Value-added tax automatic calculation method based on machine learning
Through the automatic VAT calculation method based on machine learning, the problems of inefficiency and poor adaptability in the existing technology are solved, and the automated processing of VAT is realized, which improves calculation accuracy and adaptability, reduces maintenance costs, and ensures the stability of the system.
Patent Information
- Application Number
- CN202510503866.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-22
- Publication Date
- 2025-09-05
AI Technical Summary
In the prior art, the calculation and withdrawal of VAT is inefficient, susceptible to human factors, difficult to adapt to rapidly changing market and regulatory requirements, and lack real-time data updates and sharing, affecting the efficiency and accuracy of corporate decision-making.
Using machine learning-based methods, through data collection, cleaning, feature extraction and model training, the decision tree model is used to automatically calculate VAT, and a continuous model detection mechanism is set to realize the automated processing of VAT.
It improves the automation level of VAT calculation, reduces the error rate, improves work efficiency, has strong adaptability, can quickly adapt to changes in tax laws and corporate business, reduces maintenance costs, and ensures the stability and reliability of the system.
Smart Images
Figure CN120597102A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of tax technology, and in particular to a method for automatically calculating value-added tax based on machine learning. Background Art
[0002] In the current financial processing flow, the calculation and extraction of VAT usually needs to be done manually by financial personnel.
[0003] The shortcomings of this traditional financial processing process include, but are not limited to, the following: First, manual VAT calculation and withdrawal is not only inefficient but also susceptible to human error, leading to miscalculations. Second, as a company's business volume increases, the workload of financial personnel also increases, which not only increases labor costs but also can affect the quality of their work due to overwork. Furthermore, manual processing lacks flexibility and is difficult to adapt to rapidly changing market and regulatory requirements. Finally, traditional methods make it difficult to achieve real-time data updates and sharing, which is particularly evident when multiple departments collaborate, affecting the efficiency and accuracy of corporate decision-making.
[0004] Therefore, it is necessary to design an automatic VAT calculation method based on machine learning. Summary of the Invention
[0005] The purpose of the present invention is to provide a method for automatic calculation of value-added tax based on machine learning to solve the problems raised in the above background technology.
[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:
[0007] A method for automatically calculating value-added tax based on machine learning, comprising the following steps:
[0008] S100, Data Collection: Collect data from various systems;
[0009] S200, data cleaning: filtering and cleaning based on feature selection method;
[0010] S300, Feature Extraction: Extract key dimensions from cleaned data;
[0011] S400, model training: using a decision tree model for training;
[0012] S500, model application: extracting API for system call.
[0013] According to the above technical solution, the specific method of S100 is as follows:
[0014] Data personnel obtain data from the customer's existing system, or complete it through the data synchronization function provided by the management and accounting system. Through the given mapping logic, the existing data is converted into a data structure that meets the management and accounting standards and stored in the management and accounting system for subsequent operations.
[0015] According to the above technical solution, the specific method of S200 is as follows:
[0016] Pre-processing is done through a multi-layer data cleaning mechanism, namely, denoising, normalizing, and processing outliers on the incoming data to ensure data accuracy and consistency;
[0017] The above-mentioned multi-layer data cleaning mechanism includes but is not limited to using statistical analysis methods to identify and eliminate outliers, using data interpolation technology to fill missing data, and optimizing data sets through feature selection algorithms.
[0018] According to the above technical solution, the specific method of S300 is as follows:
[0019] By studying historical documents, the information gain of different document features on the tax rate is calculated based on the entropy increase algorithm, namely:
[0020] G(D,A)=H(D)-H(D|A)
[0021] Where H(D) represents the empirical entropy of historical documents, which encompasses the impact of all features on the tax rate. H(D|A) represents the impact of feature A on the tax rate of historical documents. Taking business type as the feature, H(D|A) represents the empirical conditional entropy of the tax rate of historical documents under different business types.
