Finance and tax data analysis method and system based on big data

Abstract

The invention relates to the technical field of financial data analysis, and discloses a big data-based finance and taxation data analysis method, which comprises the following steps: firstly, collecting data, then cleaning, formatting and standardizing the collected data, and establishing a database so as to facilitate subsequent analysis and finance and taxation deep learning network establishment; step 2, respectively establishing a pre-training model and a deep learning model, respectively inputting the data of the database in the step 2 into the pre-training model and the deep learning model, performing knowledge distillation, extracting characteristics of finance and taxation data, and identifying a mode, thereby training and perfecting the pre-training model, generating a decision model, and performing decision making on the finance and taxation data. And carrying out finance and tax analysis and prediction by using the trained decision model. According to the big data-based finance and taxation data analysis method, various previous or public finance and taxation-related data, including financial statements, tax records, transaction records and the like, of other companies are stored by a finance and taxation data acquisition module through a data storage module, and a database is established.

Description

Technical Field

[0001] The present invention relates to the technical field of financial data analysis, and specifically to a financial and tax data analysis method and system based on big data. Background Art

[0002] Financial data analysis refers to the analysis of data that summarizes and evaluates a company's financial status and operating results, including debt-paying ability data, operating ability data, profitability data, and development ability data. Through analysis, we can determine whether the company's financial status is good, whether the company's business management is sound, and whether the company's business prospects are bright. At the same time, through analysis, we can also find out the crux of the company's business management and propose solutions to the problems.

[0003] Most of the existing financial and tax data analysis has the problems of too complicated data, slow calculation speed, low analysis efficiency and low accuracy, which makes it impossible to assist enterprises to make correct judgments quickly in actual decision-making. Therefore, a financial and tax data analysis method and system based on big data is proposed to solve the above problems. Summary of the invention

[0004] 1. Technical issues to be solved

[0005] In view of the shortcomings of the prior art, the present invention provides a financial and tax data analysis method and system based on big data, which has the advantage of high data processing efficiency and solves the problem that most of the existing financial and tax data analysis has the problem of too complicated data, slow calculation speed, low analysis efficiency and low accuracy, and thus cannot assist enterprises to make correct judgments quickly in actual decision-making.

[0006] (II) Technical solution

[0007] In order to achieve the above-mentioned purpose of high data processing efficiency, the present invention provides the following technical solution: a financial and tax data analysis method based on big data, characterized in that it includes the following steps:

[0008] Step 1: Data collection: collect various financial and tax-related data of the company in the past or other companies that have been made public, including financial statements, tax records, transaction records, etc.

[0009] Step 2: Clean, format and standardize the collected data and establish a database for subsequent analysis;

[0010] Step 3: Build a deep learning network for finance and taxation, including convolutional neural network (CNN) and recurrent neural network (RNN);

[0011] Step 4: Establish a pre-trained model and a deep learning model respectively, input the data of the database in step 2 into the pre-trained model and the deep learning model respectively, perform knowledge distillation, extract the characteristics of financial and tax data and identify the patterns, so as to train and improve the pre-trained model and generate a decision model;

[0012] Step 5: Use the trained decision model to conduct financial and tax analysis and forecasting.

[0013] Preferably, the training of the pre-trained model comprises the following steps:

[0014] S1 inputs the tax data in step 2 into the pre-trained model in step 4, thereby obtaining a first output result;

[0015] S2 inputs the financial and tax documents in step 2 into the deep learning model in step 4, thereby obtaining a second output result, and determining a first true loss value between the first output result and the second output result;

[0016] S3 inputs the verification data into the deep learning model to obtain a third output result, and compares the third output result with the second output result to obtain a second true loss value;

[0017] S4 determines the loss value according to the first loss value, the second loss value and knowledge distillation, and adjusts the parameters of the pre-trained model to improve the deep learning model and form a decision model.

[0018] Preferably, the data preprocessing in step 2 includes data extraction: using Left, Right, Mid, Find and other functions to extract specified values; data cleaning: using Trim function to clear extra spaces; data replacement: using Replace and Substitute functions to replace specified strings and data merging: using Concat, Phonetic, Textjoin functions to merge strings.

[0019] Preferably, the feature extraction step comprises the following steps:

[0020] 1) Extract image features using CNN;

[0021] 2) Use RNN for time series analysis.

[0022] Preferably, the convolutional neural network mainly includes a convolution layer and a pooling layer. In the convolution operation, the convolution kernel scans the input image, performs weighted summation on the pixel values ​​at each position, and obtains the corresponding pixel value of the output feature map. Specifically, assuming that the input image is X, the convolution kernel is K, and the bias is b, then the output feature map Y formula of the convolution operation is as follows:

[0023]

[0024] C represents the number of input channels; K represents the size of the convolution kernel; i and j represent the positions on the output feature map respectively; p and q represent the positions within the convolution kernel respectively; represents the convolution operation.

