Auditing data analysis method and device based on artificial intelligence

Abstract

The invention discloses an audit data analysis method and device based on artificial intelligence. The method comprises the steps that an audit analysis tool selection model of the audit analysis robot carries out intention recognition on an input audit analysis requirement, a triple used for representing the recognized intention is generated, and a corresponding audit analysis tool is screened and extracted from an audit analysis tool library on the basis of the triple; and based on audit data acquired by an audit data robot, the enterprise financial statement analytical review model of the audit analysis robot analyzes the audit data by using the audit analysis tool. The technical problem of low auditing efficiency in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to an audit data analysis method and device based on artificial intelligence. Background Art

[0002] Traditional audit analysis methods often lack intelligent support, and auditors need to spend a lot of time to identify and understand analysis needs and select appropriate analysis tools. This manual selection process is not only susceptible to the subjective judgment of auditors, but also often fails to provide timely and accurate responses when faced with complex and changing audit needs. In addition, the tool library in the existing technology often lacks systematic management and cannot quickly screen and extract suitable analysis tools according to actual needs, which in turn affects the quality and reliability of audit results. Therefore, the existing audit analysis method urgently needs a more efficient and intelligent solution to improve the overall efficiency and accuracy of audit work.

[0003] To address the above-mentioned problems, no effective solution has been proposed yet. Summary of the invention

[0004] The embodiments of the present invention provide an audit data analysis method and device based on artificial intelligence to at least solve the technical problem of low audit efficiency in the prior art.

[0005] According to one aspect of an embodiment of the present invention, there is provided an audit data analysis method based on artificial intelligence, comprising: the audit analysis tool selection model of the audit analysis robot identifies the intent of an input audit analysis requirement, generates a triplet for representing the identified intent, and screens and extracts corresponding audit analysis tools from an audit analysis tool library based on the triplet; based on the audit data acquired by the audit data robot, the enterprise financial statement analytical review model of the audit analysis robot uses the audit analysis tool to analyze the audit data.

[0006] According to another aspect of an embodiment of the present invention, there is also provided an audit data analysis device based on artificial intelligence, namely, an audit analysis robot, comprising: an audit analysis tool selection model, configured to identify the intent of an input audit analysis requirement, generate a triplet for representing the identified intent, and screen and extract corresponding audit analysis tools from an audit analysis tool library based on the triplet; and an enterprise financial statement analytical review model, configured to analyze the audit data obtained by the audit data robot using the audit analysis tool.

[0007] In the embodiment of the present invention, the audit analysis tool selection model of the audit analysis robot recognizes the intent of the input audit analysis requirements, generates a triplet for representing the recognized intent, and screens and extracts the corresponding audit analysis tools from the audit analysis tool library based on the triplet; based on the audit data obtained by the audit data robot, the enterprise financial statement analytical review model of the audit analysis robot uses the audit analysis tool to analyze the audit data. Through the above solution, the technical problem of low audit efficiency in the prior art is solved. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0009] Figure 1 is a flow chart of an audit data analysis method based on artificial intelligence according to an embodiment of the present invention;

[0010] Figure 2 is an architectural diagram of an audit analysis robot according to an embodiment of the present invention;

[0011] Figure 3 is a flow chart of another audit data analysis method based on artificial intelligence according to an embodiment of the present invention;

[0012] Figure 4 is a flow chart of another audit data analysis method based on artificial intelligence according to an embodiment of the present invention;

[0013] Figure 5 is a flow chart of a method for auditing corporate financial statements based on a large language model according to an embodiment of the present invention;

[0014] Figure 6 A schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the scheme of the present invention, the technical scheme in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present invention.

[0016] It should be noted that the terms "first", "second", etc. in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] The present application embodiment provides an audit data analysis method based on artificial intelligence, which is applied to an audit analysis robot, such as Figure 1 As shown, the method includes:

[0018] Step S102, the audit analysis tool selection model of the audit analysis robot performs intent recognition on the input audit analysis requirements, generates a triplet for representing the recognized intent, and screens and extracts corresponding audit analysis tools from the audit analysis tool library based on the triplet;

[0019] The triad consists of three tools.

