A finance and tax data analysis system
Through the data acquisition, voucher verification, input analysis, and rule matching modules of the financial and tax data analysis system, the problems of duplicate vouchers and consistency comparison in financial and tax data have been solved, and the accuracy and comprehensiveness of financial and tax risk monitoring have been achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GUIZHOU AISINO AEROSPACE INFORMATION CO LTD
- Filing Date
- 2026-05-29
- Publication Date
- 2026-07-14
AI Technical Summary
Existing financial and tax data analysis systems cannot effectively identify duplicate uploaded or reimbursed electronic vouchers, and the consistency comparison between business transactions and original vouchers is insufficient, resulting in inadequate accuracy and comprehensiveness in financial and tax risk monitoring.
The data acquisition module collects and preprocesses financial and tax data in real time, the voucher verification module performs duplicate detection and authenticity verification, the input analysis module classifies and sorts vouchers, the rule matching module performs condition matching, and the anomaly monitoring module performs consistency comparison and outputs abnormal transaction warning information.
It significantly improves the automation and reliability of financial and tax data processing, ensures the authenticity of vouchers, identifies and eliminates forged vouchers, simulates the optimal VAT deduction path, improves the timeliness and accuracy of input tax credit certification, and achieves real-time matching of risk identification with policies and regulations, avoiding misjudgments and omissions.
Smart Images
Figure CN122390896A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of financial data processing technology, and in particular to a financial and tax data analysis system. Background Technology
[0002] In the process of corporate financial and tax management, original vouchers (such as invoices, receipts, bank statements, etc.) are the direct basis for recording economic transactions, clarifying economic responsibilities, and also the foundation for accounting and tax declaration. With the expansion of business scale and the increase in transaction frequency, the amount of financial and tax data is growing explosively. Traditional financial and tax data analysis systems mainly rely on manual entry, rule verification, and batch import to perform preliminary processing of voucher information and to reconcile accounts based on the business flow records in the financial software.
[0003] Currently, common financial and tax data analysis systems typically possess functions such as voucher image storage, automatic account matching, and online invoice authenticity verification (e.g., single-document verification via tax platform interface). However, in practical applications, due to the large number, diverse formats, and complex sources of original vouchers, existing technologies generally lack the ability to proactively and in batches verify the authenticity of the vouchers themselves, especially struggling to effectively identify duplicate uploaded or reimbursed electronic vouchers (such as duplicate printed electronic invoices and duplicate scanned paper vouchers). Furthermore, in terms of consistency comparison between business transactions and original vouchers, existing technologies often rely solely on single fields such as amount or date for rough matching, failing to achieve automatic comparison of multi-dimensional information such as business logic, counterparties, and summary content. This makes it difficult to promptly detect inconsistencies between vouchers and business records (such as discrepancies between voucher and transaction amounts, voucher dates and transaction times, and differences between voucher content and business substance), thereby reducing the comprehensiveness and accuracy of financial and tax risk monitoring and posing potential tax compliance risks and financial losses to enterprises.
[0004] Chinese Patent Publication No. CN118780926A discloses a big data-based intelligent financial and tax analysis method and system. The method includes: pre-constructing a rule tree based on knowledge mining; classifying collected financial and tax information of the company and publicly available financial and tax information of other companies; matching each collected information with the rule tree to obtain sets of financial and tax information of the company and other companies; extracting metadata of the financial and tax information of the two types of companies from the two sets of information with field name association structures or keyword association structures; generating two sets of metadata of the financial and tax information of the two types of companies based on the logical relationship of the rule tree; matching each set of metadata of the financial and tax information with a normalized template; extracting normalized feature groups of the company's financial and tax information from each set of metadata of the financial and tax information; and aggregating them into normalized big data; using deep learning to analyze the normalized big data to discover the correlation patterns between the normalized big data, thereby discovering the correlation patterns between the company's financial and tax information. However, this intelligent financial and tax analysis method only collects and analyzes financial and tax data or provides tiered early warnings. It cannot perform deduplication and authenticity verification of original vouchers or consistency comparison between business transactions and vouchers, leading to omissions in financial and tax risk monitoring and reducing the accuracy of financial and tax risk monitoring. Summary of the Invention
[0005] To address this issue, the present invention provides a financial and tax data analysis system to overcome the problem that existing technologies cannot perform deduplication and authenticity verification of original vouchers or consistency comparison between business transactions and vouchers, resulting in omissions in financial and tax risk monitoring and reducing the accuracy of financial and tax risk monitoring.
[0006] To achieve the above objectives, the present invention provides a financial and tax data analysis system, comprising: The data acquisition module is used to acquire a financial and tax dataset containing original voucher data and business transaction data from the target enterprise's financial system, and to preprocess the financial and tax dataset to obtain preprocessed data. The voucher verification module is used to perform duplicate detection on the original voucher data in the preprocessed data according to the preset duplicate verification rules to obtain deduplicated voucher data, and to verify the authenticity of the deduplicated voucher data based on the preset voucher authenticity verification rules to obtain a compliant voucher dataset. The input analysis module is used to divide each compliant voucher in the compliant voucher dataset by tax amount attribute features to obtain output vouchers and input vouchers, and sort the input vouchers according to the tax amount of the output vouchers to obtain the input voucher authentication sequence. The rule matching module is used to obtain real-time updated tax rule data, and perform condition matching on the input voucher authentication sequence and the business flow data in the preprocessed data based on the tax rule data to obtain rule matching results containing preferential category markers and risk point markers; The anomaly monitoring module is used to obtain risk business transaction records associated with the risk point markers in the preprocessed business transaction data based on the risk point markers in the rule matching results, and to perform a consistency comparison between the risk business transaction records and the corresponding compliance certificates in the compliance certificate dataset to obtain a consistency comparison difference. When the consistency comparison difference exceeds a preset deviation threshold, an abnormal transaction warning message is output.
