Audit method and system for data intelligent matching and intelligent analysis
By setting the collection rules of audit fields, performing fuzzy matching and similarity detection, and building a deep learning model, the problems of time-consuming, labor-intensive and uncertainty of traditional manual audit methods are solved, and the automated processing of audit data and accurate identification of violations are realized, and audit efficiency and accuracy are improved.
Patent Information
- Application Number
- CN202411407871.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-10
- Publication Date
- 2025-03-04
AI Technical Summary
Traditional manual auditing methods are time-consuming and labor-intensive, and are susceptible to human factors, resulting in uncertainty and incompleteness of audit results, making it difficult to effectively process massive audit data.
By setting the collection rules of the audit field, performing data fuzzy matching and similarity detection, building a deep learning model for intelligent analysis, realizing automatic processing of audit data and accurate identification of violations.
It improves audit efficiency and accuracy, reduces manual intervention, shortens audit cycle, and enhances the accurate identification and processing capabilities of audit data.
Smart Images
Figure BDA0005077176070000022 
Figure BDA0005077176070000072 
Figure FDA0005077176060000011
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of audit data processing, and in particular, relates to an audit method and system for intelligent data matching and intelligent analysis. Background Art
[0002] With the rapid development of information technology, the amount of enterprise data has exploded, and traditional manual auditing methods can no longer meet the needs of efficient and accurate auditing. Existing auditing methods mostly rely on manual judgment, which is not only time-consuming and labor-intensive, but also easily affected by human factors, resulting in uncertainty and incompleteness of audit results.
[0003] The informatization of audit objects makes computer-assisted auditing inevitable, while common methods such as data query sampling and statistical analysis are mostly computerized manual audit processes, which do not make full use of advanced information technology and cannot extract hidden or unknown information. With the increase in the amount of business data of enterprises, the audited data is growing massively, and a large database of TB or even PB level has been established. With the huge amount of audit data, it is difficult to give full play to the advantages of big data by relying solely on prior knowledge and traditional experience processes. Therefore, exploring new methods suitable for audit big data to extract audit evidence has important theoretical and application value.
[0004] Therefore, it is particularly important to develop an audit system that can automatically collect, process, match and display audit data, and intelligently generate audit opinions and rectification suggestions. Summary of the invention
[0005] The purpose of the present invention is to provide an audit system and method for intelligent data matching and intelligent analysis. By setting audit field collection rules, performing data fuzzy matching and similarity detection, and building a deep learning model for intelligent analysis, it can realize automatic processing of audit data and accurate identification of violations, and finally generate detailed audit opinions and rectification suggestions.
[0006] The technical solution of the present invention is as follows:
[0007] An audit method for intelligent data matching and intelligent analysis, characterized in that it includes the following steps:
[0008] S1: Audit data import: According to the audit requirements, set the collection rules of the audit fields, formulate specific collection standards for each audit field, select the appropriate data collection tool, and extract the audited data from the data source details according to the audit fields;
[0009] S2: Audit data processing:
[0010] Select audit fields, perform fuzzy matching, analyze the audited data, set field similarity detection algorithms, and perform comparisons between audit fields in the data table;
[0011] Traverse all records in the audited data, compare the audit fields, calculate the data similarity based on the field similarity detection algorithm, and compare the calculated data similarity with the designed threshold. If the similarity exceeds the threshold, it is suspicious data;
[0012] Field similarity detection algorithms include:
[0013] (1) For character fields, a field can be regarded as a string. String similarity detection is also called string matching. Using the edit distance algorithm, the edit distance between two fields can be calculated and converted into the similarity between the fields. The edit distance is obtained by calculating the longest common subsequence. The formula is:
[0014] d(x,y)=L x +L y -2L
[0015] Where L x and L y Indicates the length of strings x and y, x belongs to the audited data, y belongs to the audit field, L represents the length of the longest common subsequence, and is converted into similarity through the following formula:
[0016]
[0017] The larger the edit distance, the less similar the fields are.