[0020] The first tool is the method tool, which is used to indicate the method of data description and ensure that the audit data is presented in the most suitable form so that users can understand and analyze it more effectively. Specifically, the method tools include text description method tools, table description method tools and visual description method tools. The text description method tool generates clear and concise text descriptions through natural language processing to provide users with in-depth understanding and background information; the table description method tool displays audit data in a structured table format, allowing users to quickly review and compare data and facilitate the identification of trends and anomalies; the visual description method tool uses visual elements such as charts and dashboards to intuitively display complex data, helping users quickly identify patterns and relationships, thereby improving analysis efficiency and accuracy.

[0021] The second tool is the purpose tool, which indicates the purpose of the audit analysis needs and clarifies the specific goals that users want to achieve through analysis. The purpose tool helps prioritize audit tasks and ensures that the selected analysis tools and methods are highly consistent with the audit objectives. For example, it can help the audit analysis robot focus on improving financial transparency, identifying potential financial risks, or ensuring compliance, thereby better meeting the needs of users.

[0022] The third tool is the account or indicator tool, which indicates the account and ratio analysis indicators related to audit analysis. The third tool ensures that the selected tool can effectively analyze specific financial data and provides a clear framework for identifying and classifying various accounts in financial statements, such as balance sheet items and income statement items, as well as related financial ratios (such as current ratios, asset turnover, etc.). The use of account or indicator tools can help the audit analysis robot to more systematically evaluate financial health and support multi-dimensional analysis, thereby providing users with deep insights.

[0023] Through the comprehensive application of these three tools, the audit analysis robot can flexibly adapt to different audit needs and achieve efficient data analysis and decision support. This structured approach ensures the transparency and effectiveness of the audit analysis process, provides users with more targeted analysis results, and helps companies make wise decisions in complex financial environments.

[0024] Step S104: Based on the audit data acquired by the audit data robot, the enterprise financial statement analytical review model of the audit analysis robot analyzes the audit data using the audit analysis tool.

[0025] First, using the subject or indicator tool in the audit analysis tool, the audit data is calculated and summarized according to each indicator and the preset function calculation method to obtain subject or indicator data, wherein the subject or indicator data includes statistical data of the subject or the indicator and a monthly trend table.

[0026] Then, using the method tool in the audit analysis tool, the subject or indicator data is displayed in the method indicated by the method tool;

[0027] Finally, the large language model is used to perform intelligent analysis on the subject or indicator data according to historical audit experience, and the corresponding analysis results are output. For example, the subject or indicator data is placed in a prompt template by using a prompt template, and the prompt template is input into a deep learning model, wherein the prompt template includes the historical audit experience; the deep learning model performs intelligent analysis on the subject or indicator data based on the content in the prompt template. The prompt template describes the historical audit experience from at least one of the three aspects of abnormal trend, source analysis, and critical value analysis.

[0028] This application introduces artificial intelligence algorithms, so that audit analysis is not limited to traditional numbers and tables, but can understand and process complex natural language data, thereby providing more comprehensive and accurate audit results. The deep integration provided by this application not only improves the intelligence level of audit work, but also provides auditors with more powerful tools to help them quickly obtain the required information when faced with large and complex financial data.

[0029] At the same time, the application of large language models in the field of corporate financial statement auditing can effectively make up for some of the shortcomings of current traditional auditing methods. Traditional auditing often faces high time and labor costs, and there is room for improvement in intelligent applications. By introducing automated data processing and intelligent analysis tools, the manpower input in auditing work can be significantly reduced and efficiency can be improved.

[0030] The following will describe in detail the method by which the audit analysis robot performs audit data analysis, taking profit and loss accounts as an example.

[0031] The audit process of corporate financial statements usually lacks the level of intelligent analysis of audit results. To address this situation, this application will use the design of an audit analysis robot to improve the level of intelligent analysis in the audit process of corporate financial statements. The audit analysis robot will extract relevant financial indicator data from financial statements, and use these indicator data to detect abnormal situations of these financial indicators through ratio analysis and trend analysis. Users can also choose the visualization method of the results of financial indicator ratio analysis and trend analysis based on their own needs, use the analysis capabilities of the large language model to achieve automated and intelligent analysis, and output relevant analysis results.

