A method and apparatus for automatically checking financial operations flash reports

Abstract

This invention discloses a method and apparatus for automatically checking financial operation reports. The method includes: collecting raw financial operation-related data from multiple data sources through a data integration interface; applying a deep learning-based financial semantic understanding model to perform entity recognition and relation extraction on unstructured text data in the raw data, outputting structured financial knowledge triples; calculating various financial indicators based on the structured data and financial knowledge triples, and constructing a financial indicator knowledge graph; applying multiple anomaly detection mechanisms to the financial indicators in parallel; weighting and fusing the initial anomaly scores from each detection mechanism, and outputting a final anomaly determination result based on the fusion result; and automatically generating a financial operation report containing data charts and textual interpretations based on the financial indicators and the final anomaly determination result using natural language generation technology.

Description

Technical Field

[0001] This invention relates to the field of financial automation processing and intelligent analysis technology, and more specifically, to a method and apparatus for automatically checking financial operation reports. Background Technology

[0002] Finance staff are required to compile and review the company's and various departments' operational data weekly, including core financial indicators such as sales, costs, and profits. Currently, this review and compilation is primarily done manually, which presents the following technical challenges: Inefficient: Manual checks consume a lot of time, especially when dealing with financial data from multiple departments and multiple periods, resulting in a huge workload; Insufficient accuracy: Manual inspection cannot guarantee the complete accuracy of the data and is prone to missing errors or anomalies; Single data source: It relies solely on structured data from the financial system and cannot make comprehensive use of unstructured text information such as financial statement notes and management discussions; Limited analytical depth: It only performs basic indicator calculations and lacks in-depth analysis of the relationships between indicators and the ability to trace anomalies; Difficulty in interpreting reports: The generated reports are only data charts and lack professional textual interpretation, requiring decision-makers to spend extra time understanding the meaning of the data.

[0003] Therefore, there is a need for a technical solution that can automatically process multi-source financial data, intelligently identify financial entities and relationships, deeply analyze the correlation of indicators, and automatically generate financial operation reports with interpretation. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method and apparatus for automatically checking financial operation reports.

[0005] According to one aspect of the present invention, a method for automatically checking financial operations reports is provided, comprising: Through the data integration interface, raw data related to financial operations is collected from multiple data sources. The raw data includes at least one of the following: structured data from the financial system, unstructured text data from the notes to the financial statements, and external market data. For unstructured text data in the original data, a deep learning-based financial semantic understanding model is applied to perform entity recognition and relation extraction, and output structured financial knowledge triples. Based on structured data and financial knowledge triples, various financial indicators are calculated, and a financial indicator knowledge graph is constructed. The knowledge graph is used to represent the computational dependencies and business relationships between various financial indicators. Multiple anomaly detection mechanisms are applied to financial indicators in parallel. These mechanisms include at least two of the following: statistical threshold detection, time series prediction detection, peer comparison detection, and historical period comparison detection. The initial anomaly scores of each detection mechanism are then output. The initial anomaly scores of each detection mechanism are weighted and fused, and the final anomaly determination result is output based on the fusion result. Based on financial indicators and the final anomaly determination results, natural language generation technology is applied to automatically generate financial operation reports that include data charts and text interpretations.

[0006] According to another aspect of the present invention, an apparatus for automatically checking financial operation reports is provided, comprising: The data acquisition module is used to collect raw data related to financial operations from multiple data sources through a data integration interface. The raw data includes at least one of the following: structured data from the financial system, unstructured text data from the notes to the financial statements, and external market data. The extraction module is used to perform entity recognition and relation extraction on unstructured text data in the original data using a deep learning-based financial semantic understanding model, and output structured financial knowledge triples. The calculation module is used to calculate various financial indicators based on structured data and financial knowledge triples, and to build a financial indicator knowledge graph. The knowledge graph is used to represent the calculation dependencies and business relationships between various financial indicators. The detection module is used to perform parallel detection of financial indicators using multiple anomaly detection mechanisms. These mechanisms include at least two of the following: statistical threshold detection, time series prediction detection, peer comparison detection, and historical period comparison detection. The module outputs the initial anomaly scores for each detection mechanism. The fusion module is used to weight and fuse the initial anomaly scores of each detection mechanism, and output the final anomaly judgment result based on the fusion result; The generation module is used to automatically generate financial operation reports containing data charts and text interpretations based on financial indicators and final anomaly determination results, using natural language generation technology.