[0022] Similarly, different dimensions of the document are used as features to calculate the information gain of each dimension on the tax rate, and finally the five categories of features with the largest information gain are extracted, including responsibility center, business type, factor, product, and product type.
[0023] According to the above technical solution, the specific method of S400 is as follows:
[0024] Construct multiple weak learners and combine the prediction results of all weak learners through ensemble learning methods to form a strong learner, thereby improving the generalization ability and stability of the model;
[0025] The above-mentioned weak learner is constructed based on a decision tree model, which performs voting or average prediction, and the root node of the decision tree model is set based on the business type.
[0026] According to the above technical solution, the specific method of S500 is as follows:
[0027] Set up a continuous model detection mechanism and a continuous model evaluation mechanism to regularly evaluate and diagnose the model's performance. Based on the set prediction error threshold, when the evaluation and diagnosis error exceeds the threshold, retrain or adjust the model.
[0028] Compared with the prior art, the beneficial effects achieved by the present invention are:
[0029] 1. High degree of automation: Machine learning technology automates the VAT extraction process, reducing the need for manual intervention and effectively lowering the error rate caused by human error. Through automated processing, the system can complete VAT extraction more efficiently and accurately, greatly improving overall work efficiency.
[0030] 2. Strong adaptability: The system can calculate VAT based on the specific needs of different industries and product types. This wide applicability enables the invention to leverage its advantages in various business scenarios and meet diverse tax processing needs.
[0031] 3. Easy to maintain: By continuously optimizing the machine learning model, it can quickly adapt to changes in tax laws and adjustments to corporate business operations, eliminating the need for frequent large-scale system upgrades or maintenance. This reduces user maintenance costs and ensures the long-term stability and reliability of the system. BRIEF DESCRIPTION OF THE DRAWINGS
[0032] Figure 1 It is a flowchart of the present invention;
[0033] Figure 2 Decision tree model and root node example diagram of the present invention;
[0034] Figure 3 A flowchart of the decision model application and automatic adjustment process of the present invention;
[0035] Figure 4 It is a flowchart of the data collection step of the present invention. DETAILED DESCRIPTION
[0036] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described 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 making creative efforts are within the scope of protection of the present invention.
[0037] like Figure 1-4 As shown, the present invention provides a technical solution: a method for automatically calculating value-added tax based on machine learning, comprising the following steps:
[0038] S100, Data Collection: Collect data from various systems;
[0039] S200, data cleaning: filtering and cleaning based on feature selection method;
[0040] S300, Feature Extraction: Extract key dimensions from cleaned data;
[0041] S400, model training: using a decision tree model for training;
[0042] S500, model application: extracting API for system call.
[0043] Specifically, the specific method of S100 is as follows:
[0044] Data personnel obtain data from the customer's existing system, or complete it through the data synchronization function provided by the management and accounting system. Through the given mapping logic, the existing data is converted into a data structure that meets the management and accounting standards and stored in the management and accounting system for subsequent operations.
[0045] Specifically, the specific method of S200 is as follows:
[0046] Pre-processing is done through a multi-layer data cleaning mechanism, namely, denoising, normalizing, and processing outliers on the incoming data to ensure data accuracy and consistency;
[0047] The above-mentioned multi-layer data cleaning mechanism includes but is not limited to using statistical analysis methods to identify and eliminate outliers, using data interpolation technology to fill missing data, and optimizing data sets through feature selection algorithms.
[0048] Specifically, the specific method of S300 is as follows:
[0049] By studying historical documents, the information gain of different document features on the tax rate is calculated based on the entropy increase algorithm, namely:
[0050] G(D,A)=H(D)-H(D|A)
[0051] Where H(D) represents the empirical entropy of historical documents, which encompasses the impact of all features on the tax rate. H(D|A) represents the impact of feature A on the tax rate of historical documents. Taking business type as the feature, H(D|A) represents the empirical conditional entropy of the tax rate of historical documents under different business types.