[0032] The specific framework of the audit analysis robot is as follows: Figure 2 As shown in the figure, the audit analysis robot mainly includes two models. One is the construction of the audit analysis tool selection model, which mainly screens and extracts appropriate and corresponding audit analysis tools from the audit analysis tool library based on the audit analysis requirements put forward by the auditors according to the intention recognition. The other is the construction of the audit financial statement analytical review model, which uses the selected audit analysis tool to obtain the data of relevant subjects through the audit data acquisition model, and outputs some analytical data results, visualization results or text analysis results through the specified audit analysis tool.

[0033] In this embodiment, the LangChain framework is also used to connect to the API of GPT-4 to realize the call of the large language model, and the large language model is used to realize the implementation of the audit analysis tool selection model and the acquisition of relevant audit data. When designing the analytical review model of the enterprise financial statement, the data is processed and calculated in the form of constructing the data processing and analysis function as described above, and finally the data visualization is realized by calling the function generated by the relevant chart. The intelligent analysis uses the text analysis and generation capabilities of the large language model itself, and improves the professional ability of the large language model in intelligent analysis in the form of prompt engineering.

[0034] like Figure 3 As shown, the audit data analysis method based on artificial intelligence includes the following steps:

[0035] Step S302, constructing an audit analysis tool selection model.

[0036] The core of the audit analysis tool selection model is to understand the identification of the analytical review requirements proposed by the auditors. This application will construct a library of relevant subjects and indicators and an audit analysis tool library so that the model can call accurate subject indicators and tools after the intention is identified. When constructing the audit analysis model, the large language model is also used to extract the analytical review requirements proposed by the auditors in the form of "['subject / indicator', 'purpose', 'method']" triple information. Among them, the "subject / indicator" library contains some ratio analysis indicators in addition to the profit and loss related subjects mentioned in the design stage of the audit working paper robot. The profit and loss subjects mainly involve financial indicators related to the income statement. At the same time, the profitability of the enterprise can also be obtained from the analysis of relevant financial indicators in the income statement. Profitability is usually measured by gross profit margin, operating profit margin, return on total assets, return on net assets and net profit margin

[60] . At the same time, in terms of "method", the results of ratio trend analysis and description are realized by text description, table description, and visual description such as line chart, pie chart, bar chart, etc. "Purpose" refers to the purpose of the analysis requirements provided by the auditor, such as the need to analyze data. If the "Purpose" is analysis, the output visual charts will be intelligently analyzed after the results are output.

[0037] The construction method of the audit analysis tool selection model is consistent with that of the audit working paper tool selection model. The GPT-4 API is used to connect with the audit analysis tool library. At the same time, the relevant tools and functions are introduced to the big language model in the form of prompt engineering for the big language model to learn. By performing intent recognition on the audit analysis requirements input by the auditors into the big language model, the "subjects / indicators", "purposes" and "methods" contained in the analysis requirements input by the auditors are identified. By carrying out prompt engineering on the big language model, some examples are used to help the big language model better understand the relevant content, so as to generate triples more accurately. The following Table 1 shows the prompts for the connection and interaction between the audit analysis tool library and the big language model.

[0038] Table 1 Audit Analysis Tool Tips

[0039]

[0040]

[0041] Through the prompts in the prompt project, auditors only need to input the audit analysis requirements they need to build into the model, and then they can select the model through the audit analysis tool to identify its "subject / indicator", "purpose" and meaning, and perform the corresponding analytical review of the company's financial statements based on the content of the triplet.

[0042] Step S304, constructing an analytical review model for corporate financial statements.

[0043] The function of the audit analysis tool model is to analyze and understand the audit analysis requirements proposed by the auditors, and obtain the key contents of the audit analysis requirements proposed by the auditors, that is, the extracted triples. After the triples are extracted, the corresponding tools, subjects or indicators in the triples and whether to analyze are confirmed to realize the call of the corresponding functions to complete the output of the analytical review results of the enterprise financial statements and the intelligent analysis of the results.

[0044] The analytical review model of corporate financial statements will include three modules. First, the previously constructed data acquisition model is used to obtain data on relevant subjects from the database, and the data is calculated and summarized according to the calculation methods of various indicators and functions, and the data of subjects or indicators and monthly trend tables are output. Secondly, the corresponding method is selected according to the needs of auditors, such as making data tables and data visualization to realize the display of analytical review results. Finally, the large language model is used to intelligently analyze the data results based on historical audit experience and output the corresponding analysis results.