[0007] According to another aspect of the present invention, a computer-readable storage medium is provided, the storage medium storing a computer program for performing the methods described in any of the above aspects of the present invention.

[0008] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: a processor; a memory for storing executable instructions of the processor; the processor being configured to read the executable instructions from the memory and execute the instructions to implement the method described in any of the preceding aspects of the present invention.

[0009] Therefore, compared with the prior art, the present invention has the following beneficial effects: Multi-source data fusion provides a more comprehensive analysis: It not only collects structured data from the financial system, but also automatically processes unstructured text such as financial statement notes through financial semantic understanding technology. At the same time, it introduces external market data, making the analysis more comprehensive and the decision-making basis more sufficient.

[0010] Financial semantic understanding and intelligent text processing: Based on a domain-fine-tuned BERT model, it can accurately identify entities and relationships in financial texts, automatically converting unstructured text into structured knowledge triples, with an accuracy rate more than 30% higher than traditional rule-based methods.

[0011] Knowledge graph construction and in-depth indicator association: Building a knowledge graph of financial indicators not only calculates indicator values, but also establishes dependencies between indicators, supports anomaly tracing, impact analysis and association recommendation, and achieves a qualitative leap from "calculation" to "understanding".

[0012] Multi-dimensional anomaly detection with high accuracy: It integrates four detection mechanisms, namely statistical threshold, time series prediction, peer comparison, and historical comparison, and outputs the final judgment through weighted voting. The anomaly detection accuracy is improved by 40%, and the false alarm rate is reduced by more than 50%.

[0013] Intelligent narrative reports, automated interpretation: Based on NLG technology, financial analysis text interpretation is automatically generated, realizing the integration of "charts + interpretation". Decision-makers do not need to spend extra time understanding the meaning of the data, improving decision-making efficiency by 60%.

[0014] Full-process automation significantly improves efficiency: Configure scheduled tasks to automatically execute the entire process of data collection, indicator calculation, anomaly detection, and report generation, reducing the workload of finance personnel by more than 80% and shortening the time for generating quick reports from several hours to minutes.

[0015] Closed-loop feedback optimization enables continuous system evolution: By recording human feedback results and dynamically adjusting fusion weights and model parameters, the system accuracy continuously improves over time, achieving self-optimization. Attached Figure Description

[0016] Exemplary embodiments of the present invention can be more fully understood by referring to the following figures: Figure 1 This is a flowchart illustrating a method for automatically checking financial operations reports provided in an exemplary embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of an apparatus for automatically checking financial operation reports provided in an exemplary embodiment of the present invention; Figure 3 This is the structure of an electronic device provided in an exemplary embodiment of the present invention. Detailed Implementation

[0017] Hereinafter, exemplary embodiments according to the present invention will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present invention, and not all embodiments of the present invention. It should be understood that the present invention is not limited to the exemplary embodiments described herein.

[0018] It should be noted that, unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of the invention.

[0019] Those skilled in the art will understand that the terms "first," "second," etc., in the embodiments of the present invention are only used to distinguish different steps, devices, or modules, and do not represent any specific technical meaning, nor do they indicate a necessary logical order between them.

[0020] It should also be understood that in the embodiments of the present invention, "multiple" can refer to two or more, and "at least one" can refer to one, two or more.

[0021] It should also be understood that any component, data or structure mentioned in the embodiments of the present invention can generally be understood as one or more unless explicitly defined or given contrary instructions in the context.

[0022] Furthermore, the term "and / or" in this invention is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character " / " in this invention generally indicates that the preceding and following related objects have an "or" relationship.