[0018] (2) For Boolean fields, first use the equation to directly determine whether the two fields are equal. If the two fields are equal, the similarity is 1; if they are different, the similarity is 0.
[0019] S3: Audit data matching:
[0020] Establish audit fields and corresponding threshold sets; establish violation basis sets in accordance with qualitative regulations; establish corresponding audit opinion sets based on violation basis sets; establish rectification opinion sets based on violation basis sets and audit opinion sets; determine the association sets of audit fields and corresponding thresholds, violation basis sets, audit opinions, and rectification opinions;
[0021] Build a database deep learning model and train the model on suspicious data so that it can learn the violation logic from the suspicious data; ensure that the relationship between audit fields, thresholds, violation basis, audit opinions and rectification opinions can be clearly expressed; automatically match suspicious data to the corresponding violation basis, audit opinions and rectification opinions;
[0022] The suspicious data, violation basis, audit opinions, and rectification opinions are stored in the suspicious data sub-database, violation basis sub-database, audit opinion sub-database, and rectification suggestion sub-database respectively;
[0023] S4: Audit data display: the corresponding suspicious data, violation basis, audit opinion, and rectification opinion are sorted in order according to the similarity score; a list of the best violation basis, audit opinion, and rectification suggestion is provided, along with other violation basis, audit opinion, and rectification suggestion that can be selected;
[0024] S5: Drafting of rectification work: When manually selected, a visual window is generated and corresponding rectification opinions are written for auditors’ reference.
[0025] In step S1, the audit fields described are: identifying data fields that are critical to the audit, such as transaction date, transaction amount, transaction type, participant information, and approval process status; establishing specific collection standards for each key field, including the format, unit, and whether the data is required, to ensure the accuracy and consistency of the data.
[0026] In step S1, the data source: clearly identify the systems or databases from which the data will come, such as ERP system, CRM system, financial system, log files; the data collection tool: select appropriate data collection tools according to the characteristics of the data source and audit requirements, such as ETL (Extract, Transform, Load) tools, API calls, and database queries.
[0027] In step S2, it is determined how to handle empty values, special characters, and non-standard format issues to ensure the accuracy and robustness of the matching process.
[0028] In step S3, the suspicious data is clustered, and appropriate clustering algorithms such as K-means and DBSCAN are selected to discover patterns and trends in the data through similarity and difference analysis; auditors use cluster analysis to identify transactions or data points with similar characteristics, thereby discovering potential anomalies or risks.
[0029] An audit system for intelligent data matching and smart analysis is a system for implementing the above method.
[0030] First, according to the specific needs of the audit, the audit fields that need to be collected are clearly identified, and detailed collection standards are formulated for each field. Then, appropriate data collection tools are selected. These tools may include database query tools, API interface call tools, etc., which are used to extract data from various data sources (such as corporate financial systems, business system databases, etc.) according to the set audit fields to form the audit data set.
[0031] In the data processing stage, key audit fields are selected for fuzzy matching to deal with the problem of incomplete consistency in the data. At the same time, field similarity detection algorithms are set, including the edit distance algorithm for character fields and the direct comparison algorithm for Boolean fields. By traversing all records in the audited data set, these algorithms are used to calculate the similarity between fields, and the calculation results are compared with the designed threshold. When the similarity exceeds the threshold, the data record is considered suspicious data, which may contain violations or abnormalities.
[0032] Establish a set of audit fields and corresponding thresholds for direct use in subsequent analysis. At the same time, in accordance with qualitative regulations (such as accounting standards, industry norms, etc.), establish a violation basis set to clarify which behaviors or data states constitute violations. According to the violation basis set, further establish the corresponding audit opinion set and rectification opinion set to provide a basis for subsequent audit conclusions and rectification measures. Subsequently, build a database deep learning model and use the model to train suspicious data so that it can automatically learn the violation logic, that is, identify which data patterns or features correspond to violations. Finally, ensure that the relationship between audit fields, thresholds, violation basis, audit opinions, and rectification opinions can be clearly expressed, and automatically match suspicious data to the corresponding violation basis, audit opinions, and rectification opinions. The matching results will be stored in their respective sub-databases for subsequent query and management.