[0045] In the data acquisition and data processing module, it is necessary to define the profit and loss account ratio analysis indicators and the indicator calculation method to provide a reference for the model to calculate the corresponding indicators. The profit and loss account ratio analysis indicators mainly include gross profit margin, operating profit margin, total asset return rate, net asset return rate and net sales profit margin. The specific calculation method is shown in Table 2. The "Account / Indicator" library constructs the data acquisition method and data calculation functions of various types of accounts and indicators for easy calling.

[0046] Table 2 Profit and loss account ratio analysis indicators and calculation methods

[0047]

[0048] In the display method selection module, the "Method" library contains a variety of methods such as tables, texts, line graphs, bar graphs, pie charts, etc. to display analysis results. The various types of display methods in the "Method" library are also called by building functions.

[0049] Finally, it is an intelligent analysis of the results of the analytical review of the company's financial statements. This module uses the prompt template to put the monthly trend data generated by each indicator into the prompt template, and then inputs the prompt template to use AI to realize intelligent analysis of the data. When constructing the prompt template, first give the big language model the identity of an audit expert to improve the accuracy of its professional field. Secondly, give the big language model the audit experience of previous auditors when analyzing the indicators. The specific historical audit experience of each indicator is shown in Table 3 below. Finally, let the big language model analyze the data, and the requirements are not limited to the historical experience of auditors, which can prompt the big language model to use its own capabilities to analyze from other aspects. Taking the monthly trend analysis of gross profit margin as an example, according to the above prompt template construction plan, the prompt template is constructed from the aspects of abnormal trend, source analysis and critical value analysis. The specific prompt template is shown in Table 4. The analysis aspects of different subjects or indicators can be set according to the specific analysis content.

[0050] Table 3 Historical audit experience in the prompt template

[0051]

[0052] Table 4 Data result intelligent analysis prompt template (taking gross profit margin as an example)

[0053]

[0054] The audit analysis robot consists of two parts: the audit analysis tool selection model and the enterprise financial statement analytical review model. Auditors can input their analysis requirements to the robot by inputting relevant audit analysis instructions. The audit analysis tool selection model can identify the instructions and generate a triple containing subjects or indicators, purposes and methods. The corresponding three tools are selected based on the contents of the triple to achieve the acquisition, calculation, visualization and intelligent analysis of financial indicator data.

[0055] The present application embodiment provides another audit data analysis method based on artificial intelligence, such as Figure 4 As shown, the method comprises the following steps:

[0056] Step S402, obtaining audit data.

[0057] First, the auditor’s query requirements are converted into an embedding vector, denoted as Vquery. At the same time, each example in the sample library is also converted into a corresponding embedding vector, denoted as V S Next, the cosine similarity between the query embedding vector and the sample embedding vector is calculated to identify the semantically closest examples.

[0058]

[0059] Among them, S represents the sample set of the sample library, Vquery is the embedding vector converted from the query requirements provided by the auditor, and V S It is the embedding vector corresponding to each sample in the sample library.

[0060] The present invention significantly improves the accuracy and efficiency of queries through a similarity calculation method based on embedded vectors. By quickly calculating the cosine similarity between the query embedded vector and the sample embedded vector, auditors can quickly obtain relevant information, thereby optimizing the information retrieval process, reducing dependence on manual intervention, improving the automation level and response speed of audit analysis, and providing auditors with more efficient and accurate data support.

[0061] In order to further improve the accuracy of sample selection, this solution also introduces multi-dimensional similarity calculation and regularization mechanism. When processing multiple dimensions of the embedding vector of each sample, the cosine similarity between the query embedding vector and the sample embedding vector in the jth dimension is first calculated. Then, the comprehensive similarity is obtained by weighted summing of the similarities of these dimensions. On this basis, the regularization term is introduced to prevent the model from overfitting.