[0023] It should also be understood that the description of the various embodiments in this invention emphasizes the differences between the various embodiments, and the similarities or similarities can be referred to each other. For the sake of brevity, they will not be described in detail.

[0024] At the same time, it should be understood that, for ease of description, the dimensions of the various parts shown in the accompanying drawings are not drawn according to actual scale.

[0025] The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the invention or its application or use.

[0026] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, they should be considered part of the specification.

[0027] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be discussed further in subsequent figures.

[0028] The embodiments of this invention can be applied to electronic devices such as terminal devices, computer systems, and servers, and can operate together with a wide range of other general-purpose or special-purpose computing system environments or configurations. Well-known examples of terminal devices, computing systems, environments, and / or configurations suitable for use with electronic devices such as terminal devices, computer systems, and servers include, but are not limited to: personal computer systems, server computer systems, thin clients, thick clients, handheld or laptop devices, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputer systems, mainframe computer systems, and distributed cloud computing environments including any of the above systems, etc.

[0029] Electronic devices such as terminal devices, computer systems, and servers can be described in the general context of computer system executable instructions (such as program modules) executed by a computer system. Typically, program modules can include routines, programs, object programs, components, logic, data structures, etc., which perform specific tasks or implement specific abstract data types. Computer systems / servers can be implemented in distributed cloud computing environments, where tasks are executed by remote processing devices linked through communication networks. In distributed cloud computing environments, program modules can reside on local or remote computing system storage media, including storage devices.

[0030] Exemplary methods Figure 1 This is a schematic flowchart of a method provided in an exemplary embodiment of the present invention. This embodiment can be applied to electronic devices, such as... Figure 1 As shown, method 100 includes the following steps: Step 101, Specifically, in view of the above-mentioned deficiencies in the prior art, the technical problems to be solved by the present invention include: 1. How to achieve automated collection and fusion of multi-source heterogeneous financial data; 2. How to perform deep semantic understanding on unstructured text such as notes to financial statements and automatically extract financial entities and relationships; 3. How to build knowledge relationships between financial indicators to support anomaly tracing and impact analysis; 4. How to integrate multiple anomaly detection mechanisms to improve the accuracy and coverage of anomaly identification; 5. How to automatically generate intelligent financial operation reports that include data charts and text interpretations.

[0031] The specific implementation steps include: I. Multi-source data acquisition Through data integration interfaces, raw financial operations-related data are collected from multiple data sources, including: Structured data: Numerical data such as sales revenue, cost, profit, assets, and liabilities collected from financial systems (such as ERP and general ledger systems); Unstructured text data: Textual descriptions collected from notes to financial statements, management discussions and analyses (MD&A), annual reports, etc. External market data: Comparative data and market trend data collected from industry databases, competitor financial reports, etc.

[0032] The collected data, after being standardized in format and aligned with time, is stored in a data warehouse for subsequent analysis.

[0033] II. Financial Semantic Understanding For unstructured text data, a deep learning-based financial semantic understanding model is applied for entity recognition and relation extraction.

[0034] (1) Model Architecture We employ BERT (Bidirectional Encoder Representations from Transformers) or RoBERTa models, which are fine-tuned based on financial domain corpora. These models are pre-trained and fine-tuned on a large number of financial reports, financial statement notes, and other texts, enabling them to accurately understand financial terminology and expressions.

[0035] (2) Entity recognition Identify various entities in financial documents, including: Revenue-generating entities: operating revenue, main business revenue, other business revenue, etc.; Cost entities: operating costs, main business costs, raw material costs, etc.; Expense-related entities: sales expenses, administrative expenses, financial expenses, research and development expenses, etc.; Asset-based entities: current assets, fixed assets, intangible assets, accounts receivable, etc. Liability entities: current liabilities, long-term liabilities, accounts payable, loans, etc.; Equity entities: owners' equity, share capital, capital reserves, retained earnings, etc. Indicators include: net profit, gross profit margin, debt-to-equity ratio, return on equity, etc.