[0007] The technical principle of this application is as follows: A financial and tax dataset containing original voucher data and business transaction data is obtained from the target enterprise's financial system. This dataset is preprocessed, and duplicate records are removed from the original voucher data after preprocessing. The authenticity of the voucher elements is then verified according to preset voucher authenticity verification rules to obtain a compliant voucher dataset. By extracting tax amount attribute features from the compliant vouchers, the dataset is divided into sales vouchers and input vouchers. The sales tax amount is calculated, and the input vouchers are sorted based on their sales tax amounts to establish an input voucher authentication sequence. Finally, by accessing real-time updated tax rule data, the input vouchers are matched sequentially. The system identifies voucher information in the certificate sequence and transaction records in the preprocessed business transaction data. Items that meet the preferential policy are marked as preferential category tags, while items that exceed risk indicators are marked as risk point tags. Business transaction data is retrieved through risk point tags to obtain risky business transaction records associated with the risk tags. Then, the key elements such as the amount and subject of the risky business transaction record are compared item by item with the elements of the corresponding voucher in the compliant voucher dataset to calculate the consistency comparison difference. The consistency comparison difference is compared with a preset deviation threshold. When the consistency comparison difference exceeds the preset deviation threshold, the system determines that the voucher and transaction data do not match and outputs an abnormal transaction warning message.
[0008] Compared with existing technologies, the beneficial effects of this application are as follows: By acquiring a financial and tax dataset containing original voucher data and business transaction data, preprocessing the dataset, and performing duplicate detection on the original voucher data in the preprocessed data to obtain deduplicated voucher data, redundant invoices can be eliminated. The authenticity of the deduplicated voucher data is then verified based on preset voucher authenticity verification rules to obtain a compliant voucher dataset, which can identify and remove forged, altered, or non-compliant invoices, preventing false invoices from contaminating the overall analysis results. Furthermore, each compliant voucher in the compliant voucher dataset is divided into sales vouchers and input vouchers based on tax amount attributes. The input vouchers are then sorted according to the tax amount of the sales vouchers to obtain an input voucher authentication sequence. This can simulate the optimal VAT deduction path, improve the timeliness and accuracy of input tax authentication, and avoid tax losses caused by authentication omissions or delays. By acquiring real-time updated tax rule data and performing conditional matching on the input voucher authentication sequence and preprocessed business transaction data, rule matching results containing preferential category markers and risk point markers are obtained. This ensures that risk identification remains consistent with the latest policies and regulations, avoiding misjudgments caused by rule lag. Using the risk point markers in the rule matching results, risky business transaction records associated with these markers are retrieved from the preprocessed business transaction data. These risky business transaction records are then compared with the corresponding compliant vouchers in the compliant voucher dataset to obtain the consistency difference. When the consistency difference exceeds a preset deviation threshold, an abnormal transaction warning is output. This enables precise reconciliation and deviation quantification of business flows, invoice flows, and financial flows, avoiding underreporting and misreporting caused by relying solely on single-dimensional indicators, ultimately significantly improving the accuracy of tax and financial risk monitoring.
[0009] Furthermore, the data acquisition module includes: The data acquisition unit is used to collect original voucher data, including VAT invoice images and electronic receipts, as well as business flow data, including bank transaction records and general ledger entries, in real time. The data integration unit is used to perform association matching between the original voucher data and business transaction data to obtain the associated and matched financial and tax data, and to check the associated and matched financial and tax data, and to add the financial and tax data in which all required fields are non-empty and the association identifier is unique to the financial and tax dataset. The data preprocessing unit preprocesses the financial and tax dataset to obtain preprocessed data.
[0010] This solution collects original vouchers and business transaction data in real time, performs correlation matching, integrity detection and deduplication, and preprocesses the financial and tax dataset to obtain standardized, accurate and non-redundant high-quality financial and tax data, which significantly improves the automation and reliability of financial and tax data processing.
[0011] Furthermore, the credential verification module includes: The voucher deduplication unit is used to extract the original voucher data according to the preset duplicate verification rules, to obtain an intermediate voucher dataset containing voucher type code, invoice date string and amount value, and to determine the hash fingerprint value based on the intermediate voucher dataset. The hash fingerprint value is then compared with the stored hash fingerprint value set to remove duplicate vouchers with the same hash fingerprint value, thus obtaining deduplicated voucher data. The authenticity verification unit is used to extract the electronic signature data and tax control anti-counterfeiting code field of each voucher from the deduplicated voucher data based on the preset voucher authenticity verification rules. It performs elliptic curve digital signature verification on the electronic signature data through a preset public key certificate, and decrypts the tax control anti-counterfeiting code field and checks its consistency with the plaintext field. The vouchers that pass the dual verification are identified as compliant vouchers, forming a compliant voucher dataset.
[0012] This solution eliminates duplicate vouchers by comparing hash fingerprints and uses dual verification of electronic signatures and tax control anti-counterfeiting codes to effectively prevent duplicate entries and forged vouchers, ensuring that each voucher is authentic and unique, forming a compliant voucher dataset, and significantly improving the accuracy of financial and tax verification.