[0052] Similarly, different dimensions of the document are used as features to calculate the information gain of each dimension on the tax rate, and finally the five categories of features with the largest information gain are extracted, including responsibility center, business type, factor, product, and product type.
[0053] Specifically, the specific method of S400 is as follows:
[0054] Construct multiple weak learners and combine the prediction results of all weak learners through ensemble learning methods to form a strong learner, thereby improving the generalization ability and stability of the model;
[0055] The above-mentioned weak learner is constructed based on a decision tree model, which performs voting or average prediction, and the root node of the decision tree model is set based on the business type.
[0056] Specifically, the specific method of S500 is as follows:
[0057] Set up a continuous model detection mechanism and a continuous model evaluation mechanism to regularly evaluate and diagnose the model's performance. Based on the set prediction error threshold, when the evaluation and diagnosis error exceeds the threshold, retrain or adjust the model.
[0058] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.
[0059] Finally, it should be noted that the above descriptions are merely preferred embodiments of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art will be able to modify the technical solutions described in the aforementioned embodiments or substitute equivalents for some of the technical features. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the present invention.
Claims
1. A method for automatic calculation of value-added tax based on machine learning, characterized in that: The following steps are involved: S100, Data Collection: Collect data from various systems; S200, data cleaning: filtering and cleaning based on feature selection method; S300, Feature Extraction: Extract key dimensions from cleaned data; S400, model training: using a decision tree model for training; S500, model application: extracting API for system call.
2. The method for automatic value-added tax calculation based on machine learning according to claim 1, characterized in that: The specific method of S100 is as follows: Data personnel obtain data from the customer's existing system, or complete it through the data synchronization function provided by the management and accounting system. Through the given mapping logic, the existing data is converted into a data structure that meets the management and accounting standards and stored in the management and accounting system for subsequent operations.
3. The method for automatic value-added tax calculation based on machine learning according to claim 1, characterized in that: The specific method of S200 is as follows: Pre-processing is done through a multi-layer data cleaning mechanism, namely, denoising, normalizing, and processing outliers on the incoming data to ensure data accuracy and consistency; The above-mentioned multi-layer data cleaning mechanism includes but is not limited to using statistical analysis methods to identify and eliminate outliers, using data interpolation technology to fill missing data, and optimizing data sets through feature selection algorithms.
4. The method for automatic value-added tax calculation based on machine learning according to claim 1, characterized in that: The specific method of S300 is as follows: By studying historical documents, the information gain of different document features on the tax rate is calculated based on the entropy increase algorithm, namely: G(D,A)=H(D)-H(D|A) Where H(D) represents the empirical entropy of historical documents, which encompasses the impact of all features on the tax rate. H(D|A) represents the impact of feature A on the tax rate of historical documents. Taking business type as the feature, H(D|A) represents the empirical conditional entropy of the tax rate of historical documents under different business types. Similarly, different dimensions of the document are used as features to calculate the information gain of each dimension on the tax rate, and finally the five categories of features with the largest information gain are extracted, including responsibility center, business type, factor, product, and product type.
5. The method for automatic value-added tax calculation based on machine learning according to claim 1, characterized in that: The specific method of S400 is as follows: Construct multiple weak learners and combine the prediction results of all weak learners through ensemble learning methods to form a strong learner, thereby improving the generalization ability and stability of the model; The above-mentioned weak learner is constructed based on a decision tree model, which performs voting or average prediction, and the root node of the decision tree model is set based on the business type.
6. The method for automatic value-added tax calculation based on machine learning according to claim 1, characterized in that: The specific method of S500 is as follows: Set up a continuous model detection mechanism and a continuous model evaluation mechanism to regularly evaluate and diagnose the model's performance. Based on the set prediction error threshold, when the evaluation and diagnosis error exceeds the threshold, retrain or adjust the model.