CN116452007A - Enterprise tax compliance risk assessment method based on capsule network - Google Patents

Enterprise tax compliance risk assessment method based on capsule network Download PDF

Info

Publication number
CN116452007A
CN116452007A CN202310708709.6A CN202310708709A CN116452007A CN 116452007 A CN116452007 A CN 116452007A CN 202310708709 A CN202310708709 A CN 202310708709A CN 116452007 A CN116452007 A CN 116452007A
Authority
CN
China
Prior art keywords
tax
risk
network
index
capsule
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN202310708709.6A
Other languages
Chinese (zh)
Other versions
CN116452007B (en
Inventor
胡为民
黄婵娟
何永定
张丽
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Dib Enterprise Risk Management Technology Co ltd
Original Assignee
Shenzhen Dib Enterprise Risk Management Technology Co ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Dib Enterprise Risk Management Technology Co ltd filed Critical Shenzhen Dib Enterprise Risk Management Technology Co ltd
Priority to CN202310708709.6A priority Critical patent/CN116452007B/en
Publication of CN116452007A publication Critical patent/CN116452007A/en
Application granted granted Critical
Publication of CN116452007B publication Critical patent/CN116452007B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/22—Matching criteria, e.g. proximity measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/10—Tax strategies

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Human Resources & Organizations (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Biophysics (AREA)
  • Strategic Management (AREA)
  • Computing Systems (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Software Systems (AREA)
  • Economics (AREA)
  • Development Economics (AREA)
  • Marketing (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Finance (AREA)
  • Accounting & Taxation (AREA)
  • General Business, Economics & Management (AREA)
  • Technology Law (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Game Theory and Decision Science (AREA)
  • Tourism & Hospitality (AREA)
  • Educational Administration (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The invention belongs to the technical field of enterprise wind control, and particularly relates to an enterprise tax compliance risk assessment method based on a capsule network, which comprises the following steps: s1: acquiring tax index data of an enterprise, preprocessing and acquiring formatted data characteristicsThe method comprises the steps of carrying out a first treatment on the surface of the S2: formatting based data featuresFeature learning of tax indexes is carried out by adopting a residual network, and fused tax feature vectors are obtainedThe method comprises the steps of carrying out a first treatment on the surface of the S3: formatting based data featuresFeature extraction is carried out on tax index data by adopting a capsule network, the relation among different risk indexes is captured, and a multi-dimensional risk feature vector is obtainedThe method comprises the steps of carrying out a first treatment on the surface of the S4: combining fused tax feature vectorsAnd risk feature vectorWeight optimization is carried out by adopting a mean square error loss function; s5: and calculating the probability of risk of each risk index of the enterprise by using the cosine similarity, and outputting a risk assessment result corresponding to each risk index to represent the enterprise tax compliance of the enterprise on each risk index.