[0013] Furthermore, the input analysis module includes: The tax amount attribute extraction unit is used to extract the compliant voucher dataset to obtain voucher tax amount attribute records containing voucher identifier, tax amount and input / output tax flags; The voucher division unit is used to combine vouchers marked as sales into a sales voucher set and vouchers marked as input into an input voucher set according to the input and output tax indicators in the tax amount attribute record of the vouchers, and to sum up the tax amount of all vouchers in the sales voucher set to obtain the total sales tax amount. The authentication sequence generation unit is used to calculate the authentication priority score for each input voucher in the input voucher set with reference to the total output tax amount, and sort each input voucher in the input voucher set in descending order based on the authentication priority score to obtain the input voucher authentication sequence.
[0014] This scheme calculates and sorts the authentication priority of input vouchers by dividing them into sales and input vouchers and using the total sales tax amount as a reference. This allows for the rational planning of the input authentication order, improves tax processing efficiency, and ensures that input tax is deducted in a timely and effective manner.
[0015] Furthermore, in the authentication sequence generation unit, the mathematical expression for the authentication priority score is: In the formula, Indicates the first The certification priority scoring of each input invoice. Indicates the first The tax amount on each input invoice. This represents the total output tax. Indicates the first The deadline for certification of each input invoice. Indicates the current system date. This indicates the preset benchmark period value. This represents the natural logarithm function.
[0016] This solution combines the amount of tax on the voucher, the overall scale of output tax, the certification deadline, the current system time, and the preset time limit to comprehensively evaluate the certification priority. This approach can meet the timeliness requirements of tax business, scientifically and rationally arrange the order of voucher certification, and avoid the risk of overdue voucher certification.
[0017] Furthermore, the rule matching module includes: The rule parsing unit is used to parse real-time updated tax rule data to obtain preferential condition templates containing applicable condition fields and risk condition templates containing risk indicator fields. The condition matching unit is used to traverse the business transaction data in the preprocessed data, match the fields in the discount condition template with the attributes of the business transaction records in the preprocessed data, output the discount category label, and perform field consistency checks on each input voucher and the corresponding business transaction record in the input voucher authentication sequence based on the risk condition template, and generate an initial risk point label. The rule integration unit is used to calculate a risk score based on the initial risk point marker, and merge the discount category marker and the risk point marker containing the risk score to obtain a rule matching result.
[0018] In this solution, by parsing real-time updated tax rule data into structured preferential condition templates and risk condition templates, it can automatically traverse business transaction data for accurate matching, output preferential category tags, and at the same time perform consistency checks on input vouchers and business transaction records in preprocessed data to generate risk point tags. Then, it further calculates risk scores and merges preferential tags and risk scores into a complete rule matching result, which greatly improves the accuracy of tax business audit.
[0019] Furthermore, in the rule integration unit, the mathematical expression for the risk score is: In the formula, Indicates the first Each initial risk point is marked with a corresponding risk score. Indicates the first The number of risk condition templates corresponding to each initial risk point marker. This indicates the total number of risk condition templates. This represents the absolute difference between the amount on the input invoice and the corresponding transaction amount. This represents the normalized parameter for the amount deviation. Represents the natural constant. This represents the weighting coefficient used to adjust the degree of rule matching in the overall risk score. This represents the weighting coefficient used to adjust for the degree of difference in monetary amounts in the overall risk score.
[0020] This solution combines rule adaptation with the amount deviation of vouchers and business transactions, sets weights to adjust the impact ratio of each dimension, and uses a normalization method for amount deviation to comprehensively consider risk factors from multiple dimensions, making risk score calculation more in line with actual business scenarios.
[0021] Furthermore, the anomaly monitoring module includes: The risk record extraction unit is used to extract risk business transaction records with the same transaction identifier as the risk point marker in the preprocessed data's business transaction data based on the risk point marker in the matching result of the rule; The consistency comparison unit is used to compare the fields of the risk business transaction record with the compliance certificate corresponding to the certificate identifier in the compliance certificate dataset to obtain a consistency comparison difference value including the transaction amount difference value and the tax rate difference value. The early warning output unit is used to generate an abnormal transaction early warning containing risk point markers, voucher numbers, difference fields and difference values when the transaction amount difference value in the consistency comparison difference is greater than the first preset deviation threshold C1, or the tax rate difference value is greater than the second preset deviation threshold C2, and output the abnormal transaction early warning information through a message queue, wherein C1≠C2.