[0033] During the data display phase, suspicious data and their corresponding violation bases, audit opinions, and rectification opinions are sorted according to the similarity score. In this way, auditors can give priority to those data records with the highest similarity and the most likely violations. At the same time, the system provides a list of the best violation bases, audit opinions, and rectification suggestions, with other options available, so that auditors can choose or adjust according to the actual situation.
[0034] When the auditor selects the corresponding violation basis and audit opinion based on the audit results, the system generates a visual window, allowing the auditor to enter or edit specific rectification opinions. These rectification opinions will serve as guidance and basis for subsequent rectification work, helping the audited unit to correct violations in a timely manner and improve management processes.
[0035] The advantages of the present invention are as follows:
[0036] 1. Improve audit efficiency
[0037] By setting the collection rules and data collection tools for audit fields, the automatic import of audit data is achieved, which reduces manual intervention and improves the efficiency of data collection.
[0038] By using field similarity detection algorithms and deep learning models, audit data can be automatically processed and intelligently analyzed to quickly identify suspicious data and violations, greatly shortening the audit cycle.
[0039] 2. Enhance audit accuracy
[0040] Through fuzzy matching and similarity detection algorithms, subtle differences and potential problems in the data can be accurately identified, improving the accuracy of the audit.
[0041] The deep learning model can learn violation logic from suspicious data, continuously optimize matching rules, and improve recognition accuracy and reliability.
[0042] 3. Optimize audit experience
[0043] Providing visual windows and sorting functions enables auditors to view audit results and rectification suggestions intuitively and conveniently, improving the user experience of audit work.
[0044] It comes with optional violation evidence, audit opinions and rectification suggestions, allowing auditors to make choices and adjustments based on actual conditions, increasing the flexibility and pertinence of the audit work.
[0045] 4. Promote compliance management
[0046] Through automated data processing and intelligent analysis, real-time monitoring and early warning of audit objects can be achieved, which helps companies to detect and correct violations in a timely manner.
[0047] The rectification suggestions generated based on big data and intelligent algorithms provide enterprises with a more scientific and reasonable compliance management solution, which helps to improve the compliance management level of enterprises. DETAILED DESCRIPTION
[0048] Example 1
[0049] An audit method for data intelligent matching and intelligent analysis includes the following steps:
[0050] S1: Audit data import: According to the audit requirements, set the collection rules of the audit fields, formulate specific collection standards for each audit field, select the appropriate data collection tool, and extract the audited data from the data source details according to the audit fields;
[0051] The audit fields described above: identify data fields that are critical to the audit, such as transaction date, transaction amount, transaction type, participant information, and approval process status; formulate specific collection standards for each key field: including the format, unit, and whether the data is required, to ensure the accuracy and consistency of the data.
[0052] The data source mentioned above: clarify which systems or databases the data will come from, such as ERP system, CRM system, financial system, log files; the data collection tool mentioned above: select appropriate data collection tools according to the characteristics of the data source and audit requirements, such as ETL (Extract, Transform, Load) tools, API calls, and database queries.
[0053] First of all, auditors need to fully understand the specific objectives and requirements of the audit and clarify the scope and focus of the audit. This is the basis for formulating subsequent audit field collection rules and data collection strategies.
[0054] Audit Field Identification:
[0055] Based on the audit objectives, identify the data fields that are critical to the audit. These fields may include but are not limited to transaction date, transaction amount, transaction type, participant information (such as customer name, supplier name, etc.), approval process status, etc.
[0056] The selection of these fields should be based on the specific needs and business logic of the audit to ensure that the business activities of the audited entity can be fully and accurately reflected.