[0025] Preferably, when identifying tax violations in a classification model, a cross entropy loss or DiceLoss can be used to optimize the model's predictive ability for different types of tax violations. The cross entropy loss calculation formula is:

[0026]

[0027] Where P is the actual distribution and Q is the expected distribution.

[0028] Preferably, the Dice Loss calculation formula is:

[0029]

[0030] Among them, pi represents the value of the i-th pixel in the predicted result, yi represents the value of the i-th pixel in the true result, X represents the pixel label of the true segmented image, and Y represents the pixel category of the segmented image predicted by the model. This formula is based on the pixel-level loss function and uses the similarity between the predicted result and the true result as the measurement standard.

[0031] Preferably, when using the regression model to predict tax revenue, MSE or MAE can be used as a loss function to evaluate the accuracy of the model prediction. The mean square error calculation formula is:

[0032]

[0033] y i is the true value, i is the predicted value, N is the number of data points, and this formula is used to calculate the mean square error between the predicted value and the true value, which is a commonly used indicator for evaluating the prediction performance of the model.

[0034] Preferably, the fiscal and taxation data analysis system comprises a main controller module, a fiscal and taxation data acquisition module, a fiscal and taxation data storage module, a fiscal and taxation feature extraction module, a model training module and a decision model construction module, and the fiscal and taxation data storage module, the fiscal and taxation feature extraction module, the model training module and the decision model construction module are controlled by the main controller module;

[0035] Through the financial and taxation data collection module, various financial and taxation-related data that have been previously or publicly disclosed by other companies, including financial statements, tax records, transaction records, etc., are stored and a database is established through the data storage module;

[0036] Through the model training module, the pre-trained model is trained and the accuracy of fiscal and tax feature extraction is improved. The fiscal and tax feature extraction includes graphic data and text data, and finally a decision model is generated;

[0037] Through the decision module, when the data to be analyzed is input into the model decision module, the analysis results can be generated.

[0038] (III) Beneficial effects

[0039] Compared with the prior art, the present invention provides a financial and tax data analysis method and system based on big data, which has the following beneficial effects:

[0040] 1. The financial and tax data analysis method based on big data, by setting step one, collects a variety of past data including text and images, financial statements, tax records, transaction records, tax invoices, etc., including normal data and abnormal data. By setting step two, the collected data is cleaned and standardized, including data deletion and filling, so that it has uniformity, reduces the redundancy of data calculation, and is conducive to improving the prediction speed. By setting step three, the image features are extracted through the convolutional neural network, and the time series features are analyzed through the recurrent neural network. The two-way joint action makes the feature extraction more efficient. By setting step four, a student model, i.e., a pre-training model, and a teacher model, i.e., a deep learning model, are established respectively. Through knowledge distillation, the student model is adjusted and improved to have the ability to automatically update and ensure better data analysis. By setting step five, the data to be analyzed is input into the decision model to obtain a quick response, thereby obtaining the data analysis results to assist the company's next stage of work.

[0041] 2. The financial and taxation data analysis system based on big data uses the financial and taxation data acquisition module to collect various financial and taxation-related data from the past or that have been disclosed by other companies, including financial statements, tax records, transaction records, etc., and stores them through the data storage module and establishes a database. Through the model training module, the pre-trained model is trained to improve the accuracy of financial and taxation feature extraction. The financial and taxation feature extraction includes graphic data and text data, and finally generates a decision model. Through the decision module, when the data to be analyzed is input into the model decision module, the analysis result can be generated. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of the financial and tax data analysis method of the present invention;

[0043] Figure 2 It is a structural schematic diagram of the financial and taxation data analysis system of the present invention. DETAILED DESCRIPTION

[0044] 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.

[0045] See also Figure 1-2 , a financial and tax data analysis method based on big data, characterized in that it includes the following steps:

[0046] Step 1: Data collection: collect various financial and tax-related data of the company in the past or other companies that have been made public, including financial statements, tax records, transaction records, etc.

[0047] Step 2: Clean, format and standardize the collected data and establish a database for subsequent analysis;

[0048] Step 3: Build a deep learning network for finance and taxation, including convolutional neural network (CNN) and recurrent neural network (RNN);

[0049] Step 4: Establish a pre-trained model and a deep learning model respectively, input the data of the database in step 2 into the pre-trained model and the deep learning model respectively, perform knowledge distillation, extract the characteristics of financial and tax data and identify the patterns, so as to train and improve the pre-trained model and generate a decision model;

[0050] Step 5: Use the trained decision model to conduct financial and tax analysis and forecasting.