[0022] In this solution, by matching the corresponding risk business transaction records based on the risk point markers and then comparing the numerical differences with the compliance vouchers across all fields, deviations in transaction amounts and tax rates can be accurately identified. Abnormal situations exceeding the first or second preset deviation thresholds can be promptly determined. Key risk information is integrated to generate abnormal transaction warning information, and the warning is issued through a message queue, effectively improving the accuracy of business compliance risk control. Attached Figure Description
[0023] Figure 1 This is a schematic diagram of the structure of a financial and tax data analysis system according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the data acquisition module in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the voucher verification module in an embodiment of the present invention; Figure 4This is a schematic diagram of the structure of the input analysis module in an embodiment of the present invention; Figure 5 This is a schematic diagram of the rule matching module in an embodiment of the present invention; Figure 6 This is a schematic diagram of the anomaly monitoring module in an embodiment of the present invention. Detailed Implementation
[0024] The following detailed description illustrates the specific implementation method: like Figure 1 As shown, it is a schematic diagram of the structure of a financial and tax data analysis system according to an embodiment of the present invention, including: The data acquisition module is used to acquire a financial and tax dataset containing original voucher data and business transaction data from the target enterprise's financial system, and to preprocess the financial and tax dataset to obtain preprocessed data. The voucher verification module is used to perform duplicate detection on the original voucher data in the preprocessed data according to the preset duplicate verification rules to obtain deduplicated voucher data, and to verify the authenticity of the deduplicated voucher data based on the preset voucher authenticity verification rules to obtain a compliant voucher dataset. The input analysis module is used to divide each compliant voucher in the compliant voucher dataset by tax amount attribute features to obtain output vouchers and input vouchers, and sort the input vouchers according to the tax amount of the output vouchers to obtain the input voucher authentication sequence. The rule matching module is used to obtain real-time updated tax rule data, and perform condition matching on the input voucher authentication sequence and the business flow data in the preprocessed data based on the tax rule data to obtain rule matching results containing preferential category markers and risk point markers; The anomaly monitoring module is used to obtain risk business transaction records associated with the risk point markers in the preprocessed business transaction data based on the risk point markers in the rule matching results, and to perform a consistency comparison between the risk business transaction records and the corresponding compliance certificates in the compliance certificate dataset to obtain a consistency comparison difference. When the consistency comparison difference exceeds a preset deviation threshold, an abnormal transaction warning message is output.
[0025] like Figure 2 As shown, it is a structural diagram of the data acquisition module in an embodiment of the present invention, including: The data acquisition unit is used to collect original voucher data, including VAT invoice images and electronic receipts, as well as business flow data, including bank transaction records and general ledger entries, in real time. The data integration unit is used to perform association matching between the original voucher data and business transaction data to obtain the associated and matched financial and tax data, and to check the associated and matched financial and tax data, and to add the financial and tax data in which all required fields are non-empty and the association identifier is unique to the financial and tax dataset. The data preprocessing unit preprocesses the financial and tax dataset to obtain preprocessed data.
[0026] Furthermore, in the data acquisition unit, VAT invoice images are acquired in real time through the image acquisition channel connected to the scanning terminal. The acquisition process uses a preset scanning resolution, which is determined based on the minimum accuracy requirement of the optical character recognition engine, for example, setting the preset scanning resolution to 300 dots per inch. Electronic receipts are received in real time through the data push channel of the bank's electronic receipt system application interface, and the electronic receipt data is parsed. Optical character recognition is performed on the VAT invoice images to obtain the optical character recognition results, which are then added to the original voucher data along with the electronic receipt data. Bank transaction records are collected through real-time message subscription via the bank-enterprise direct connection interface, using a preset message subscription buffer capacity, which is determined based on the peak number of transactions, for example, setting the preset message subscription buffer capacity to 2048 records. General ledger entries are acquired in real time through the change log capture mechanism of the enterprise resource planning system database, and processed using a preset log scanning cycle, which is determined based on the database load capacity, for example, setting the preset log scanning cycle to 3 seconds. The acquired bank transaction records and general ledger entries are combined to form business transaction data. Both original voucher data and business transaction data are pushed into a temporary storage area for integration and processing.
[0027] In the data integration unit, correlation matching is processed using a preset set of matching rules, which includes a monetary tolerance threshold and a date tolerance range. The monetary tolerance threshold is determined based on the smallest currency unit, for example, setting it to 0.01 yuan. The date tolerance range is determined based on the allowable deviation of the accounting entry date, for example, setting it to one calendar day before or after. During correlation matching, the face value, invoice date, transaction amount, and transaction date of the VAT invoice image and electronic receipt are extracted from the original voucher data and compared with the transaction amount, transaction date, voucher amount, and voucher date of the bank transaction records and general ledger entries in the business transaction data. Simultaneously, the consistency of the counterparty name is verified. Records that satisfy the matching rule set generate correlation identifiers. These correlation identifiers are generated according to preset identifier combination rules, which are determined based on the uniqueness of the invoice number and bank transaction number. For example, the correlation identifier combination rule is set so that the invoice number and bank transaction number are connected by an underscore. The matched financial and tax data is then inspected to ensure that all required fields are non-empty. The list of required fields is determined based on the integrity specifications of the financial and tax dataset, for example, setting all required fields as invoice code, invoice number, total price including tax, bank transaction number, and transaction date. Simultaneously, the system checks for duplicate association identifiers within the financial and tax dataset by iterating through existing association identifier indices. Finally, the matched financial and tax data, where all required fields are non-empty and no association identifiers are duplicated, is added to the financial and tax dataset.
[0028] In the data preprocessing unit, the financial and tax dataset is first processed for missing values. The preset missing value imputation strategy is determined based on the account attributes. For example, for the amount field under the travel expense account, the missing value imputation strategy is set to fill with the monthly average of that account. Then, outlier processing is performed. The preset outlier judgment boundary is determined based on the interquartile range method. For example, the outlier judgment boundary is set as records below the lower quartile minus 1.5 times the interquartile range or above the upper quartile plus 1.5 times the interquartile range are considered outliers and removed. Next, the text fields are processed for whitespace removal and encoding standardization. The preset accounting account code standardization mapping table is determined based on the enterprise account system. For example, the account code "1001" is uniformly mapped to "cash on hand". The date field is converted according to the preset date format, which is determined based on the data storage specifications. For example, the preset date format is set to 8-digit pure numeric form, such as 20260128. Finally, the financial and tax dataset is checked for completely duplicate related identifiers. If they exist, only the first occurrence of the identifier is retained. After completion, the preprocessed data is output.
