponding tag sequence , and combined with the state transfer matrix A of the CRF model, we get the vector sequence Scoring :

[0073] ;

[0074] in, Represents the adjacent state transfer matrix score value in the CRF model, Represents the output vector fraction value;

[0075] By dividing the score The probability formula of the model obtained after normalization is as follows:

[0076] ;

[0077] in, represents all possible labeling sequences;

[0078] Select the predictions labeled with the maximum likelihood estimation function to obtain financial entities, tax entities, and business entities, where each entity category corresponds to an output Interval value, the formula is as follows:

[0079] .

[0080] In some embodiments, analyzing the mechanism relationship between entities of each category based on the dependency syntax analysis method and constructing an entity association graph based on the mechanism relationship includes the following process:

[0081] Dependency extraction: Perform dependency syntactic analysis on sentences containing entities to extract dependency relationships between entities;

[0082] Relationship classification: According to the type of dependency and context information, the relationship between entities is classified into specific mechanism relationships, including causal relationships, parallel relationships, and inclusion relationships;

[0083] Relationship screening: remove noise and irrelevant relationships and retain only those relationships that are useful for building entity association graphs;

[0084] Entity node construction: The identified entities are used as nodes in the graph;

[0085] Relationship edge construction: Based on the extracted mechanism relationship, construct edges connecting entities in the graph and mark the edge types.

[0086] It is worth mentioning that dependency syntactic analysis is an important technology in natural language processing. Its core idea is to analyze the dependency relationship between words in the grammatical structures such as "subject, predicate, object", "attributive, adverbial, complement" of the sentence. This method is to obtain the structural organization of the sentence by analyzing the interdependence between words in the text sentence. When performing dependency structure analysis on text data, the core verb in the sentence will not be affected by other verbs, and the core verb has the right to dominate other verbs. The sentences of dependency syntactic analysis meet several principles: ① There is a dependency connection between words in the sentence; ② Each phrase in the sentence is related to other phrases as an independent whole; ③ Each sentence has a general dominator, and other sentences depend on the general dominator; ④ There is no dependency relationship between the other words on both sides of the general dominator; ⑤ There are three phrases ABC in the sentence, A is attached to B, C is between A and B, then C phrase is also attached to a component of phrase A and phrase B or directly attached to phrase B; ⑥ Each sentence contains an independent component; ⑦ Each dependency word structure appears only once. The dependency relationships between entities are represented by directed arcs, which eventually form a dependency structure diagram.

[0087] In some embodiments, performing audit analysis on the entity data pool to obtain audit analysis results includes the following process:

[0088] Obtain the transaction number, commodity name, transaction amount, transaction quantity, tax rate and tax amount in the entity data pool, and the registered transaction number, registered commodity name, registered transaction amount, registered transaction quantity, registered tax rate and registered tax amount in the cloud service center;

[0089] Calculate the same rate JY between the transaction number and the registered transaction number, the same rate JE between the transaction amount and the registered transaction amount, the same rate SL between the transaction quantity and the registered transaction quantity, the same rate SW between the tax rate and the registered tax rate, and the same rate SEW between the tax amount and the registered tax amount respectively;

[0090] Calculate the proportion of product name types to registered product name types SP;

[0091] Record JY, JE, SL, SW, SEW and SP as the audit analysis results.

[0092] It is worth mentioning that the cloud service center, as a comprehensive service platform, mainly relies on cloud computing technology to provide various cloud-based computing, storage, network, security and data analysis services to enterprises and individual users.

[0093] Furthermore, generating a risk monitoring index based on the audit analysis results and conducting risk warning for corporate taxation based on the risk monitoring index includes the following processes:

[0094] Calculate the corporate tax sameness index SWX based on JY, JE, SL, SW, and SEW:

[0095] ;

[0096] Substitute the corporate tax sameness index SWX and SP into the risk monitoring index calculation formula to obtain the risk monitoring index LMS, the calculation formula is as follows:

[0097] ;

[0098] in, , is the weight coefficient, set by the system;

[0099] When the risk monitoring index When it is greater than the preset threshold, a risk warning is generated and fed back to the enterprise.

[0100] The above embodiments can be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center by wired (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0101] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0102] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0103] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only some logical function divisions. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0104] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0105] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. The enterprise tax audit analysis system based on cloud computing is characterized by: The system includes a data acquisition module, an entity association graph construction module, a cloud computing audit analysis module, and a risk monitoring module; The data acquisition module is used to acquire enterprise tax data and pre-process the enterprise tax data, wherein the enterprise tax data includes financial data, tax data and business data; The entity association graph construction module is used to extract text entities from the pre-processed corporate tax data through an improved long short-term memory network model to obtain corporate tax entities, classify corporate tax entities into financial entities, tax entities and business entities, analyze the mechanism relationship between entities of each category based on the dependency syntax analysis method, and construct an entity association graph based on the mechanism relationship; The cloud computing audit analysis module is used to cluster the transaction numbers, commodity names, transaction amounts, transaction quantities, tax rates and tax amounts corresponding to various entities into entity data pools based on the entity association graph, and to perform audit analysis on the entity data pools to obtain audit analysis results, wherein each entity of each enterprise corresponds to an entity data pool; The risk monitoring module is used to generate a risk monitoring index based on the audit analysis results, and to issue risk warnings for corporate taxation based on the risk monitoring index.