[0057] Collection standard formulation:
[0058] Establish specific collection standards for each key field, including the data format (such as date format, number format, etc.), unit (such as currency unit, measurement unit, etc.), whether it is required, etc.
[0059] The formulation of these standards should be based on the standardization and consistency requirements of data to ensure that the collected data can meet the needs of audit analysis.
[0060] Clear data source
[0061] Identify the systems or databases from which the data will come. These data sources may include the company's ERP system (enterprise resource planning system), CRM system (customer relationship management system), financial system, and log files.
[0062] Understand the data structure, storage method, access rights and other information of each data source to provide a basis for the selection and configuration of subsequent data collection tools.
[0063] Data collection tool selection and application
[0064] Data collection tool selection:
[0065] Select appropriate data collection tools based on the characteristics of the data source and audit requirements. These tools may include ETL (Extract, Transform, Load) tools, API call interfaces, database query tools, etc.
[0066] ETL tools are suitable for scenarios where data needs to be extracted, transformed, and loaded from multiple data sources into a target database; API call interfaces are suitable for situations where data sources provide standard API interfaces; and database query tools are suitable for scenarios where databases need to be accessed directly for querying.
[0067] Data collection applications:
[0068] According to the selected data collection tool, configure the corresponding parameters and scripts to achieve the purpose of extracting data from the data source details according to the audit field.
[0069] During the data collection process, it is necessary to ensure the integrity and accuracy of the data to avoid data loss or errors.
[0070] The collected data will be organized into a format that meets audit requirements and stored in the audited database for subsequent audit analysis and processing.
[0071] S2: Audit data processing:
[0072] Select audit fields, perform fuzzy matching, analyze the audited data, set field similarity detection algorithms, and perform comparisons between audit fields in the data table; set how to handle null values, special characters, and non-standard format issues to ensure the accuracy and robustness of the matching process.
[0073] Traverse all records in the audited data, compare the audit fields, calculate the data similarity based on the field similarity detection algorithm, and compare the calculated data similarity with the designed threshold. If the similarity exceeds the threshold, it is suspicious data;
[0074] Field similarity detection algorithms include:
[0075] (1) For character fields, a field can be regarded as a string. String similarity detection is also called string matching. Using the edit distance algorithm, the edit distance between two fields can be calculated and converted into the similarity between the fields. The edit distance is obtained by calculating the longest common subsequence. The formula is:
[0076] d(x,y)=L x +L y -2L
[0077] Where L x and L y Indicates the length of strings x and y, x belongs to the audited data, y belongs to the audit field, L represents the length of the longest common subsequence, and is converted into similarity through the following formula:
[0078]
[0079] The larger the edit distance, the less similar the fields are.
[0080] (2) For Boolean fields, first use the equation to directly determine whether the two fields are equal. If the two fields are equal, the similarity is 1; if they are different, the similarity is 0.
[0081] In the process of audit data processing, we first need to select audit fields that are closely related to the audit objectives. These fields usually include key business data such as transaction date, transaction amount, transaction type, participant information, approval process status, etc. By accurately selecting audit fields, we can ensure the pertinence and effectiveness of the audit work.
[0082] Fuzzy matching is a method of judging the similarity between two or more objects through certain rules or algorithms without a complete match. In audit data processing, fuzzy matching is mainly used to handle issues such as null values, special characters, and non-standard formats in audit fields to ensure the accuracy and robustness of the matching process.
[0083] For different types of audit fields, different field similarity detection algorithms are needed to calculate data similarity.
[0084] For character fields:
[0085] A character field can be considered as a string, so string similarity detection is string matching.
[0086] An edit distance algorithm (such as Levenshtein distance) is used to calculate the edit distance between two fields. The edit distance represents the minimum number of operations (including insertion, deletion, replacement of characters, etc.) required to convert one string into another.