[0062] Specifically, the most relevant example can be calculated by the following method. First, the cosine similarity between the query embedding vector and the sample embedding vector in the jth dimension of the m total dimensions is calculated, and the similarities of all m dimensions are weighted and summed to obtain the comprehensive similarity between the embedding vector and the sample. Subsequently, the regularization term is calculated based on the square of the L2 norm of the embedding vector of the sample Si and the regularization coefficient. Finally, the most relevant example is determined based on the comprehensive similarity and the regularization term. Specifically, the calculation formula for the cosine similarity between the query embedding vector and the sample embedding vector in the jth dimension is as follows:

[0063]

[0064] Where Vquery is the query embedding vector, Vsi is the sample embedding vector, Wi is the weight factor of sample Si used to reflect the importance or quality of the sample; λ is the regularization term coefficient used to control the strength of the regularization term; γ j is the nonlinear activation function used to adjust each embedding dimension; σ is the activation function used to control the final example score. m represents the number of dimensions of the embedding vector, and aj is the weight coefficient of the jth similarity calculation dimension.

[0065] By introducing multi-dimensional similarity calculation and regularization mechanism, the present invention significantly improves the accuracy and stability of example selection. The multi-dimensional similarity calculation method enables the model to capture the semantic relationship between samples more comprehensively, further improving the recognition accuracy of related examples. In addition, the introduction of the regularization mechanism effectively prevents the overfitting of the model and ensures the good generalization ability of the model on new data. Finally, combined with the quality and importance of the sample, after the nonlinear activation function processing of the dimension, the obtained example score is more accurate and reliable.

[0066] Step S404, select an audit working paper tool, and use the audit working paper tool to prepare audit working papers.

[0067] First, auditors select the most suitable tools from the audit tool library based on the specific audit needs and the financial subjects involved. Tools are selected based on the type of working papers, such as detailed statements, monthly detailed statements, or income cut-off tests, as well as the required functional features to ensure that the tools can meet specific audit standards and requirements.

[0068] Once the audit working papers tool is selected, auditors can use it to prepare working papers. The tool will help auditors automatically organize, process and present relevant audit data to ensure the accuracy and consistency of the data. In the process of preparing audit working papers, the tool can efficiently integrate audit data from different sources and generate professional working papers according to preset templates and formats. This automated process not only saves time, but also reduces the possibility of human errors, ensuring the reliability and compliance of audit results.

[0069] Ultimately, the generated audit working papers will provide auditors with clear and structured information, facilitating subsequent audit analysis, review and decision-making, and improving the efficiency and effectiveness of the overall audit work. Through such a systematic process, the application of audit working paper tools has significantly optimized audit practice and improved the quality and efficiency of audit work.

[0070] Step S406: the enterprise financial statement analytical review model uses the audit analysis tool to analyze the audit data.

[0071] The specific analysis steps are similar to steps S302-S304 and will not be repeated here.

[0072] After the audit data analysis is completed, the display and visualization of the results are also crucial. Through a variety of presentation methods, auditors can understand the data more intuitively and provide a basis for decision-making.

[0073] The audit analysis model will intelligently select the most suitable visualization tool based on the "method" information in the extracted triples. For example, if the purpose of the analysis is to reveal trends, the model will give priority to line charts; if it is to show the proportional relationship between the parts, the model will choose pie charts. In this way, the model not only considers the initial needs of auditors, but also analyzes the characteristics of the data (such as data volume, time series, etc.) to ensure that the best display method is selected.

[0074] The audit analysis model also supports a variety of display methods, such as tables, line charts, bar charts, and heat maps. For example, heat maps can effectively display sales performance in different regions, while bar charts are suitable for comparing performance in different time periods. The model's visualization module will achieve these displays by calling different graphics libraries (such as Matplotlib, Seaborn, etc.).

[0075] To enhance the user experience, the model supports dynamically updated visualizations. When auditors select different parameters or time ranges in the report, the system can update the display in real time. In addition, the design of the interactive dashboard allows auditors to explore data more easily and gain a deeper understanding of the relationship between various indicators. Finally, the model will automatically generate an audit report by combining the analysis results with the visualization. The report will include a summary of key indicators, trend analysis, anomaly identification, and intelligent analysis suggestions. Through natural language generation technology, the system can convert data and analysis results into easy-to-understand text form, which facilitates communication between auditors and management.

[0076] Through the above methods, audit analysis tools can not only improve the efficiency of audit work, but also enhance auditors' understanding and control over financial data.