2. The cloud computing-based enterprise tax audit analysis system according to claim 1 is characterized in that: Preprocessing of corporate tax data specifically includes the following processes: Data cleaning: Use the is.na function to identify missing values. If the number of missing values ​​is less than the preset number, directly delete the records or variables containing missing values. Use global constants, mean, mode, or regression model methods to fill in missing values. Use statistical analysis methods to identify outliers, correct or delete outliers, and use data deduplication technology to ensure that there are no duplicate records in the data set. Data integration: extract the required tax data from various data sources, solve the problem of inconsistent data formats, units, and codes between different data sources, convert the data into a unified format, and load the converted data into a unified data storage device; Data reduction: Summarize or aggregate the data to reduce the level of detail, compress the data using principal component analysis techniques, reduce the dimensionality of the data, and extract a representative sample from the original data set for subsequent analysis or modeling.

3. The cloud computing-based enterprise tax audit analysis system according to claim 1 is characterized in that: The improved long short-term memory network model is used to extract text entities from the preprocessed corporate tax data to obtain corporate tax entities. The corporate tax entities are classified into financial entities, tax entities, and business entities, including the following processes: The pre-processed corporate tax data is embedded into a vector sequence , Indicates The input vector of words, where word embedding is to randomly initialize the One-hot vector of a single character or word from the embedding matrix to a low-dimensional dense vector; The vector sequence Input into the improved long short-term memory network model to obtain the text feature matrix P: the vector sequence Input into the improved long short-term memory network model to obtain the forward hidden layer sequence and the backward hidden layer sequence ,Bundle , Integrate by position to get the final complete hidden layer sequence ,Will The columns are connected to the linear layer, mapped from n-dimension to k-dimension, and the vector dimensions correspond to the length of the label set, respectively, and mapped into the text feature matrix P, where the value of k is the number of standard set labels; The feature matrix P is classified and labeled through the CRF model to obtain financial entities, tax entities and business entities.

4. The cloud computing-based enterprise tax audit analysis system according to claim 3 is characterized in that: The feature matrix P is classified and labeled through the CRF model to obtain financial entities, tax entities, and business entities, including the following process: Get entity meaning elements based on feature matrix P ,in, represents the i-th word in the j-th annotation; By input and vector sequence The corresponding tag sequence , and combined with the state transfer matrix A of the CRF model, we get the vector sequence Scoring : ； in, Represents the adjacent state transfer matrix score value in the CRF model, Represents the output vector fraction value; By dividing the score The probability formula of the model obtained after normalization is as follows: ； in, represents all possible labeling sequences; Select the predictions labeled with the maximum likelihood estimation function to obtain financial entities, tax entities, and business entities, where each entity category corresponds to an output Interval value, the formula is as follows: 。 5. The cloud computing-based enterprise tax audit analysis system according to claim 1 is characterized in that: Analyzing the mechanism relationship between entities of each category based on the dependency syntax analysis method and constructing the entity association graph based on the mechanism relationship includes the following processes: Dependency extraction: Perform dependency syntactic analysis on sentences containing entities to extract dependency relationships between entities; Relationship classification: According to the type of dependency and context information, the relationship between entities is classified into specific mechanism relationships, including causal relationships, parallel relationships, and inclusion relationships; Relationship screening: remove noise and irrelevant relationships and retain only those relationships that are useful for building entity association graphs; Entity node construction: The identified entities are used as nodes in the graph; Relationship edge construction: Based on the extracted mechanism relationship, construct edges connecting entities in the graph and mark the edge types.

6. The cloud computing-based enterprise tax audit analysis system according to claim 1 is characterized in that: Performing audit analysis on the entity data pool and obtaining the audit analysis results includes the following processes: Obtain the transaction number, commodity name, transaction amount, transaction quantity, tax rate and tax amount in the entity data pool, and the registered transaction number, registered commodity name, registered transaction amount, registered transaction quantity, registered tax rate and registered tax amount in the cloud service center; Calculate the same rate JY between the transaction number and the registered transaction number, the same rate JE between the transaction amount and the registered transaction amount, the same rate SL between the transaction quantity and the registered transaction quantity, the same rate SW between the tax rate and the registered tax rate, and the same rate SEW between the tax amount and the registered tax amount respectively; Calculate the proportion of product name types to registered product name types SP; Record JY, JE, SL, SW, SEW and SP as the audit analysis results.

7. The cloud computing-based enterprise tax audit analysis system according to claim 1 is characterized in that: Generating a risk monitoring index based on the audit analysis results and conducting risk warnings for corporate taxation based on the risk monitoring index includes the following processes: Calculate the corporate tax sameness index SWX based on JY, JE, SL, SW, and SEW: ； Substitute the corporate tax sameness index SWX and SP into the risk monitoring index calculation formula to obtain the risk monitoring index LMS, the calculation formula is as follows: ； in, , is the weight coefficient, set by the system; When the risk monitoring index When it is greater than the preset threshold, a risk warning is generated and fed back to the enterprise.

Images