[0087] The edit distance can be converted into the similarity between fields, which is usually calculated by subtracting the ratio of the edit distance and the string length from 1. The closer the similarity is to 1, the more similar the two fields are.
[0088] For Boolean fields:
[0089] Boolean fields have only two values: True or False.
[0090] If two Boolean fields are equal (both true or both false), the similarity is 1; if they are different, the similarity is 0.
[0091] For the suspicious data identified, auditors need to conduct further analysis and investigation. This may include reviewing the original data, tracking business processes, verifying the operational behavior of relevant personnel, etc. Based on the investigation results, auditors can make corresponding rectification suggestions or take other necessary measures.
[0092] S3: Audit data matching:
[0093] Establish audit fields and corresponding threshold sets; establish violation basis sets in accordance with qualitative regulations; establish corresponding audit opinion sets based on violation basis sets; establish rectification opinion sets based on violation basis sets and audit opinion sets; determine the association sets of audit fields and corresponding thresholds, violation basis sets, audit opinions, and rectification opinions;
[0094] Cluster suspicious data, select appropriate clustering algorithms such as K-means and DBSCAN, and discover patterns and trends in the data through similarity and difference analysis; auditors use cluster analysis to identify transactions or data points with similar characteristics, thereby discovering potential anomalies or risks.
[0095] Build a database deep learning model and train the model on suspicious data so that it can learn the violation logic from the suspicious data; ensure that the relationship between audit fields, thresholds, violation basis, audit opinions and rectification opinions can be clearly expressed; automatically match suspicious data to the corresponding violation basis, audit opinions and rectification opinions;
[0096] The suspicious data, violation basis, audit opinions, and rectification opinions are stored in the suspicious data sub-database, violation basis sub-database, audit opinion sub-database, and rectification suggestion sub-database respectively;
[0097] The working principle of the entire audit data processing and analysis system can be summarized as follows: first, establish audit fields and corresponding threshold sets to provide a basis for data processing and analysis; then formulate violation basis sets, audit opinion sets, and rectification opinion sets to provide standards for judging whether the data is in violation; then, through the application of cluster analysis and deep learning models, automatically identify anomalies or risks in the data; finally, store relevant data in different sub-databases, and implement cross-database related queries so that auditors can obtain comprehensive audit information. This system has greatly improved the efficiency and accuracy of audit work and provided strong support for corporate risk management.
[0098] S4: Audit data display: the corresponding suspicious data, violation basis, audit opinion, and rectification opinion are sorted in order according to the similarity score; a list of the best violation basis, audit opinion, and rectification suggestion is provided, along with other violation basis, audit opinion, and rectification suggestion that can be selected;
[0099] According to the similarity score results, suspicious data, violation bases, audit opinions and rectification suggestions are sorted. The purpose of sorting is to determine the priority of processing and give priority to the most important or urgent issues. Based on the sorting results and the professional judgment of the auditors, an optimal list of violation bases, audit opinions and rectification suggestions is generated. At the same time, other optional violation bases, audit opinions and rectification suggestions will also be provided so that adjustments and optimizations can be made according to actual conditions.
[0100] S5: Drafting of rectification work: When manually selected, a visual window is generated and corresponding rectification opinions are written for auditors’ reference.
[0101] The similarity scoring and sorting process of audit data display, as well as the generation of the optimal and alternative recommendation lists, is an important part of the audit work. This process ensures the accuracy and effectiveness of the audit work through scientific methods and professional judgment, provides targeted improvement suggestions for the audited units, and promotes their standardized operation and healthy development.
[0102] Example 2
[0103] The audit method for intelligent data matching and intelligent analysis, based on Example 1, also includes the following contents:
[0104] S1: Audit result management: Newly generated audit opinions and rectification suggestions are manually confirmed and then returned to the database for storage.
[0105] Data confidentiality management: The entire database is always encrypted. Each piece of data is encrypted and stored.