Description

Enterprise tax compliance risk assessment method based on capsule network
Technical Field
The invention belongs to the technical field of enterprise wind control, and particularly relates to an enterprise tax compliance risk assessment method based on a capsule network.
Background
The tax compliance of enterprises refers to the process of timely and accurately reporting and paying tax according to national tax regulations and policy regulations, and is critical to the stable development of enterprises. The enterprise tax compliance risk assessment can effectively help enterprises to identify and control risks, and the tax compliance level is improved.
Currently, tax risks are mainly analyzed by financial staff on tax related data of enterprises to judge the tax risks of the enterprises. But often associated with subjectivity of financial staff, there are subjectivity and limitations; and because the manual evaluation efficiency is limited, the requirement on participation in evaluation data is high, the data needs to be cleaned and processed in advance, a large amount of complex data is difficult to process, and the accuracy of an analysis result is difficult to guarantee. Meanwhile, the influence of a single risk factor is considered in the traditional risk assessment, the relation among different risk factors cannot be comprehensively and comprehensively considered, and larger errors are easily caused in analysis results, so that the accuracy of tax risk assessment is lower.
Disclosure of Invention
The invention provides an enterprise tax compliance risk assessment method based on a capsule network, which is characterized in that a residual error network and the capsule network are adopted to learn enterprise tax index type data to obtain a fused tax characteristic vector and a risk characteristic vector, and meanwhile, enterprise tax index risk items can be obtained; optimizing the residual error network model and the capsule network model by adopting a mean square error loss function, calculating the probability of risk of each risk index of the enterprise by adopting cosine similarity, and outputting a corresponding risk assessment result to represent the tax compliance status of the enterprise on the corresponding risk index, so as to realize automatic and intelligent assessment of the tax compliance risk of the enterprise, help the enterprise to discover the potential risk in time and improve the tax compliance level; the technical defect of low accuracy of tax risk assessment results caused by subjectivity of financial staff and limitation of single factor consideration can be overcome, and the accuracy of tax compliance risk assessment is improved.
An enterprise tax compliance risk assessment method based on a capsule network comprises the following steps:
s1: acquiring tax index data of an enterprise, preprocessing the tax index data, and acquiring formatted data characteristics；
S2: formatting based data featuresFeature learning of tax indexes is carried out by adopting a residual network, and a fused tax feature vector +.>；
S3: formatting based data featuresFeature extraction is carried out on tax index data by adopting a capsule network, the relation among different risk indexes is captured, and a multi-dimensional risk feature vector +_is obtained>；
S4: combining fused tax feature vectorsAnd risk feature vector->Weight optimization is carried out by adopting a mean square error loss function;
s5: and calculating the probability of risk of each risk index of the enterprise by using the cosine similarity, and outputting a risk assessment result corresponding to each risk index for representing the enterprise tax compliance of the enterprise on each risk index.
The enterprise tax index type data is learned and processed by adopting a residual network and a capsule network to obtain a fused tax characteristic vector and a risk characteristic vector, and meanwhile, enterprise tax index risk items can be obtained; optimizing the residual error network model and the capsule network model by adopting a mean square error loss function, calculating the probability of risk of each risk index of the enterprise by adopting cosine similarity, and outputting a corresponding risk assessment result to represent the tax compliance status of the enterprise on the corresponding risk index, so as to realize automatic and intelligent assessment of the tax compliance risk of the enterprise, help the enterprise to discover the potential risk in time and improve the tax compliance level; the technical defect of low accuracy of tax risk assessment results caused by subjectivity of financial staff and limitation of single factor consideration can be overcome, and the accuracy of tax compliance risk assessment is improved.
Further, in the step S1, the data indexes in the tax index data of the enterprise include: the rate of increase in business income, the rate of gross interest, the rate of fee, the rate of effective tax, the amount of tax payable, the amount of overdue unpaid tax, the number of tax violations, and the rate of tax earnings for enterprises.
Further, the saidS1, preprocessing tax index data of an enterprise to obtain formatted data characteristicsThe process of (1) comprises the following steps:
s11: based on the obtained tax index data, carrying out normalization processing on each tax index data;
s12: according to time seriesHFor a pair ofWThe tax index types are arranged, and a two-dimensional vector of a time sequence and the tax index types is constructed and used as formatted data characteristics。
Further, in S11, the calculation expression for normalizing the tax index data is:
；
in the method, in the process of the invention,representing normalized tax index data values ranging from 0 to 1->Representing the original tax index data value, +.>Representing the maximum value in the corresponding tax index, +.>Representing the minimum value in the corresponding tax index.
Further, in the step S2, the data characteristics are based on formattingFeature learning of tax indexes is carried out by adopting a residual network, and a fused tax feature vector +.>The process of (1) specifically comprises:
respectively constructing a residual network for the number of each tax index type, and sharingA plurality of; i.e. set the residual network as a functionWherein->The method comprises the steps of carrying out a first treatment on the surface of the The residual error network comprises a convolution layer and a full connection layer;
s21: characterizing dataIs of tax index type, i.e +.>Respectively serving as input of a corresponding residual error network convolution layer;
s22: after the residual network convolution layer carries out convolution processing on the tax index types, calculating feature vectors corresponding to the tax index types by combining jump connectionAnd splice the feature vector corresponding to each tax index type +.>Constructing a new feature matrix->；