[0051] exist Figure 1 In , the training of the pre-trained model includes the following steps:

[0052] S1 inputs the tax data in step 2 into the pre-trained model in step 4, thereby obtaining a first output result;

[0053] S2 inputs the financial and tax documents in step 2 into the deep learning model in step 4, thereby obtaining a second output result, and determining a first true loss value between the first output result and the second output result;

[0054] S3 inputs the verification data into the deep learning model to obtain a third output result, and compares the third output result with the second output result to obtain a second true loss value;

[0055] S4 determines the loss value according to the first loss value, the second loss value and knowledge distillation, and adjusts the parameters of the pre-trained model to improve the deep learning model and form a decision model.

[0056] exist Figure 1In the data preprocessing of step 2, the data preprocessing includes data extraction: using Left, Right, Mid, Find and other functions to extract the specified value; data cleaning: using Trim function to remove extra spaces; data replacement: using Replace and Substitute functions to replace the specified string; and data merging: using Concat, Phonetic, Textjoin functions to merge strings.

[0057] exist Figure 1 In the feature extraction process, the following steps are involved:

[0058] 1) Extract image features using CNN;

[0059] 2) Use RNN for time series analysis.

[0060] exist Figure 1 In the convolutional neural network, the convolutional neural network mainly includes convolutional layers and pooling layers. In the convolution operation, the convolution kernel scans the input image, performs weighted summation on the pixel values ​​at each position, and obtains the corresponding pixel value of the output feature map. Specifically, let the input image be X, the convolution kernel be K, and the bias be b. Then the output feature map Y formula of the convolution operation is as follows:

[0061]

[0062] C represents the number of input channels; K represents the size of the convolution kernel; i and j represent the positions on the output feature map respectively; p and q represent the positions within the convolution kernel respectively; represents the convolution operation.

[0063] Preferably, when identifying tax violations in a classification model, cross entropy loss or Dice Loss can be used to optimize the model's predictive ability for different types of tax violations. The cross entropy loss calculation formula is:

[0064]

[0065] Where P is the actual distribution and Q is the expected distribution.

[0066] exist Figure 1 In the Dice Loss calculation formula:

[0067]

[0068] Among them, pi represents the value of the i-th pixel in the predicted result, yi represents the value of the i-th pixel in the true result, X represents the pixel label of the true segmented image, and Y represents the pixel category of the segmented image predicted by the model. This formula is based on the pixel-level loss function and uses the similarity between the predicted result and the true result as the measurement standard.

[0069] exist Figure 1In the case of using a regression model to predict tax revenue, MSE or MAE can be used as the loss function to evaluate the accuracy of the model prediction. The mean square error calculation formula is:

[0070]

[0071] y i is the true value, i is the predicted value, N is the number of data points, and this formula is used to calculate the mean square error between the predicted value and the true value, which is a commonly used indicator for evaluating the prediction performance of the model.

[0072] exist Figure 2 In the present invention, the financial and taxation data analysis system includes a main controller module, a financial and taxation data acquisition module, a financial and taxation data storage module, a financial and taxation feature extraction module, a model training module and a decision model construction module, and the financial and taxation data storage module, the financial and taxation feature extraction module, the model training module and the decision model construction module are controlled by the main controller module;

[0073] Through the financial and taxation data collection module, various financial and taxation-related data that have been previously or publicly disclosed by other companies, including financial statements, tax records, transaction records, etc., are stored and a database is established through the data storage module;

[0074] Through the model training module, the pre-trained model is trained and the accuracy of fiscal and tax feature extraction is improved. The fiscal and tax feature extraction includes graphic data and text data, and finally a decision model is generated;

[0075] Through the decision module, when the data to be analyzed is input into the model decision module, the analysis results can be generated

[0076] To sum up, the financial and taxation data analysis system based on big data uses the financial and taxation data acquisition module to collect various financial and taxation-related data from the past or that have been disclosed by other companies, including financial statements, tax records, transaction records, etc., and stores them through the data storage module and establishes a database. Through the model training module, the pre-trained model is trained to improve the accuracy of financial and taxation feature extraction. The financial and taxation feature extraction includes graphic data and text data, and finally a decision model is generated. Through the decision module, when the data to be analyzed is input into the model decision module, the analysis result can be generated.

[0077] In addition, data collection is first performed, including various financial and tax-related data from the company's past or other companies' public disclosures. The collected data is then cleaned, formatted, and standardized to establish a database for subsequent analysis. A financial and tax deep learning network is built, including a convolutional neural network (CNN) and a recurrent neural network (RNN). A pre-trained model and a deep learning model are established respectively. The data from the database in step 2 are input into the pre-trained model and the deep learning model respectively to perform knowledge distillation, extract the features of the financial and tax data, and identify the patterns, thereby training and improving the pre-trained model, generating a decision model, and using the trained decision model for financial and tax analysis and prediction.