[0106] The newly generated audit opinions and rectification suggestions are first carefully confirmed by humans to ensure their accuracy and applicability. After confirmation, these opinions and suggestions will be returned to the database for subsequent query, analysis and tracking. This step is a key link in the audit process, which ensures the authority and reliability of the audit results.
[0107] The overall database is processed using encryption technology to ensure the confidentiality and security of the data. Each piece of data is encrypted when stored, so that even if the database is illegally accessed, unauthorized personnel cannot directly read the data content. This encryption mechanism effectively prevents data leakage and abuse, and protects the legitimate rights and interests of the audited unit and related parties.
[0108] S2: Audit data export, the export content is:
[0109] Index problem details: index the entire problem details from four aspects: factual statement, violation basis, audit opinion, and rectification suggestions;
[0110] Select data to export as a table: After finding the required data, you can select one or more data and export them as a detailed list of audit findings;
[0111] Select data to export as files: After finding the required data, you can select one or more data to produce audit record files and audit working papers.
[0112] The audit data export function allows users to index the problem details from four aspects: factual statement, violation basis, audit opinion, and rectification suggestion. Users can quickly locate specific audit issues through these indexes, which is convenient for subsequent analysis and processing. The establishment of indexes improves the efficiency and accuracy of data retrieval.
[0113] Users can select one or more data as needed and export them as a detailed list of audit findings. This tabular export method makes it easier for users to organize and analyze data, and also makes it easier to share and communicate audit results with other personnel.
[0114] In addition to exporting to tables, users can also choose to export audit data as audit record files and audit working papers. These files contain important information and evidence in the audit process and are of great significance for subsequent audit review and file management. The file export function provides users with a flexible way to process data.
[0115] S3: Manual personalized modification entry: If there are discrepancies or differences between the violation basis, audit opinions, rectification suggestions and factual descriptions derived from suspicious data in certain audit fields or a certain category of audit fields and the issues related to the opinions, a manual personalized modification list will be provided, which will also be used to train the database deep learning model.
[0116] In some cases, the violation basis, audit opinions, and rectification suggestions automatically generated by the system may be inconsistent or different from the actual situation. To solve this problem, the system provides a manual personalized modification entry. Users can make personalized modifications to the violation basis, audit opinions, and rectification suggestions derived from suspicious data based on actual conditions, and generate a modification list. These modifications not only solve specific problems, but also provide valuable training data for the database deep learning model. Through continuous learning and optimization, the deep learning model can gradually improve the accuracy and applicability of the automatically generated violation basis, audit opinions, and rectification suggestions.
Claims
1. An audit method for intelligent data matching and intelligent analysis, characterized in that: The following steps are involved: S1: Audit data import: According to the audit requirements, set the collection rules of the audit fields, formulate specific collection standards for each audit field, select the appropriate data collection tool, and extract the audited data from the data source details according to the audit fields; S2: Audit data processing: Select audit fields for fuzzy matching, analyze the audited data, set the field similarity detection algorithm, and perform comparisons between audit fields in the data table; Traverse all records in the audited data, compare the audit fields, and finally calculate the data similarity based on the field similarity detection algorithm. Compare the calculated data similarity with the designed threshold. If the similarity exceeds the threshold, it is suspicious data. Field similarity detection algorithms include: (1) For character fields, a field can be regarded as a string. String similarity detection is also called string matching. The edit distance algorithm is used to calculate the edit distance between two fields and convert it into the similarity between the fields. The edit distance is obtained by calculating the longest common subsequence. The formula is: d(x,y)=L x +L y -2L Where L x and L y Represents the length of strings x and y, x belongs to the audited data, y belongs to the audit field, L represents the length of the longest common subsequence, and is converted into similarity through the following formula: similarity = The larger the edit distance, the less similar the fields are. (2) For Boolean fields, first use the equation to directly determine whether the two fields are equal. If the two fields are equal, the similarity