Feature vectors corresponding to each tax index typeThe calculated expression of (2) is:
；
splicing feature vectors corresponding to each tax index typeConstructing a new feature matrix->The calculation expression is as follows:
；
wherein a new feature matrixIs shaped as +.>，/>Representing the length of the feature vector, +.>The width of the spliced feature vector, namely the number of tax index types is represented;
s23: the constructed characteristic matrixAs the input of the residual network full connection layer, and adopts an activation function to perform activation processing for outputting the fused tax characteristic vector +.>；
Fusing tax feature vectorsThe calculated expression of (2) is:
；
in the method, in the process of the invention,is an activation function; />The weight matrix is the full connection layer in the residual network and is in the shape of；/>The bias vector is the bias vector of the full connection layer in the residual error network; />Is one-dimensional fused tax characteristic vector with the vector length of +.>。
And performing feature learning through a residual network, and providing key information for enterprise tax compliance risk assessment for subsequent enterprise tax compliance risk assessment.
Further, in the step S3, the data characteristics are based on formattingFeature extraction is carried out on tax index data by adopting a capsule network, the relation among different risk indexes is captured, and a multi-dimensional risk feature vector +_is obtained>The process of (1) specifically comprises:
s31: characterizing dataAs input of capsule network convolution layer, obtain new feature map +.>To extract data features->Is a local feature of (2);
novel feature mapThe calculated expression of (2) is:
；
in the method, in the process of the invention,is an activation function; />A weight matrix for a convolutional layer in the capsule network; />Is the bias vector of the convolution layer in the capsule network; />Is a convolution operation;
s32: based on the obtained feature mapCapturing vectors representing the presence and properties of a feature using the capsule layer of a capsule network>；
Wherein the capsule layer isEach capsule layer is composed of a plurality of capsule units; output vector of capsule layer->For representing the presence and properties of a feature, the computational expression is:
；
in the method, in the process of the invention,is->Output vector of each capsule layer, wherein +.>；/>A calculation function for the capsule layer;
s33: will beThe output vector of each capsule layer is used as the input of the dynamic routing layer, and the connection weight between the capsule layer and the risk feature vector is updated through iteration>For outputting risk feature vectors ++>；
Connection weightThe calculated expression of (2) is:
；
risk feature vectorThe calculated expression of (2) is:
；
in the method, in the process of the invention,indicate->Capsule layer and->The connection weight among the individual risk feature vectors; />Unnormalized data for connected weights; />Is a normalization function; />Is a risk feature vector, the vector length of which is +.>Length of->In accordance with, wherein->，/>The number of the risk feature vectors is; />To activate the function.
The existence and the attribute of the risk features in the data features are captured through the capsule network, and key information is provided for enterprise tax compliance risk assessment for subsequent enterprise tax compliance risk assessment.
Further, in S4, the tax feature vector is combinedAnd risk feature vector->The process of weight optimization by adopting the mean square error loss function specifically comprises the following steps:
s41: based on real tax feature vectorAnd true risk feature vector->Respectively calculating losses of a residual network and a capsule network, and calculating weighted losses of the residual network and the capsule network by combining weights;
residual network loss functionThe calculated expression of (2) is:
；
capsule network loss functionThe calculated expression of (2) is:
；
residual network and capsule network weighting lossThe calculated expression of (2) is:
；
in the method, in the process of the invention,、/>is a super parameter and is used for controlling the weight loss of a residual error network and a capsule network;
s42: adopting a gradient descent back propagation algorithm to respectively update weights of a residual error network and a capsule network;
the computational expression for the gradient of the residual network is:
；
the computational expression of the gradient of the capsule network is:
；
the process of updating the residual network weights is as follows:
；
the process of updating the capsule network weight is as follows:
；
in the method, in the process of the invention,is the learning rate, used to control the step size of the parameter update.
By iteratively optimizing the model weight parameters, the accuracy of the extracted tax characteristics and risk characteristics can be improved, so that enterprise tax compliance risk assessment can be more accurately carried out.
Further, in the step S5, the process of calculating the probability of risk of the enterprise using the cosine similarity and outputting the risk assessment result of the tax compliance of the enterprise includes:
s51: combining fused tax feature vectorsAnd risk feature vector->Calculating cosine similarity between the two;
the cosine similarity is calculated by the following expression:
；
in the method, in the process of the invention,calculating a function for cosine similarity, wherein the function is used for representing the probability that the risk index item has risk; />For fusing tax feature vectors->And risk feature vector->Is a dot product of (2); />To fuse tax feature vectorsIs a mold of (2); />Is a risk feature vector->Is a mold of (2);
s52: setting a similarity thresholdJudging whether the enterprise has risks or not based on the cosine similarity value;
；
in the method, in the process of the invention,indicate->A risk-like indicator is a result of risk; when->When it indicates +.>Risk of the risk-like index is present when +.>When it indicates +.>The risk-like index is free of risk.
A computer device, comprising: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication over the bus when the computer device is running, the machine-readable instructions when executed by the processor performing the method of any of the above.
A computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the method of any of the preceding claims.
The beneficial effects of the invention are as follows:
according to the method, the residual error network and the capsule network are adopted to learn the enterprise tax index type data, so that the fusion tax characteristic vector and the risk characteristic vector are obtained, and meanwhile, the enterprise tax index risk item can be obtained; optimizing the residual error network model and the capsule network model by adopting a mean square error loss function, calculating the probability of risk of each risk index of the enterprise by adopting cosine similarity, and outputting a corresponding risk assessment result to represent the tax compliance status of the enterprise on the corresponding risk index, so as to realize automatic and intelligent assessment of the tax compliance risk of the enterprise, help the enterprise to discover the potential risk in time and improve the tax compliance level; the technical defects of low efficiency and low accuracy of tax risk assessment results caused by subjectivity of financial staff and limitation of single factor consideration can be overcome, and the processing efficiency and the result accuracy of tax compliance risk assessment are improved.
Drawings
FIG. 1 is a flow chart of the present invention;
fig. 2 is a schematic structural diagram of a computer device according to the present invention.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments of the present invention, but not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
It is noted that various aspects of the embodiments are described below within the scope of the following claims. It should be apparent that the aspects described herein may be embodied in a wide variety of forms and that any specific structure and/or function described herein is merely illustrative. Based on the present disclosure, one skilled in the art will appreciate that one aspect described herein may be implemented independently of any other aspect, and that two or more of these aspects may be combined in various ways. For example, an apparatus may be implemented and/or a method practiced using any number of the aspects set forth herein. In addition, such apparatus may be implemented and/or such methods practiced using other structure and/or functionality in addition to one or more of the aspects set forth herein.
Furthermore, in the following description, specific details are provided for the purpose of providing a thorough understanding of the examples, and the particular meaning of the terms described above in this application will be understood to those of ordinary skill in the art in the context of the present application.
Example 1
FIG. 1 shows an enterprise tax compliance risk assessment method based on a capsule network, wherein a residual network and the capsule network are adopted to learn enterprise tax index type data to obtain a fused tax feature vector and a risk feature vector, and meanwhile, enterprise tax index risk items can be obtained; optimizing the residual error network model and the capsule network model by adopting a mean square error loss function, calculating the probability of risk of each risk index of the enterprise by adopting cosine similarity, and outputting a corresponding risk assessment result to represent the tax compliance status of the enterprise on the corresponding risk index, so as to realize automatic and intelligent assessment of the tax compliance risk of the enterprise, help the enterprise to discover the potential risk in time and improve the tax compliance level; the technical defect of low accuracy of tax risk assessment results caused by subjectivity of financial staff and limitation of single factor consideration can be overcome, and the accuracy of tax compliance risk assessment is improved. The method specifically comprises the following steps:
s1: acquiring tax index data of an enterprise, preprocessing the tax index data, and acquiring formatted data characteristics；
Wherein, the data index in tax index data of enterprise includes: the business income increasing rate (%), the gross interest rate (%), the cost rate (%), the effective tax rate (%), the tax amount (element) to be paid, the overdue non-paid tax amount (element), the tax violation number (secondary), and the tax burden rate (%) of the enterprise.
Preprocessing tax index data of an enterprise to obtain formatted data characteristicsThe process of (1) comprises the following steps:
s11: based on the obtained tax index data, carrying out normalization processing on each tax index data;
the calculation expression for carrying out normalization processing on the tax index data is as follows:
；
in the method, in the process of the invention,representing normalized tax index data values ranging from 0 to 1->Representing the original tax index data value, +.>Representation pairMaximum value in tax index, +.>Representing the minimum value in the corresponding tax index.
S12: according to time seriesHFor a pair ofWThe tax index types are arranged, and a two-dimensional vector of a time sequence and the tax index types is constructed and used as formatted data characteristics。
The two-dimensional vector is constructed asWherein->For the length of the time series, +.>Is tax index type; two-dimensional vector->In which the rows represent time nodes and the columns represent tax index types, which are taken as formatted data features +.>For the input of the subsequent residual network and capsule network.
S2: formatting based data featuresFeature learning of tax indexes is carried out by adopting a residual network, and a fused tax feature vector +.>The method comprises the steps of carrying out a first treatment on the surface of the The process specifically comprises the following steps:
respectively constructing a residual network for the number of each tax index type, and sharingA plurality of; i.e. set residual errorThe network being a functionWherein->The method comprises the steps of carrying out a first treatment on the surface of the The residual error network comprises a convolution layer and a full connection layer;
s21: characterizing dataIs of tax index type, i.e +.>Respectively serving as input of a corresponding residual error network convolution layer;
characterizing dataSplitting according to columns, and taking each column of tax index type as an independent input characteristic to be respectively input into a corresponding residual error network, so that each residual error network pays attention to different tax indexes independently.
S22: after the residual network convolution layer carries out convolution processing on the tax index types, calculating feature vectors corresponding to the tax index types by combining jump connectionAnd splice the feature vector corresponding to each tax index type +.>Constructing a new feature matrix->；
Feature vectors corresponding to each tax index typeThe calculated expression of (2) is:
；
splicing feature vectors corresponding to each tax index typeConstructing a new feature matrix->The calculation expression is as follows:
；
wherein a new feature matrixIs shaped as +.>，/>Representing the length of the feature vector, +.>The width of the spliced feature vector, namely the number of tax index types is represented;
s23: the constructed characteristic matrixAs the input of the residual network full connection layer, and adopts an activation function to perform activation processing for outputting the fused tax characteristic vector +.>；
Fusing tax feature vectorsThe calculated expression of (2) is:
；
in the method, in the process of the invention,is an activation function; />The weight matrix is the full connection layer in the residual network and is in the shape of；/>The bias vector is the bias vector of the full connection layer in the residual error network; />Is one-dimensional fused tax characteristic vector with the vector length of +.>。
And performing feature learning through a residual network, and providing key information for enterprise tax compliance risk assessment for subsequent enterprise tax compliance risk assessment.
S3: formatting based data featuresFeature extraction is carried out on tax index data by adopting a capsule network, the relation among different risk indexes is captured, and a multi-dimensional risk feature vector +_is obtained>；
Further, in the step S3, the data characteristics are based on formattingFeature extraction is carried out on tax index data by adopting a capsule network, the relation among different risk indexes is captured, and a multi-dimensional risk feature vector +_is obtained>The process of (1) specifically comprises:
constructing a capsule network for multitasking learning; capsule networkComprising a convolution layer, a plurality of capsule layers, and a dynamic routing layer. Acquiring multi-dimensional risk feature vectors by using capsule networkThe process of (1) specifically comprises:
s31: characterizing dataAs input of capsule network convolution layer, obtain new feature map +.>To extract data features->Is a local feature of (2);
novel feature mapThe calculated expression of (2) is:
；
in the method, in the process of the invention,is an activation function; />A weight matrix for a convolutional layer in the capsule network; />Is the bias vector of the convolution layer in the capsule network; />Is a convolution operation;
s32: based on the obtained feature mapCapturing vectors representing the presence and properties of a feature using the capsule layer of a capsule network>；
Wherein the capsule layer isEach capsule layer is composed of a plurality of capsule units; output vector of capsule layer->For representing the presence and properties of a feature, the computational expression is:
；
in the method, in the process of the invention,is->Output vector of each capsule layer, wherein +.>；/>A calculation function for the capsule layer;
s33: will beThe output vector of each capsule layer is used as the input of the dynamic routing layer, and the connection weight between the capsule layer and the risk feature vector is updated through iteration>For outputting risk feature vectors ++>；
Connection weightThe calculated expression of (2) is:
；
risk feature vectorThe calculated expression of (2) is:
；
in the method, in the process of the invention,indicate->Capsule layer and->The connection weight among the individual risk feature vectors; />Unnormalized data for connected weights; />Is a normalization function; />Is a risk feature vector, the vector length of which is +.>Length of->In accordance with, wherein->，/>The number of the risk feature vectors is; />To activate the function.
In the enterprise tax compliance risk assessment process, the risk indexes comprise overdue risk of unpaid tax, insufficient tax payment risk, tax planning risk, tax collection management system change risk and the like.
The existence and the attribute of the risk features in the data features are captured through the capsule network, and key information is provided for enterprise tax compliance risk assessment for subsequent enterprise tax compliance risk assessment.
S4: combining fused tax feature vectorsAnd risk feature vector->Weight optimization is carried out by adopting a mean square error loss function; the process specifically comprises the following steps:
s41: based on real tax feature vectorAnd true risk feature vector->Respectively calculating losses of a residual network and a capsule network, and calculating weighted losses of the residual network and the capsule network by combining weights;
residual network loss functionThe calculated expression of (2) is:
；
capsule network loss functionThe calculated expression of (2) is:
；
residual network and capsule network weighting lossThe calculated expression of (2) is:
；
in the method, in the process of the invention,、/>is a super parameter and is used for controlling the weight loss of a residual error network and a capsule network;
s42: adopting a gradient descent back propagation algorithm to respectively update weights of a residual error network and a capsule network;
the computational expression for the gradient of the residual network is:
；
the computational expression of the gradient of the capsule network is:
；
the process of updating the residual network weights is as follows:
；
the process of updating the capsule network weight is as follows:
；
in the method, in the process of the invention,is the learning rate, used to control the step size of the parameter update.
By iteratively optimizing the model weight parameters, the accuracy of the extracted tax characteristics and risk characteristics can be improved, so that enterprise tax compliance risk assessment can be more accurately carried out.
S5: calculating the probability of risk of each risk index of the enterprise by using the cosine similarity, and outputting a risk assessment result corresponding to each risk index for representing the enterprise tax compliance of the enterprise on each risk index; the process specifically comprises the following steps:
s51: combining fused tax feature vectorsAnd risk feature vector->The cosine similarity between the two is calculated to evaluate the probability of risk of each risk index;
the cosine similarity is calculated by the following expression:
；
in the method, in the process of the invention,calculating a function for cosine similarity, wherein the function is used for representing the probability that the risk index item has risk; />For fusing tax feature vectors->And risk feature vector->Is a dot product of (2); />To fuse tax feature vectorsIs a mold of (2); />Is a risk feature vector->Is a mold of (2);
s52: setting a similarity thresholdJudging whether the enterprise has risks or not based on the cosine similarity value;
；
in the method, in the process of the invention,indicate->A risk-like indicator is a result of risk; when->When it indicates +.>Risk of the risk-like index is present when +.>When it indicates +.>The risk-like index is free of risk.