[0077] This application also provides an audit data analysis device based on artificial intelligence, such as Figure 5 As shown, it includes: an audit analysis tool selection model 52, which is configured to identify the intent of the input audit analysis requirements, generate a triplet for representing the identified intent, and screen and extract corresponding audit analysis tools from the audit analysis tool library based on the triplet; an enterprise financial statement analytical review model 54, which is configured to analyze the audit data based on the audit data obtained by the audit data robot using the audit analysis tool.

[0078] It should be noted that the audit data analysis device provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. In addition, the audit data analysis device provided in the above embodiment belongs to the same concept as the audit data analysis method embodiment. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0079] Figure 6 FIG. 1 shows a schematic diagram of the structure of an electronic device suitable for implementing the embodiment of the present disclosure. It should be noted that: Figure 6 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0080] like Figure 6As shown, the electronic device includes a central processing unit (CPU) 1001, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 1002 or a program loaded from a storage part 1008 into a random access memory (RAM) 1003. In the RAM 1003, various programs and data required for system operation are also stored. The CPU 1001, the ROM 1002, and the RAM 1003 are connected to each other via a bus 1004. An input / output (I / O) interface 1005 is also connected to the bus 1004.

[0081] The following components are connected to the I / O interface 1005: an input section 1006 including a keyboard, a mouse, etc.; an output section 1007 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 1008 including a hard disk, etc.; and a communication section 1009 including a network interface card such as a LAN card, a modem, etc. The communication section 1009 performs communication processing via a network such as the Internet. A drive 1010 is also connected to the I / O interface 1005 as needed. A removable medium 1011, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 1010 as needed, so that a computer program read therefrom is installed into the storage section 1008 as needed.

[0082] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. An audit data analysis method based on artificial intelligence, characterized in that: include: The audit analysis tool selection model of the audit analysis robot performs intent recognition on the input audit analysis requirements, generates a triplet for representing the recognized intent, and screens and extracts corresponding audit analysis tools from the audit analysis tool library based on the triplet; Based on the audit data acquired by the audit data robot, the enterprise financial statement analytical review model of the audit analysis robot uses the audit analysis tool to analyze the audit data.

2. The method according to claim 1, characterized in that The triplet includes three tools, the first tool of which is a method tool for indicating the data description method, the second tool is a purpose tool for indicating the purpose of the audit analysis requirements, and the third tool is a subject or indicator tool for indicating relevant subjects and ratio analysis indicators.

3. The method according to claim 2, characterized in that The method tools include text description method tools, table description method tools, and visual description method tools.

4. The method according to claim 2, characterized in that: Based on the audit data acquired by the audit data robot, the enterprise financial statement analytical review model of the audit analysis robot uses the audit analysis tool to analyze the audit data, including: Using the subject or indicator tool in the audit analysis tool, the audit data is calculated and summarized according to each indicator and a preset function calculation method to obtain subject or indicator data, wherein the subject or indicator data includes statistical data of the subject or the indicator and a monthly trend table; Using the method tool in the audit analysis tool, displaying the subject or indicator data in the manner indicated by the method tool; Use a large language model to conduct intelligent analysis on the subject or indicator data based on historical audit experience and output the corresponding analysis results.

5. The method according to claim 4, characterized in that Use the big language model to conduct intelligent analysis on the subject or indicator data based on historical audit experience, including: By means of a prompt template, the subject or indicator data is placed into a prompt template, and the prompt template is input into a deep learning model, wherein the prompt template includes the historical audit experience; The deep learning model performs intelligent analysis on the subject or indicator data based on the content in the prompt template.

6. The method according to claim 5, characterized in that The prompt template describes the historical audit experience from at least one of the three aspects of abnormal trend, source analysis and critical value analysis.

7. An audit data analysis device based on artificial intelligence, characterized in that: include: An audit analysis tool selection model is configured to perform intent recognition on an input audit analysis requirement, generate a triplet for representing the recognized intent, and screen and extract a corresponding audit analysis tool from an audit analysis tool library based on the triplet; The enterprise financial statement analytical review model is configured to analyze the audit data based on the audit data obtained by the audit data robot using the audit analysis tool.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored program, wherein when the program is executed, the device where the computer-readable storage medium is located is controlled to execute the method according to any one of claims 1 to 6.

9. A computer device, characterized in that: include: Memory and processor, The memory stores a computer program; The processor is used to execute the computer program stored in the memory, and when the computer program is run, the processor is enabled to execute the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images