[0036] (3) Relation extraction Identify various relationships between financial entities and output structured financial knowledge triples: Calculation relationships: e.g. (Net profit, calculated from..., operating revenue - operating costs - taxes - period expenses); Year-on-year relationship: e.g. (Operating revenue, year-on-year growth, 15%); Month-on-month comparison: e.g. (net profit, month-on-month growth, 8.3%) Percentage relationship: e.g. (Sales expenses as a percentage of operating revenue, 12%). Attribute relationships: such as (gross profit margin, numerical value, 35.2%).

[0037] Financial knowledge triples are represented as: (Entity 1, Relationship, Entity 2) or (Entity, Attribute, Attribute Value).

[0038] III. Indicator Calculation and Knowledge Graph Construction (1) Calculation of financial indicators Based on structured data and financial knowledge triples, various financial indicators are calculated, including: Profitability Indicators: For example, Gross Profit (G) = S - C, Net Profit (NP) = P - Tax, Profit Margin (ROE) = (P / Equity) * 100, and Debt-to-Equity Ratio (D / E) = Debt / Equity.

[0039] Solvency indicators: debt-to-equity ratio, current ratio; Operational performance indicators: accounts receivable turnover and inventory turnover.

[0040] (2) Construction of financial indicator knowledge graph Construct a knowledge graph representing the computational dependencies and business relationships among financial indicators: Node definition: Various financial indicators are used as nodes in the knowledge graph. Each node contains attributes such as indicator name, value, unit, and timestamp. Edge definition: The computational dependencies between indicators are used as directed edges in the knowledge graph, such as "net profit" depending on "operating revenue", "operating cost", "taxes", "period expenses", etc. Knowledge Graph Construction: Based on standard calculation formulas in the financial field, predefine the dependencies between indicators to construct an initial knowledge graph; as the system runs, continuously extract new relationships from financial knowledge triples to expand the graph.

[0041] (3) Application of knowledge graphs Indicator tracing: When a financial indicator becomes abnormal, reverse tracing is performed using a knowledge graph to identify the underlying indicators that caused the abnormality. Impact Analysis: Through positive propagation analysis using knowledge graphs, the scope and extent of the impact of an anomaly in a certain indicator on higher-level indicators are predicted. Related recommendations: Based on the relationships in the knowledge graph, other indicators related to the current analysis indicator are recommended to assist in comprehensive analysis.

[0042] IV. Multi-dimensional anomaly detection Multiple anomaly detection mechanisms are applied in parallel to detect financial indicators, including: (1) Statistical threshold detection Based on preset threshold ranges, determine whether financial indicators exceed the normal range: Among them, L lower and L upper These are the lower and upper limits set according to business rules.

[0043] (2) Time series prediction detection A time-series forecasting model is trained based on historical data, and the residuals between the predicted and actual values ​​are calculated. Historical time series can be trained using ARIMA models, Prophet models, or LSTM neural networks; The trained model is used to predict the current indicator value to obtain the predicted value; Calculate the historical residual e of the predicted value t ; The dynamic threshold is calculated based on the statistical distribution of historical residuals, when e t > T t When T is determined to be abnormal, t This is a dynamic threshold.

[0044] Dynamic threshold calculation formula: in, The mean of the historical residuals. denoted as the standard deviation of the historical residuals, and k is an adjustable parameter.

[0045] (3) Peer comparison testing Obtain comparable indicator data from companies in the same industry for the same period, and calculate the deviation of the company's indicators from the industry average: When the deviation exceeds the preset threshold, it is judged as abnormal.

[0046] (4) Historical comparison test Compare the indicator data with the same period last year or the previous period, and calculate the year-on-year change rate or month-on-month change rate: When the rate of change exceeds a preset threshold, it is considered abnormal.

[0047] V. Abnormal Fusion Judgment The initial anomaly scores from each detection mechanism are weighted and fused to output the final anomaly determination result.

[0048] (1) Weight allocation Based on the historical accuracy of each anomaly detection mechanism, the fusion weights are dynamically allocated: in, Let be the historical accuracy of the i-th detection mechanism within a recent time period t.