[0078] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.

[0079] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A financial and tax data analysis method based on big data, characterized by: The steps include: Step 1: Data collection: collect various financial and tax-related data of the company in the past or other companies that have been made public, including financial statements, tax records, transaction records, etc. Step 2: Clean, format and standardize the collected data and establish a database for subsequent analysis; Step 3: Build a deep learning network for finance and taxation, including convolutional neural network (CNN) and recurrent neural network (RNN); Step 4: Establish a pre-trained model and a deep learning model respectively, input the data of the database in step 2 into the pre-trained model and the deep learning model respectively, perform knowledge distillation, extract the characteristics of financial and tax data and identify the patterns, so as to train and improve the pre-trained model and generate a decision model; Step 5: Use the trained decision model to conduct financial and tax analysis and forecasting.

2. The method for analyzing financial and tax data based on big data according to claim 1, characterized in that: The training of the pre-trained model includes the following steps: S1 inputs the tax data in step 2 into the pre-trained model in step 4, thereby obtaining a first output result; S2 inputs the financial and tax documents in step 2 into the deep learning model in step 4, thereby obtaining a second output result, and determining a first true loss value between the first output result and the second output result; S3 inputs the verification data into the deep learning model to obtain a third output result, and compares the third output result with the second output result to obtain a second true loss value; S4 determines the loss value according to the first loss value, the second loss value and knowledge distillation, and adjusts the parameters of the pre-trained model to improve the deep learning model and form a decision model.

3. The method for analyzing financial and tax data based on big data according to claim 1, characterized in that: The data preprocessing in step 2 includes data extraction: using Left, Right, Mid, Find and other functions to extract specified values; data cleaning: using Trim function to clear extra spaces; data replacement: using Replace and Substitute functions to replace specified strings and data merging: using Concat, Phonetic, Textjoin functions to merge strings.

4. The method for analyzing financial and tax data based on big data according to claim 1, characterized in that: The feature extraction step comprises the following steps: 1) Extract image features using CNN; 2) Use RNN for time series analysis.

5. The method for analyzing financial and tax data based on big data according to claim 1, characterized in that: The convolutional neural network mainly includes convolution layers and pooling layers. In the convolution operation, the convolution kernel scans the input image, performs weighted summation on the pixel values ​​at each position, and obtains the corresponding pixel value of the output feature map. Specifically, assuming that the input image is X, the convolution kernel is K, and the bias is b, then the output feature map Y formula of the convolution operation is as follows: C represents the number of input channels; K represents the size of the convolution kernel; i and j represent the positions on the output feature map respectively; p and q represent the positions within the convolution kernel respectively; represents the convolution operation.

6. The method for analyzing financial and tax data based on big data according to claim 1, characterized in that: When identifying tax violations in a classification model, cross entropy loss or Dice Loss can be used to optimize the model's predictive ability for different types of tax violations. The cross entropy loss calculation formula is: Where P is the actual distribution and Q is the expected distribution.

7. The method for analyzing financial and tax data based on big data according to claim 1, characterized in that: The DiceLoss calculation formula: Among them, pi represents the value of the i-th pixel in the predicted result, yi represents the value of the i-th pixel in the true result, X represents the pixel label of the true segmented image, and Y represents the pixel category of the segmented image predicted by the model. This formula is based on the pixel-level loss function and uses the similarity between the predicted result and the true result as the measurement standard.

8. The method for analyzing financial and tax data based on big data according to claim 1, characterized in that: When using the regression model to predict tax revenue, MSE or MAE can be used as the loss function to evaluate the accuracy of the model prediction. The mean square error calculation formula is: y i is the true value, y^i is the predicted value, and N is the number of data points. This formula is used to calculate the mean square error between the predicted value and the true value, which is a common indicator for evaluating the prediction performance of the model.

9. A financial and tax data analysis system based on big data, comprising the financial and tax data analysis method of claims 1 to 7 and also comprising a financial and tax analysis system, characterized in that: The fiscal and taxation data analysis system includes a main controller module, a fiscal and taxation data acquisition module, a fiscal and taxation data storage module, a fiscal and taxation feature extraction module, a model training module and a decision model construction module, and the fiscal and taxation data storage module, the fiscal and taxation feature extraction module, the model training module and the decision model construction module are controlled by the main controller module; Through the financial and taxation data collection module, various financial and taxation-related data that have been previously or publicly disclosed by other companies, including financial statements, tax records, transaction records, etc., are stored and a database is established through the data storage module; Through the model training module, the pre-trained model is trained and the accuracy of fiscal and tax feature extraction is improved. The fiscal and tax feature extraction includes graphic data and text data, and finally a decision model is generated; Through the decision module, when the data to be analyzed is input into the model decision module, the analysis results can be generated.

Images