[0029] like Figure 3 As shown, it is a structural diagram of the credential verification module in an embodiment of the present invention, including: The voucher deduplication unit is used to extract the original voucher data according to the preset duplicate verification rules, to obtain an intermediate voucher dataset containing voucher type code, invoice date string and amount value, and to determine the hash fingerprint value based on the intermediate voucher dataset. The hash fingerprint value is then compared with the stored hash fingerprint value set to remove duplicate vouchers with the same hash fingerprint value, thus obtaining deduplicated voucher data. The authenticity verification unit is used to extract the electronic signature data and tax control anti-counterfeiting code field of each voucher from the deduplicated voucher data based on the preset voucher authenticity verification rules. It performs elliptic curve digital signature verification on the electronic signature data through a preset public key certificate, and decrypts the tax control anti-counterfeiting code field and checks its consistency with the plaintext field. The vouchers that pass the dual verification are identified as compliant vouchers, forming a compliant voucher dataset.
[0030] Furthermore, in the voucher deduplication unit, the preset duplicate verification rules are determined based on the combination of unique identifier elements of the voucher. For example, the preset duplicate verification rules are set to extract the invoice code field of the VAT invoice as the voucher type code, extract the invoice date field and format it as an 8-digit numeric string as the invoice date string, and extract the total price and tax field and convert it into an integer value in cents as the amount. Each record in the intermediate voucher dataset calculates a hash fingerprint value using a secure hash algorithm. The secure hash algorithm is determined based on the digest length and collision resistance requirements. For example, the secure hash algorithm is set to SHA-256, generating a 64-bit hexadecimal hash fingerprint value. The hash fingerprint value is compared with the already stored hash fingerprint value set. The already stored hash fingerprint value set is maintained in memory using a Bloom filter structure. The bit array length and the number of hash functions of the Bloom filter are determined based on the expected total number of vouchers and the acceptable error rate. For example, the bit array length of the Bloom filter is set to 536,870,912 bits, and the number of hash functions is set to 7. When a hash fingerprint value matches any element in the stored hash fingerprint value set, the corresponding duplicate voucher is removed. Unmatched vouchers are retained as deduplicated voucher data, and their hash fingerprint values are simultaneously stored in the stored hash fingerprint value set.
[0031] In the authenticity verification unit, the preset voucher authenticity verification rules are determined based on the location of anti-counterfeiting elements in the VAT invoice data specification. For example, the preset voucher authenticity verification rules are set as follows: extract the content of the XML node labeled "signature value" from the structured data of the VAT invoice as the electronic signature data, and extract the content of the XML node labeled "anti-counterfeiting code" as the tax control anti-counterfeiting code field. Elliptic curve digital signature verification is performed on the electronic signature using a pre-set public key certificate. The pre-set public key certificate is imported from the root certificate of the tax digital certificate management authority. The elliptic curve digital signature verification uses an elliptic curve cryptography scheme for signature verification. The parameters of the elliptic curve are determined according to actual management requirements; for example, the elliptic curve is set to an SM2 curve. The tax control anti-counterfeiting code field is decrypted and its consistency with the plaintext field is checked. The decryption operation is performed using a pre-set symmetric decryption key, which is negotiated and generated by the tax control device when the VAT invoice is issued and pre-stored in the verification unit's security module. After decryption, plaintext fields containing invoice code, invoice number, invoice date, amount excluding tax, and seller's taxpayer identification number are obtained. The decrypted plaintext fields are then compared item by item with the corresponding fields in the deduplicated voucher data for consistency. Vouchers that pass elliptic curve digital signature verification and whose decrypted plaintext fields match the tax control anti-counterfeiting code field are identified as compliant vouchers, forming a compliant voucher dataset.
[0032] like Figure 4 As shown, it is a structural schematic diagram of the input analysis module in an embodiment of the present invention, including: The tax amount attribute extraction unit is used to extract the compliant voucher dataset to obtain voucher tax amount attribute records containing voucher identifier, tax amount and input / output tax flags; The voucher division unit is used to combine vouchers marked as sales into a sales voucher set and vouchers marked as input into an input voucher set according to the input and output tax indicators in the tax amount attribute record of the vouchers, and to sum up the tax amount of all vouchers in the sales voucher set to obtain the total sales tax amount. The authentication sequence generation unit is used to calculate the authentication priority score for each input voucher in the input voucher set with reference to the total output tax amount, and sort each input voucher in the input voucher set in descending order based on the authentication priority score to obtain the input voucher authentication sequence.
[0033] Specifically, in the authentication sequence generation unit, the mathematical expression for the authentication priority score is: In the formula, Indicates the first The certification priority scoring of each input invoice. Indicates the first The tax amount on each input invoice. This represents the total output tax. Indicates the first The deadline for certification of each input invoice. Indicates the current system date. This indicates the preset benchmark period value. This represents the natural logarithm function.
[0034] Furthermore, in the tax amount attribute extraction unit, the extraction process of the compliant voucher dataset is executed according to preset field extraction mapping rules. These rules are determined based on the voucher structure definition in the VAT electronic invoice data specification. For example, the voucher identifier is set as a string concatenated with the invoice code and invoice number fields using underscores; the tax amount is set as the value of the tax amount field on the invoice; and the input / output flag is set as the invoice purpose type field. When the invoice purpose type field value is the output flag code, the input / output flag is marked as output; when the invoice purpose type field value is the input flag code, the input / output flag is marked as input. The tax amount attribute extraction unit combines the read voucher identifier, tax amount, and input / output flag into a single voucher tax amount attribute record.