[0049] (2) Weighted fusion The initial anomaly scores of each detection mechanism are multiplied by their corresponding weights and then summed to obtain the comprehensive anomaly score.

[0050] (3) Anomaly detection When the overall anomaly score exceeds the preset fusion threshold T fusion At that time, output the final anomaly determination result: (4) Feedback learning Record the manual feedback results (confirmed anomaly or false alarm) for each anomaly determination, which will be used to adjust the fusion weights and optimize the detection model in the future.

[0051] VI. Intelligent Narrative Report Generation Based on financial indicators and anomaly detection results, a financial operations report is automatically generated, including data charts and textual interpretations.

[0052] (1) Data chart generation Based on a preset report template, generate data charts containing key financial indicators, including: Trend chart: Shows the changing trends of key indicators such as sales, costs, and profits over time; Composition diagram: showing the cost structure, revenue structure, expense structure, etc.; Comparison chart: Showing a comparison between actual values, budgeted values, and values ​​from the same period last year; Dashboard: Displays the current status of key indicators and provides alerts for anomalies.

[0053] (2) Text Interpretation Generation For financial indicators that have changed significantly, a natural language interpretation is automatically generated: Template matching: Based on the direction and magnitude of the indicator's change, it automatically matches a preset interpretation template, such as "This month's operating revenue increased by 8.3% month-on-month, mainly due to the increase in sales of product line A"; Knowledge graph enhancement: For complex and abnormal scenarios, based on the dependencies in the financial indicator knowledge graph, the causes and effects of the anomalies are analyzed, and a more in-depth interpretation is generated; NLG Model Generation: For scenarios requiring personalized interpretation, a natural language generation model based on T5, BART, or GPT architecture is invoked to fine-tune the financial report corpus, converting structured data into fluent natural language descriptions.

[0054] (3) Report integration It integrates data charts and text interpretations into a complete financial operations report, supports multiple output formats such as PDF, Excel, and HTML, and can be automatically pushed via email, WeChat for Business, and other means.

[0055] VII. Automated Scheduling Configure a scheduled task to automatically execute all the above steps according to a preset cycle (such as weekly or monthly) to achieve automated generation of financial operation reports.

[0056] The scheduling system supports: Task orchestration: Define the execution order and dependencies of each step; Error handling: When a step fails, automatically retry or send an alarm; Log recording: Records detailed logs for each execution, supporting problem tracing; Version management: Manage the versions of report templates and models, and support rollback.

[0057] Compared with the prior art, the present invention has the following beneficial effects: Multi-source data fusion provides a more comprehensive analysis: It not only collects structured data from the financial system, but also automatically processes unstructured text such as financial statement notes through financial semantic understanding technology. At the same time, it introduces external market data, making the analysis more comprehensive and the decision-making basis more sufficient.

[0058] Financial semantic understanding and intelligent text processing: Based on a domain-fine-tuned BERT model, it can accurately identify entities and relationships in financial texts, automatically converting unstructured text into structured knowledge triples, with an accuracy rate more than 30% higher than traditional rule-based methods.

[0059] Knowledge graph construction and in-depth indicator association: Building a knowledge graph of financial indicators not only calculates indicator values, but also establishes dependencies between indicators, supports anomaly tracing, impact analysis and association recommendation, and achieves a qualitative leap from "calculation" to "understanding".

[0060] Multi-dimensional anomaly detection with high accuracy: It integrates four detection mechanisms, namely statistical threshold, time series prediction, peer comparison, and historical comparison, and outputs the final judgment through weighted voting. The anomaly detection accuracy is improved by 40%, and the false alarm rate is reduced by more than 50%.

[0061] Intelligent narrative reports, automated interpretation: Based on NLG technology, financial analysis text interpretation is automatically generated, realizing the integration of "charts + interpretation". Decision-makers do not need to spend extra time understanding the meaning of the data, improving decision-making efficiency by 60%.