is 1; if they are different, the similarity is 0. S3: Audit data matching: Establish audit fields and corresponding threshold sets; establish violation basis sets in accordance with qualitative regulations; establish corresponding audit opinion sets based on violation basis sets; establish rectification opinion sets based on violation basis sets and audit opinion sets; determine the association sets of audit fields and corresponding thresholds, violation basis sets, audit opinions, and rectification opinions; Build a database deep learning model and train the model on suspicious data so that it can learn the violation logic from the suspicious data; ensure that the relationship between audit fields, thresholds, violation basis, audit opinions and rectification opinions can be clearly expressed; automatically match suspicious data to the corresponding violation basis, audit opinions and rectification opinions; The suspicious data, violation basis, audit opinions, and rectification opinions are stored in the suspicious data sub-database, violation basis sub-database, audit opinion sub-database, and rectification suggestion sub-database respectively; S4: Audit data display: the corresponding suspicious data, violation basis, audit opinion, and rectification opinion are sorted in order according to the similarity score; a list of the best violation basis, audit opinion, and rectification suggestion is provided, along with other violation basis, audit opinion, and rectification suggestion that can be selected; S5: Drafting of rectification work: When manually selected, a visual window is generated and corresponding rectification opinions are written for auditors’ reference.
2. According to claim 1, the audit method for intelligent data matching and intelligent analysis is characterized in that: In step S1, the audit fields described are: identifying data fields that are critical to the audit, such as transaction date, transaction amount, transaction type, participant information, and approval process status; establishing specific collection standards for each key field, including the format, unit, and whether the data is required, to ensure the accuracy and consistency of the data.
3. According to claim 1, the audit method for intelligent data matching and intelligent analysis is characterized in that: In step S1, the data source: clearly identify the systems or databases from which the data will come, such as ERP system, CRM system, financial system, log files; the data collection tool: select appropriate data collection tools according to the characteristics of the data source and audit requirements, such as ETL (Extract, Transform, Load) tools, API calls, and database queries.
4. According to claim 1, the audit method for intelligent data matching and intelligent analysis is characterized in that: In step S2, it is determined how to handle empty values, special characters, and non-standard format issues to ensure the accuracy and robustness of the matching process.
5. According to claim 1, the audit method for intelligent data matching and intelligent analysis is characterized in that: In step S3, the suspicious data is clustered and the K-means clustering algorithm is selected to discover patterns and trends in the data through similarity and difference analysis. Auditors use cluster analysis to identify transactions or data points with similar characteristics to discover potential anomalies or risks.
6. The audit method for intelligent data matching and intelligent analysis according to claim 1 is characterized in that: It is characterized in that it also includes the following steps: Audit result management: Newly generated audit opinions and rectification suggestions are manually confirmed and then returned to the database for storage.
7. The audit method for intelligent data matching and intelligent analysis according to claim 1 is characterized in that: The following steps are also included: Data confidentiality management: The entire database is always encrypted. Each piece of data is encrypted and stored.
8. The audit method for intelligent data matching and intelligent analysis according to claim 1 is characterized in that: The following steps are also included: Audit data export, the export content is: Index problem details: index the entire problem details from four aspects: factual statement, violation basis, audit opinion, and rectification suggestions; Select data to export as a table: After finding the required data, you can select one or more data and export them as a detailed list of audit findings; Select data to export as files: After finding the required data, you can select one or more data to produce audit record files and audit working papers.
9. The audit method for intelligent data matching and intelligent analysis according to claim 1 is characterized in that: The following steps are also included: Manual personalized modification entry: If there are discrepancies or differences between the violation basis, audit opinions, rectification suggestions and factual descriptions derived from suspicious data in certain audit fields or a certain category of audit fields and the issues related to the opinions, a manual personalized modification list will be provided, which will also be used to train the database deep learning model.
10. An audit system for intelligent data matching and intelligent analysis, characterized in that: A system for implementing any of the methods of claims 1-5.