[0035] In the voucher segmentation unit, a container for output vouchers and a container for input vouchers are constructed, both initialized as empty sets. A variable for accumulating total output tax is set and initialized to zero. Each voucher's tax amount attribute record is processed sequentially. The input / output flag value is read. When the input / output flag is the output flag, the voucher identifier and tax amount of that voucher's tax amount attribute record are stored in the output voucher set, and the tax amount of that voucher's tax amount attribute record is accumulated to the total output tax amount. When the input / output flag equals the input flag, the voucher identifier and tax amount of that voucher's tax amount attribute record are stored in the input voucher set. After processing all voucher tax amount attribute records, the output voucher set contains all output vouchers and their tax amounts, and the input voucher set contains all input vouchers and their tax amounts. The total output tax amount is the arithmetic sum of the tax amounts in the output voucher set.
[0036] In the authentication sequence generation unit, the preset baseline period is determined based on the validity period of input voucher authentication as stipulated in tax regulations, for example, setting the preset baseline period to 365 days. The current system date is the server date at the time of calculation, for example, May 6, 2024. The authentication sequence generation unit reads the... The authentication deadline for each input voucher is the last date stipulated by the tax authorities for which the voucher can be authenticated and deducted. The authentication sequence generation unit calculates the difference in days between the authentication deadline and the current system date. After calculating the authentication priority score for each input voucher in the input voucher set, the authentication sequence generation unit sorts all input vouchers in descending order of their scores. The order of the sorted results constitutes the input voucher authentication sequence.
[0037] like Figure 5 As shown, it is a structural diagram of the rule matching module in an embodiment of the present invention, including: The rule parsing unit is used to parse real-time updated tax rule data to obtain preferential condition templates containing applicable condition fields and risk condition templates containing risk indicator fields. The condition matching unit is used to traverse the business transaction data in the preprocessed data, match the fields in the discount condition template with the attributes of the business transaction records in the preprocessed data, output the discount category label, and perform field consistency checks on each input voucher and the corresponding business transaction record in the input voucher authentication sequence based on the risk condition template, and generate an initial risk point label. The rule integration unit is used to calculate a risk score based on the initial risk point marker, and merge the discount category marker and the risk point marker containing the risk score to obtain a rule matching result.
[0038] Specifically, in the rule integration unit, the mathematical expression for the risk score is: In the formula, Indicates the first Each initial risk point is marked with a corresponding risk score. Indicates the first The number of risk condition templates corresponding to each initial risk point marker. This indicates the total number of risk condition templates. This represents the absolute difference between the amount on the input invoice and the corresponding transaction amount. This represents the normalized parameter for the amount deviation. Represents the natural constant. This represents the weighting coefficient used to adjust the degree of rule matching in the overall risk score. This represents the weighting coefficient used to adjust for the degree of difference in monetary amounts in the overall risk score.
[0039] Furthermore, in the rule parsing unit, a rule pattern library is pre-built. This library contains a set of conditional trigger words and a set of risk indicator words. Conditional trigger words include terms such as "applicable," "compliant," and "satisfied," while risk indicator words include terms such as "inconsistent," "exceeding limits," and "missing." After receiving tax rule data in real time, the text of each tax rule is first segmented and tagged with parts of speech. Then, the conditional clauses and risk clauses of the rule are located based on the trigger words in the rule pattern library. The attribute names, comparison operators, and threshold values in each conditional clause are extracted to generate a preferential condition template containing applicable condition fields. For example, the applicable condition fields in the preferential condition template include "industry type equals manufacturing" and "sales amount greater than 1 million yuan." The indicator type, verification logic, and deviation range in each risk clause are extracted to generate a risk condition template containing risk indicator fields. For example, the risk indicator fields in the risk condition template include "absolute difference between input voucher amount and business transaction amount greater than 1,000 yuan" and "voucher type missing." After parsing, the preferential condition template and the risk condition template are stored as structured data objects, respectively.
[0040] In the condition matching unit, the preprocessed business transaction data is first read. Each business transaction record contains multiple attribute fields, such as transaction amount, transaction date, and taxpayer identification number. The condition matching unit iterates through each business transaction record. For the current business transaction record, it compares the value of each applicable condition field in the preferential condition template with the corresponding attribute of the business transaction record. For example, if the preferential condition template requires "taxable income greater than 1 million yuan", then it checks whether the value of the "taxable income" field in the business transaction record is greater than 1 million yuan. When all applicable condition fields match successfully, the condition matching unit outputs a preferential category tag for the business transaction record. The preferential category tag includes the applicable preferential type name and version number. Meanwhile, the condition matching unit performs field consistency checks on each input voucher in the input voucher authentication sequence based on the risk condition template. The input voucher authentication sequence records fields such as the number, amount, and invoice date of each input voucher. The condition matching unit compares the amount field of the current input voucher with the amount field recorded in the corresponding business transaction record. If the two are inconsistent, an initial risk point mark is generated. The initial risk point mark includes the inconsistent field name, the specific numerical difference, and the risk condition template number that was hit.