[0062] Full-process automation significantly improves efficiency: Configure scheduled tasks to automatically execute the entire process of data collection, indicator calculation, anomaly detection, and report generation, reducing the workload of finance personnel by more than 80% and shortening the time for generating quick reports from several hours to minutes.

[0063] Closed-loop feedback optimization enables continuous system evolution: By recording human feedback results and dynamically adjusting fusion weights and model parameters, the system accuracy continuously improves over time, achieving self-optimization.

[0064] Exemplary device Figure 2 This is a schematic diagram of the structure of an apparatus for automatically checking financial operation reports provided in an exemplary embodiment of the present invention. Figure 2 As shown, the device 200 includes: The data acquisition module 210 is used to collect raw data related to financial operations from multiple data sources through a data integration interface. The raw data includes at least one of the following: structured data from the financial system, unstructured text data from the notes to the financial statements, and external market data. Extraction module 220 is used to perform entity recognition and relation extraction on unstructured text data in the original data using a deep learning-based financial semantic understanding model, and output structured financial knowledge triples. The calculation module 230 is used to calculate various financial indicators based on structured data and financial knowledge triples, and to construct a financial indicator knowledge graph. The knowledge graph is used to represent the calculation dependencies and business relationships between various financial indicators. The detection module 240 is used to perform parallel detection of financial indicators using multiple anomaly detection mechanisms. These mechanisms include at least two of the following: statistical threshold detection, time series prediction detection, peer comparison detection, and historical period comparison detection. The module outputs the initial anomaly scores for each detection mechanism. The fusion module 250 is used to weight and fuse the initial anomaly scores of each detection mechanism, and output the final anomaly judgment result based on the fusion result; The generation module 260 is used to automatically generate financial operation reports containing data charts and text interpretations based on financial indicators and the final anomaly judgment results, using natural language generation technology.

[0065] Exemplary electronic devices Figure 3 This is the structure of an electronic device provided in an exemplary embodiment of the present invention. For example... Figure 3As shown, the electronic device 30 includes one or more processors 31 and memory 32.

[0066] The processor 31 may be a central processing unit (CPU) or other form of processing unit with data processing and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions.

[0067] The memory 32 may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc. One or more computer program instructions may be stored on the computer-readable storage medium, and the processor 31 may execute the program instructions to implement the methods of the software programs of the various embodiments of the present invention described above, and / or other desired functions. In one example, the electronic device may also include an input device 33 and an output device 34, these components being interconnected via a bus system and / or other forms of connection mechanisms (not shown).

[0068] In addition, the input device 33 may also include, for example, a keyboard, a mouse, etc.

[0069] The output device 34 can output various information to the outside. The output device 34 may include, for example, a display, a speaker, a printer, and a communication network and its connected remote output devices, etc.

[0070] Of course, for the sake of simplicity, Figure 3 Only some of the components of the electronic device relevant to the present invention are shown, omitting components such as buses, input / output interfaces, etc. In addition, the electronic device may include any other suitable components depending on the specific application.

[0071] Exemplary computer program products and computer-readable storage media In addition to the methods and apparatus described above, embodiments of the present invention may also be computer program products, which include computer program instructions that, when executed by a processor, cause the processor to perform the steps in the methods according to various embodiments of the present invention described in the "Exemplary Methods" section above.

[0072] The computer program product can be written in any combination of one or more programming languages ​​to perform the operations of the embodiments of the present invention. The programming languages ​​include object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C or similar languages. The program code can be executed entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server.

[0073] Furthermore, embodiments of the present invention may also be computer-readable storage media storing computer program instructions thereon, which, when executed by a processor, cause the processor to perform the steps of the methods for information mining of historical change records according to various embodiments of the present invention as described in the "Exemplary Methods" section above.