By outputting the evaluation results of the risk index types, intelligent evaluation of the tax compliance risk of the enterprise is realized, the enterprise is helped to find potential risks in time, and the tax compliance level is improved.
Example 2
Based on the same technical concept, as shown in fig. 2, the present embodiment further provides a computer device corresponding to the method provided in the foregoing embodiment, including a processor 2, a memory 1, and a bus, where the memory stores machine-readable instructions executable by the processor, and when the computer device is running, the processor and the memory communicate through the bus, and the machine-readable instructions are executed by the processor to perform any one of the methods described above.
The memory 1 includes at least one type of readable storage medium including flash memory, a hard disk, a multimedia card, a card type memory (e.g., SD or DX memory, etc.), a magnetic memory, a magnetic disk, an optical disk, etc.
The memory 1 may in some embodiments be an internal storage unit of an enterprise tax compliance risk assessment system, such as a hard disk. The memory 1 may in other embodiments also be an external storage device of an enterprise tax compliance risk assessment system, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash Card (Flash Card), etc. Further, the memory 1 may also include both an internal storage unit and an external storage device of the enterprise tax compliance risk assessment system. The memory 1 may be used not only for storing application software installed in the enterprise tax compliance risk assessment system and various types of data, such as codes of the enterprise tax compliance risk assessment system program, etc., but also for temporarily storing data that has been output or is to be output.
The processor 2 may in some embodiments be a central processing unit (Central Processing Unit, CPU), controller, microcontroller, microprocessor or other data processing chip for running program code or processing data stored in the memory 1, for example executing an enterprise tax compliance risk assessment program or the like.
The disclosed embodiments also provide a computer readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method described in the method embodiments above. Wherein the storage medium may be a volatile or nonvolatile computer readable storage medium.
The computer program product for applying the page content refreshing method provided by the embodiment of the present invention includes a computer readable storage medium storing program codes, and the instructions included in the program codes may be used to execute the steps of the method described in the method embodiment, specifically, refer to the method embodiment and are not repeated herein.
The disclosed embodiments also provide a computer program which, when executed by a processor, implements any of the methods of the previous embodiments. The computer program product may be realized in particular by means of hardware, software or a combination thereof. In an alternative embodiment, the computer program product is embodied as a computer storage medium, and in another alternative embodiment, the computer program product is embodied as a software product, such as a software development kit (Software Development Kit, SDK), or the like.
It is to be understood that the same or similar parts in the above embodiments may be referred to each other, and that in some embodiments, the same or similar parts in other embodiments may be referred to.
It should be noted that in the description of the present invention, the terms first, second, and the like are used for descriptive purposes only and are not to be construed as indicating or implying relative importance. Furthermore, in the description of the present invention, unless otherwise indicated, a plurality of meanings means at least two.
Any process or method descriptions in flow charts or otherwise described herein may be understood as representing modules, segments, or portions of code which include one or more executable instructions for implementing specific logical functions or steps of the process, and further implementations are included within the scope of the preferred embodiment of the present invention in which functions may be executed out of order from that shown or discussed, including substantially concurrently or in reverse order, depending on the functionality involved, as would be understood by those reasonably skilled in the art of the present invention.
It is to be understood that portions of the present invention may be implemented in hardware, software, firmware, or a combination thereof. In the above-described embodiments, the various steps or methods may be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, may be implemented using any one or combination of the following techniques, as is well known in the art: discrete logic circuits having logic gates for implementing logic functions on data signals, application specific integrated circuits having suitable combinational logic gates, programmable Gate Arrays (PGAs), field Programmable Gate Arrays (FPGAs), and the like.
Those of ordinary skill in the art will appreciate that all or a portion of the steps carried out in the method of the above-described embodiments may be implemented by a program to instruct related hardware, where the program may be stored in a computer readable storage medium, and where the program, when executed, includes one or a combination of the steps of the method embodiments.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing module, or each unit may exist alone physically, or two or more units may be integrated in one module. The integrated modules may be implemented in hardware or in software functional modules. The integrated modules may also be stored in a computer readable storage medium if implemented in the form of software functional modules and sold or used as a stand-alone product.
The above-mentioned storage medium may be a read-only memory, a magnetic disk or an optical disk, or the like.
In the description of the present specification, a description referring to terms "one embodiment," "some embodiments," "examples," "specific examples," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present invention. In this specification, schematic representations of the above terms do not necessarily refer to the same embodiments or examples. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.
While embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and not to be construed as limiting the invention, and that variations, modifications, alternatives and variations may be made to the above embodiments by one of ordinary skill in the art within the scope of the invention.