[0041] In the rule integration unit, a weighting coefficient is used to adjust the degree of rule matching in the overall risk score. The sensitivity requirement is determined based on historical risk verification results; for example, in the scenario of VAT special invoice risk monitoring, [the sensitivity requirement is...]. Set to 0.6; this is used to adjust the weighting of the degree of difference in monetary amount in the overall risk score. The determination is based on the degree of impact of the amount deviation on tax compliance, for example, Set to 0.4; Amount Deviation Normalization Parameter The parameter used to convert the absolute difference in amount into a dimensionless exponentially decaying term is the amount deviation normalization parameter. It is determined based on the order of magnitude of the average transaction amount per enterprise, for example, The value is set to 10,000 yuan. The rule integration unit attaches a calculated risk score to each initial risk point marker, forming a risk point marker with a risk score. Then, it merges the discount category marker with the risk point marker with the risk score, and finally outputs the rule matching result.
[0042] like Figure 6 As shown, it is a structural schematic diagram of the anomaly monitoring module in an embodiment of the present invention, including: The risk record extraction unit is used to extract risk business transaction records with the same transaction identifier as the risk point marker in the preprocessed data's business transaction data based on the risk point marker in the matching result of the rule; The consistency comparison unit is used to compare the fields of the risk business transaction record with the compliance certificate corresponding to the certificate identifier in the compliance certificate dataset to obtain a consistency comparison difference value including the transaction amount difference value and the tax rate difference value. The early warning output unit is used to generate an abnormal transaction early warning containing risk point markers, voucher numbers, difference fields, and difference values when the transaction amount difference value in the consistency comparison difference is greater than a preset transaction amount difference value, or the tax rate difference value is greater than a preset tax rate difference value, and output the abnormal transaction early warning information through a message queue.
[0043] Furthermore, in the risk record extraction unit, rule matching results are passed in as structured messages. Each rule matching result contains a risk point marker and a corresponding transaction identifier. The preprocessed business transaction data is stored in a relational database, and each business transaction record contains at least the fields of transaction identifier, transaction amount, tax rate, and voucher identifier. When the risk record extraction unit starts, it loads all business transaction records through batch queries and constructs an in-memory hash map using the transaction identifier as the key and the complete business transaction record as the value, ensuring that each query can be completed in constant time. Subsequently, the risk record extraction unit sequentially reads the rule matching result set, extracts the transaction identifier associated with each risk point marker, and uses the transaction identifier to retrieve the hash map. If a matching business transaction record is found, the risk record extraction unit copies the business transaction record and marks it as a risky business transaction record, while binding the corresponding risk point marker. If no corresponding business transaction record is found for the transaction identifier, the rule matching result is discarded and an exception log is recorded.
[0044] In the consistency comparison unit, each compliance document in the compliance document dataset has a unique document identifier. The consistency comparison unit pre-parses the compliance document dataset, using the document identifier as the key and the compliance document object as the value to construct a hash index table. For each input risk transaction record, the consistency comparison unit reads the document identifier field value from the risk transaction record and uses the document identifier to look up the compliance document in the hash index table. After finding the corresponding compliance document, the consistency comparison unit reads the transaction amount and tax rate field values from the risk transaction record and the compliance document. The transaction amount difference is determined by calculating the absolute value of the difference between the transaction amount value of the risk transaction record and the transaction amount value of the compliance document, and the tax rate difference is determined by calculating the absolute value of the difference between the tax rate value of the risk transaction record and the tax rate value of the compliance document. If the document identifier does not match in the compliance document dataset, the risk transaction record is ignored.
[0045] In the early warning output unit, the preset transaction amount difference value is determined based on the maximum tolerable amount deviation identified in the financial reconciliation over the past twelve months; for example, the preset transaction amount difference value is set to 800.00 yuan. The preset tax rate difference value is determined based on the allowable error range for VAT rate declaration published by the tax authorities; for example, the preset tax rate difference value is set to 0.05%. The early warning output unit receives each comparison result tuple and extracts the consistency comparison difference value. When the transaction amount difference value in the consistency comparison difference is greater than the preset transaction amount difference value, or the tax rate difference value is greater than the preset tax rate difference value, the early warning output unit generates an abnormal transaction early warning information. The abnormal transaction alert information includes the risk point marker associated with the current comparison, and the voucher identifier is entered as the voucher number. The difference field is dynamically set according to the threshold type: if the transaction amount difference exceeds the limit, the difference field is filled with "transaction amount", and the difference value is filled with the transaction amount difference value; if the tax rate difference exceeds the limit, the difference field is filled with "tax rate", and the difference value is filled with the tax rate difference value; if both difference values exceed the limit, the abnormal transaction alert includes both sets of difference fields and difference values. The message queue is determined according to the persistence and delivery guarantee requirements of the alert message transmission. For example, the message queue is set to a topic subscription mode that supports persistent message storage, with the topic name being "Abnormal Transaction Alert Topic". The abnormal transaction alert information is encapsulated and published through this abnormal transaction alert topic.
[0046] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.
Claims
1. A financial and tax data analysis system, characterized in that: include: The data acquisition module is used to acquire a financial and tax dataset containing original voucher data and business transaction data from the target enterprise's financial system, and to preprocess the financial and tax dataset to obtain preprocessed data. The voucher verification module is used to perform duplicate detection on the original voucher data in the preprocessed data according to the preset duplicate verification rules to obtain deduplicated voucher data, and to verify the authenticity of the deduplicated voucher data based on the preset voucher authenticity verification rules to obtain a compliant voucher dataset. The input analysis module is used to divide each compliant voucher in the compliant voucher dataset by tax amount attribute features to obtain output vouchers and input vouchers, and sort the input vouchers according to the tax amount of the output vouchers to obtain the input voucher authentication sequence. The rule matching module is used to obtain real-time updated tax rule data, and perform condition matching on the input voucher authentication sequence and the business flow data in the preprocessed data based on the tax rule data to obtain rule matching results containing preferential category markers and risk point markers; The anomaly monitoring module is used to obtain risk business transaction records associated with the risk point markers in the preprocessed business transaction data based on the risk point markers in the rule matching results, and to perform a consistency comparison between the risk business transaction records and the corresponding compliance certificates in the compliance certificate dataset to obtain a consistency comparison difference. When the consistency comparison difference exceeds a preset deviation threshold, an abnormal transaction warning message is output.