[0074] The computer-readable storage medium may be any combination of one or more readable media. A readable medium may be a readable signal medium or a readable storage medium. A readable storage medium may be, for example, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, system, or device, or any combination thereof. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0075] The basic principles of the present invention have been described above with reference to specific embodiments. However, it should be noted that the advantages, benefits, and effects mentioned in the present invention are merely examples and not limitations, and should not be considered as essential features of each embodiment of the present invention. Furthermore, the specific details disclosed above are for illustrative and facilitative purposes only, and are not limitations. These details do not limit the present invention to the necessity of employing the aforementioned specific details.

[0076] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For system embodiments, since they largely correspond to method embodiments, the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0077] The block diagrams of devices, systems, devices, and systems involved in this invention are merely illustrative examples and are not intended to require or imply that they must be connected, arranged, or configured in the manner shown in the block diagrams. As those skilled in the art will recognize, these devices, systems, devices, and systems can be connected, arranged, and configured in any manner. Words such as “comprising,” “including,” “having,” etc., are open-ended terms meaning “including but not limited to,” and are used interchangeably with them. The terms “or” and “and” as used herein refer to the terms “and / or,” and are used interchangeably with them unless the context clearly indicates otherwise. The term “such as” as used herein refers to the phrase “such as but not limited to,” and is used interchangeably with it.

[0078] The methods and systems of the present invention may be implemented in many ways. For example, they may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above-described order of steps for the methods is for illustrative purposes only, and the steps of the methods of the present invention are not limited to the order specifically described above unless otherwise specifically stated. Furthermore, in some embodiments, the present invention may also be implemented as a program recorded on a recording medium, the program comprising machine-readable instructions for implementing the methods according to the present invention. Thus, the present invention also covers recording media storing programs for performing the methods according to the present invention.

[0079] It should also be noted that in the systems, apparatus, and methods of the present invention, the components or steps can be disassembled and / or recombined. These disassemblies and / or recombinations should be considered equivalents of the present invention. The above description of the disclosed aspects is provided to enable any person skilled in the art to make or use the invention. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other aspects without departing from the scope of the invention. Therefore, the invention is not intended to be limited to the aspects shown herein, but rather to be carried out within the widest scope consistent with the principles and novel features disclosed herein.

[0080] The above description has been given for purposes of illustration and description. Furthermore, this description is not intended to limit the embodiments of the invention to the forms disclosed herein. Although numerous exemplary aspects and embodiments have been discussed above, those skilled in the art will recognize certain variations, modifications, alterations, additions, and sub-combinations therein.

Claims

1. A method of automatically checking financial operational flash reports, characterized by, include: Through the data integration interface, raw data related to financial operations is collected from multiple data sources. The raw data includes at least one of the following: structured data from the financial system, unstructured text data from the notes to the financial statements, and external market data. For the unstructured text data in the original data, a deep learning-based financial semantic understanding model is applied to perform entity recognition and relation extraction, and output structured financial knowledge triples. Based on the structured data and the financial knowledge triples, various financial indicators are calculated, and a financial indicator knowledge graph is constructed. The knowledge graph is used to represent the computational dependencies and business relationships between the various financial indicators. Multiple anomaly detection mechanisms are applied to the financial indicators in parallel. These multiple anomaly detection mechanisms include at least two of the following: statistical threshold detection, time series prediction detection, peer comparison detection, and historical period comparison detection. The initial anomaly score of each detection mechanism is output. The initial anomaly scores of each detection mechanism are weighted and fused, and the final anomaly determination result is output based on the fusion result. Based on the aforementioned financial indicators and the final anomaly determination results, natural language generation technology is applied to automatically generate a financial operations report that includes data charts and textual interpretations.

2. The method of claim 1, wherein, The financial semantic understanding model is a BERT model or RoBERTa model fine-tuned based on financial domain corpus, used to identify entity types and entity relationships in financial text. The entity types include revenue entities, cost entities, expense entities, asset entities, liability entities, and equity entities. The entity relationships include calculation relationships, year-on-year relationships, month-on-month relationships, and percentage relationships.