Claims (10)

1. The enterprise tax compliance risk assessment method based on the capsule network is characterized by comprising the following steps of:
s1: acquiring tax index data of an enterprise, preprocessing the tax index data, and acquiring formatted data characteristics；
S2: formatting based data featuresFeature learning of tax indexes is carried out by adopting a residual network, and a fused tax feature vector +.>；
S3: formatting based data featuresFeature extraction is carried out on tax index data by adopting a capsule network, the relation among different risk indexes is captured, and a multi-dimensional risk feature vector +_is obtained>；
S4: combining fused tax feature vectorsAnd risk feature vector->Weight optimization is carried out by adopting a mean square error loss function;
s5: and calculating the probability of risk of each risk index of the enterprise by using the cosine similarity, and outputting a risk assessment result corresponding to each risk index for representing the enterprise tax compliance of the enterprise on each risk index.
2. The method for evaluating the tax compliance risk of an enterprise based on the capsule network according to claim 1, wherein in S1, the data index in the tax index data of the enterprise comprises: the rate of increase in business income, the rate of gross interest, the rate of fee, the rate of effective tax, the amount of tax payable, the amount of overdue unpaid tax, the number of tax violations, and the rate of tax earnings for enterprises.
3. The method for evaluating the tax compliance risk of an enterprise based on a capsule network according to claim 1, wherein in S1, tax index data of the enterprise is preprocessed to obtain formatted data featuresThe process of (1) comprises the following steps:
s11: based on the obtained tax index data, carrying out normalization processing on each tax index data;
s12: according to time seriesHFor a pair ofWThe tax index types are arranged, and a two-dimensional vector of a time sequence and the tax index types is constructed and used as formatted data characteristics。
4. The method for evaluating the compliance risk of tax in enterprises based on the capsule network as set forth in claim 3, wherein in S11, the calculation expression for normalizing the tax index data is:
；
in the method, in the process of the invention,representing normalized tax index data values ranging from 0 to 1->Representing the original tax index data value, +.>Representing the maximum value in the corresponding tax index, +.>Representing the minimum value in the corresponding tax index.
5. A method for assessing compliance with tax in enterprises based on a capsule network as recited in claim 3, wherein in S2, the compliance with tax compliance risk is based on formatted data featuresFeature learning of tax indexes is carried out by adopting a residual network, and a fused tax feature vector +.>The process of (1) specifically comprises:
respectively constructing a residual network for the number of each tax index type, and sharingA plurality of; i.e. let the residual network be a function->Wherein->The method comprises the steps of carrying out a first treatment on the surface of the The residual error network comprises a convolution layer and a full connection layer;
s21: characterizing dataTax index class of each column of (1)I.e.)>Respectively serving as input of a corresponding residual error network convolution layer;
s22: after the residual network convolution layer carries out convolution processing on the tax index types, calculating feature vectors corresponding to the tax index types by combining jump connectionAnd splice the feature vector corresponding to each tax index type +.>Constructing a new feature matrix->；
Feature vectors corresponding to each tax index typeThe calculated expression of (2) is:
；
splicing feature vectors corresponding to each tax index typeConstructing a new feature matrix->The calculation expression is as follows:
；
wherein a new feature matrixIs shaped as +.>，/>Representing the length of the feature vector, +.>The width of the spliced feature vector, namely the number of tax index types is represented;
s23: the constructed characteristic matrixAs the input of the residual network full connection layer, and adopts an activation function to perform activation processing for outputting the fused tax characteristic vector +.>；
Fusing tax feature vectorsThe calculated expression of (2) is:
；
in the method, in the process of the invention,is an activation function; />The weight matrix is the full connection layer in the residual network, and the shape of the weight matrix is +.>；Bias vector for full connection layer in residual error network；/>Is one-dimensional fused tax characteristic vector with the vector length of +.>。
6. The method for assessing compliance with tax and tax risk of a corporation based on a capsule network of claim 5, wherein said S3 is based on formatted data featuresFeature extraction is carried out on tax index data by adopting a capsule network, the relation among different risk indexes is captured, and a multi-dimensional risk feature vector +_is obtained>The process of (1) specifically comprises:
s31: characterizing dataAs input of capsule network convolution layer, obtain new feature map +.>To extract data features->Is a local feature of (2);
novel feature mapThe calculated expression of (2) is:
；
in the method, in the process of the invention,is an activation function; />A weight matrix for a convolutional layer in the capsule network; />Is the bias vector of the convolution layer in the capsule network; />Is a convolution operation;
s32: based on the obtained feature mapCapturing vectors representing feature presence and attributes using a capsule layer of a capsule network；
Wherein the capsule layer isEach capsule layer is composed of a plurality of capsule units; output vector of capsule layer->For representing the presence and properties of a feature, the computational expression is:
；
in the method, in the process of the invention,is->Output vector of each capsule layer, wherein +.>；/>A calculation function for the capsule layer;
s33: will beThe output vector of each capsule layer is used as the input of the dynamic routing layer, and the connection weight between the capsule layer and the risk feature vector is updated through iteration>For outputting risk feature vectors ++>；
Connection weightThe calculated expression of (2) is:
；
risk feature vectorThe calculated expression of (2) is:
；
in the method, in the process of the invention,indicate->Capsule layer and->The connection weight among the individual risk feature vectors; />Unnormalized data for connected weights; />Is a normalization function; />Is a risk feature vector, the vector length of which is integrated with the tax feature vectorLength of->In accordance with, wherein->，/>The number of the risk feature vectors is; />To activate the function.
7. The method for evaluating compliance risk of enterprise tax based on capsule network as claimed in claim 6, wherein in S4, the feature vectors of tax are combined togetherAnd risk feature vector->The process of weight optimization by adopting the mean square error loss function specifically comprises the following steps:
s41: based on real tax feature vectorAnd true risk feature vector->Respectively calculating losses of a residual network and a capsule network, and calculating weighted losses of the residual network and the capsule network by combining weights;
residual network loss functionThe calculated expression of (2) is:
；
capsule network loss functionThe calculated expression of (2) is:
；
residual network and capsule network weighting lossThe calculated expression of (2) is:
；
in the method, in the process of the invention,、/>is a super parameter and is used for controlling the weight loss of a residual error network and a capsule network;
s42: adopting a gradient descent back propagation algorithm to respectively update weights of a residual error network and a capsule network;
the computational expression for the gradient of the residual network is:
；
the computational expression of the gradient of the capsule network is:
；
the process of updating the residual network weights is as follows:
；
the process of updating the capsule network weight is as follows:
；
in the method, in the process of the invention,is the learning rate, used to control the step size of the parameter update.
8. The method for evaluating the risk of enterprise tax compliance based on the capsule network according to claim 7, wherein in S5, the process of calculating the probability of enterprise risk using cosine similarity and outputting the result of evaluating the risk of enterprise tax compliance comprises:
s51: combining fused tax feature vectorsAnd risk feature vector->Calculating cosine similarity between the two;
the cosine similarity is calculated by the following expression:
；
in the method, in the process of the invention,calculating a function for cosine similarity, wherein the function is used for representing the probability that the risk index item has risk; />For fusing tax feature vectors->And risk feature vector->Is a dot product of (2); />For fusing tax feature vectors->Is a mold of (2);is a risk feature vector->Is a mold of (2);
s52: setting a similarity thresholdJudging whether the enterprise has risks or not based on the cosine similarity value;
；
in the method, in the process of the invention,indicate->A risk-like indicator is a result of risk; when->When representing the first of the enterprisesRisk of the risk-like index is present when +.>When it indicates +.>The risk-like index is free of risk.
9. A computer device, comprising: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication over the bus when the computer device is running, the machine-readable instructions when executed by the processor performing the method of any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that it has stored thereon a computer program which, when executed by a processor, performs the method according to any of claims 1 to 8.
CN202310708709.6A 2023-06-15 2023-06-15 Enterprise tax compliance risk assessment method based on capsule network Active CN116452007B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310708709.6A CN116452007B (en) 2023-06-15 2023-06-15 Enterprise tax compliance risk assessment method based on capsule network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310708709.6A CN116452007B (en) 2023-06-15 2023-06-15 Enterprise tax compliance risk assessment method based on capsule network