2. The financial and tax data analysis system according to claim 1, characterized in that: The data acquisition module includes: The data acquisition unit is used to collect original voucher data, including VAT invoice images and electronic receipts, as well as business flow data, including bank transaction records and general ledger entries, in real time. The data integration unit is used to perform association matching between the original voucher data and business transaction data to obtain the associated and matched financial and tax data, and to check the associated and matched financial and tax data, and to add the financial and tax data in which all required fields are non-empty and the association identifier is unique to the financial and tax dataset. The data preprocessing unit preprocesses the financial and tax dataset to obtain preprocessed data.
3. The financial and tax data analysis system according to claim 1, characterized in that: The credential verification module includes: The voucher deduplication unit is used to extract the original voucher data according to the preset duplicate verification rules, to obtain an intermediate voucher dataset containing voucher type code, invoice date string and amount value, and to determine the hash fingerprint value based on the intermediate voucher dataset. The hash fingerprint value is then compared with the stored hash fingerprint value set to remove duplicate vouchers with the same hash fingerprint value, thus obtaining deduplicated voucher data. The authenticity verification unit is used to extract the electronic signature data and tax control anti-counterfeiting code field of each voucher from the deduplicated voucher data based on the preset voucher authenticity verification rules. It performs elliptic curve digital signature verification on the electronic signature data through a preset public key certificate, and decrypts the tax control anti-counterfeiting code field and checks its consistency with the plaintext field. The vouchers that pass the dual verification are identified as compliant vouchers, forming a compliant voucher dataset.
4. The financial and tax data analysis system according to claim 1, characterized in that: The input analysis module includes: The tax amount attribute extraction unit is used to extract the compliant voucher dataset to obtain voucher tax amount attribute records containing voucher identifier, tax amount and input / output tax flags; The voucher division unit is used to combine vouchers marked as sales into a sales voucher set and vouchers marked as input into an input voucher set according to the input and output tax indicators in the tax amount attribute record of the vouchers, and to sum up the tax amount of all vouchers in the sales voucher set to obtain the total sales tax amount. The authentication sequence generation unit is used to calculate the authentication priority score for each input voucher in the input voucher set with reference to the total output tax amount, and sort each input voucher in the input voucher set in descending order based on the authentication priority score to obtain the input voucher authentication sequence.
5. The financial and tax data analysis system according to claim 4, characterized in that: In the authentication sequence generation unit, the mathematical expression for the authentication priority score is: In the formula, Indicates the first The certification priority scoring of each input invoice. Indicates the first The tax amount on each input invoice. This represents the total output tax. Indicates the first The deadline for certification of each input invoice. Indicates the current system date. This indicates the preset benchmark period value. This represents the natural logarithm function.
6. The financial and tax data analysis system according to claim 1, characterized in that: The rule matching module includes: The rule parsing unit is used to parse real-time updated tax rule data to obtain preferential condition templates containing applicable condition fields and risk condition templates containing risk indicator fields. The condition matching unit is used to traverse the business transaction data in the preprocessed data, match the fields in the discount condition template with the attributes of the business transaction records in the preprocessed data, output discount category tags, and perform field consistency checks on each input voucher and the corresponding business transaction record in the input voucher authentication sequence based on the risk condition template, and generate initial risk point tags. The rule integration unit is used to calculate a risk score based on the initial risk point marker, and merge the discount category marker and the risk point marker containing the risk score to obtain a rule matching result.
7. A financial and tax data analysis system according to claim 6, characterized in that: In the rule integration unit, the mathematical expression for the risk score is: In the formula, Indicates the first Each initial risk point is marked with a corresponding risk score. Indicates the first The number of risk condition templates corresponding to each initial risk point marker. This indicates the total number of risk condition templates. This represents the absolute difference between the amount on the input invoice and the corresponding transaction amount. This represents the normalized parameter for the amount deviation. Represents the natural constant. This represents the weighting coefficient used to adjust the degree of rule matching in the overall risk score. This represents the weighting coefficient used to adjust for the degree of difference in monetary amounts in the overall risk score.
8. The financial and tax data analysis system according to claim 1, characterized in that: The anomaly monitoring module includes: The risk record extraction unit is used to extract risk business transaction records with the same transaction identifier as the risk point marker in the preprocessed data's business transaction data based on the risk point marker in the matching result of the rule; The consistency comparison unit is used to compare the fields of the risk business transaction record with the compliance certificate corresponding to the certificate identifier in the compliance certificate dataset to obtain a consistency comparison difference value including the transaction amount difference value and the tax rate difference value. The early warning output unit is used to generate an abnormal transaction early warning containing risk point markers, voucher numbers, difference fields, and difference values when the transaction amount difference value in the consistency comparison difference is greater than a preset transaction amount difference value, or the tax rate difference value is greater than a preset tax rate difference value, and output the abnormal transaction early warning information through a message queue.