3. The method of claim 1, wherein, The specific steps for constructing the financial indicator knowledge graph include: Based on standard calculation formulas in the financial field, define the calculation dependencies between various financial indicators; The calculated financial metrics are used as nodes in the knowledge graph, and the computational dependencies are used as edges in the knowledge graph. The financial metric knowledge graph is used for: When a certain financial indicator becomes abnormal, the knowledge graph is used to conduct a source tracing analysis to identify the lower-level indicators that caused the abnormality. Impact analysis is performed using the knowledge graph to predict the impact of this indicator anomaly on higher-level indicators.

4. The method of claim 1, wherein, The multi-dimensional anomaly detection mechanism specifically includes: Statistical threshold detection: Based on a preset threshold range, determine whether financial indicators exceed the normal range; Time series prediction detection: The time series prediction model is trained based on historical data, and the residual between the predicted value and the actual value is calculated. When the residual exceeds the dynamic threshold, it is judged as an anomaly. Peer comparison detection: Obtain the same period indicator data of comparable companies in the same industry, calculate the deviation of the company's indicators from the industry average, and judge it as abnormal when the deviation exceeds the threshold; Historical comparison detection: Compare the indicator data with the same period last year or the previous period to calculate the year-on-year change rate or month-on-month change rate. When the change rate exceeds the threshold, it is judged as abnormal.

5. The method of claim 4, wherein, The time series prediction detection uses an ARIMA model, a Prophet model, or an LSTM neural network to train historical time series, and makes anomaly judgments based on the residuals between predicted and actual values. The dynamic threshold is adaptively calculated based on the statistical distribution of historical residuals.

6. The method of claim 1, wherein, The initial anomaly scores from each detection mechanism are weighted and fused, and the final anomaly determination result is output based on the fusion result, including: The fusion weights are dynamically allocated based on the historical accuracy of each anomaly detection mechanism. The initial anomaly scores of each detection mechanism are multiplied by their corresponding weights and then summed to obtain the comprehensive anomaly score. When the overall anomaly score exceeds a preset fusion threshold, the final anomaly determination result is output. Record the manual feedback results of each anomaly detection for subsequent adjustment of fusion weights.

7. The method of claim 1, wherein, The production steps for the aforementioned financial operations report include: Based on the preset report template, generate data charts containing key financial indicators; For financial indicators that have undergone significant changes, a preset interpretation template is automatically matched based on the direction and magnitude of the change; For complex and abnormal scenarios, a natural language generation model is invoked to generate personalized text interpretations based on the dependencies in the financial indicator knowledge graph and the final anomaly determination results. It integrates data charts and text interpretations into a complete financial operations report, supporting multiple output formats.

8. A device for automatically checking financial operation reports, characterized in that, include: The data acquisition module is used to collect raw data related to financial operations from multiple data sources through a data integration interface. The raw data includes at least one of the following: structured data from the financial system, unstructured text data from the notes to the financial statements, and external market data. The extraction module is used to perform entity recognition and relation extraction on the unstructured text data in the original data using a deep learning-based financial semantic understanding model, and output structured financial knowledge triples. The calculation module is used to calculate various financial indicators based on the structured data and the financial knowledge triples, and to construct a financial indicator knowledge graph, which is used to represent the calculation dependencies and business relationships between the various financial indicators. The detection module is used to apply multiple anomaly detection mechanisms to the financial indicators in parallel. The multiple anomaly detection mechanisms include at least two of the following: statistical threshold detection, time series prediction detection, peer comparison detection, and historical period comparison detection. The module outputs the initial anomaly score for each detection mechanism. The fusion module is used to weight and fuse the initial anomaly scores of each detection mechanism, and output the final anomaly judgment result based on the fusion result; The generation module is used to automatically generate a financial operation report containing data charts and text interpretations based on the financial indicators and the final anomaly determination results, using natural language generation technology.

9. A computer-readable storage medium, characterized in that, The storage medium stores a computer program for performing the method described in any one of claims 1-7.

10. An electronic device, characterized in that, The electronic device includes: processor; Memory used to store the processor's executable instructions; The processor is configured to read the executable instructions from the memory and execute the instructions to implement the method described in any one of claims 1-7.

Images