Publications (2)

Publication Number Publication Date
CN116452007A true CN116452007A (en) 2023-07-18
CN116452007B CN116452007B (en) 2023-09-19

Family

ID=87130632

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310708709.6A Active CN116452007B (en) 2023-06-15 2023-06-15 Enterprise tax compliance risk assessment method based on capsule network

Country Status (1)

Country Link
CN (1) CN116452007B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117237114A (en) * 2023-11-10 2023-12-15 深圳市迪博企业风险管理技术有限公司 Financing trade compliance detection method based on twin evolution

Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110264085A (en) * 2019-06-21 2019-09-20 黄河科技学院 A kind of assessment of industrial and commercial administration business risk and early warning system
US20200320704A1 (en) * 2018-07-24 2020-10-08 Shenzhen Institutes Of Advanced Technology Method and device of processing plaques in magnetic resonance imaging of vessel wall, and computer device
CN112733701A (en) * 2021-01-07 2021-04-30 中国电子科技集团公司信息科学研究院 Robust scene recognition method and system based on capsule network
CN113269433A (en) * 2021-05-20 2021-08-17 深圳易财信息技术有限公司 Tax risk prediction method, apparatus, medium, and computer program product
CN113449819A (en) * 2021-08-27 2021-09-28 中国测绘科学研究院 Credit evaluation model method based on capsule network and storage medium thereof
CN115471899A (en) * 2022-10-21 2022-12-13 青岛科技大学 Face expression recognition method combining ResNet18 and capsule network and performing pruning
WO2022267456A1 (en) * 2021-06-24 2022-12-29 平安科技（深圳）有限公司 Vehicle damage assessment method, apparatus and device, and storage medium
CN115761358A (en) * 2022-11-24 2023-03-07 深圳先进技术研究院 Method for classifying myocardial fibrosis based on residual capsule network
CN116227939A (en) * 2023-05-04 2023-06-06 深圳市迪博企业风险管理技术有限公司 Enterprise credit rating method and device based on graph convolution neural network and EM algorithm

Patent Citations (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200320704A1 (en) * 2018-07-24 2020-10-08 Shenzhen Institutes Of Advanced Technology Method and device of processing plaques in magnetic resonance imaging of vessel wall, and computer device
CN110264085A (en) * 2019-06-21 2019-09-20 黄河科技学院 A kind of assessment of industrial and commercial administration business risk and early warning system
CN112733701A (en) * 2021-01-07 2021-04-30 中国电子科技集团公司信息科学研究院 Robust scene recognition method and system based on capsule network
CN113269433A (en) * 2021-05-20 2021-08-17 深圳易财信息技术有限公司 Tax risk prediction method, apparatus, medium, and computer program product
WO2022267456A1 (en) * 2021-06-24 2022-12-29 平安科技（深圳）有限公司 Vehicle damage assessment method, apparatus and device, and storage medium
CN113449819A (en) * 2021-08-27 2021-09-28 中国测绘科学研究院 Credit evaluation model method based on capsule network and storage medium thereof
CN115471899A (en) * 2022-10-21 2022-12-13 青岛科技大学 Face expression recognition method combining ResNet18 and capsule network and performing pruning
CN115761358A (en) * 2022-11-24 2023-03-07 深圳先进技术研究院 Method for classifying myocardial fibrosis based on residual capsule network
CN116227939A (en) * 2023-05-04 2023-06-06 深圳市迪博企业风险管理技术有限公司 Enterprise credit rating method and device based on graph convolution neural network and EM algorithm

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
胡向东 等: "基于胶囊网络的工业互联网入侵检测方法", 《电子学报》, vol. 50, no. 6, pages 1457 - 1465 *
霍鑫怡 等: "基于卷积注意力和胶囊网络的 SAR 少样本目标识别方法", 《中国科学院大学学报》, vol. 39, no. 6, pages 783 - 792 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117237114A (en) * 2023-11-10 2023-12-15 深圳市迪博企业风险管理技术有限公司 Financing trade compliance detection method based on twin evolution
CN117237114B (en) * 2023-11-10 2024-03-08 深圳市迪博企业风险管理技术有限公司 Financing trade compliance detection method based on twin evolution

Also Published As

Publication number Publication date
CN116452007B (en) 2023-09-19

Similar Documents

Publication Publication Date Title
US11681925B2 (en) Techniques for creating, analyzing, and modifying neural networks
CN108171260B (en) Picture identification method and system
US20180253657A1 (en) Real-time credit risk management system
CN110968701A (en) Relationship map establishing method, device and equipment for graph neural network
CN107220217A (en) Characteristic coefficient training method and device that logic-based is returned
CN111199474B (en) Risk prediction method and device based on network map data of two parties and electronic equipment
CN111222976B (en) Risk prediction method and device based on network map data of two parties and electronic equipment
US11640539B2 (en) Techniques for visualizing the operation of neural networks using samples of training data
CN111325619A (en) Credit card fraud detection model updating method and device based on joint learning
WO2021190379A1 (en) Method and device for realizing automatic machine learning
CN116452007B (en) Enterprise tax compliance risk assessment method based on capsule network
CN110084609B (en) Transaction fraud behavior deep detection method based on characterization learning
CN109740642A (en) Invoice category recognition methods, device, electronic equipment and readable storage medium storing program for executing
CN107229731A (en) Method and apparatus for grouped data
US20220215292A1 (en) Method to identify incorrect account numbers
CN113850381A (en) Graph neural network training method and device
US11615321B2 (en) Techniques for modifying the operation of neural networks
CN110555713A (en) method and device for determining sales prediction model
CN111798304A (en) Risk loan determination method and device, electronic equipment and storage medium
Zhu et al. Loan default prediction based on convolutional neural network and LightGBM
CN112446777B (en) Credit evaluation method, device, equipment and storage medium
CN112365352A (en) Anti-cash-out method and device based on graph neural network
CN109635289A (en) Entry classification method and audit information abstracting method
CN114266653A (en) Client loan risk estimation method for integrated learning
CN114548307A (en) Classification model training method and device, and classification method and device

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant