Financial performance assessment supervision method and system based on artificial intelligence

Abstract

The invention provides a financial performance assessment supervision method and system based on artificial intelligence, and the method comprises the steps: carrying out the first verification of the obtained structural data and non-structural data of a financial system based on the consistency of data, and obtaining a first verification result, performing secondary verification on the structured data and the unstructured data based on a preset baseline model to obtain a second verification result; based on the first verification result and the second verification result, screening and removing the structured data and the unstructured data to obtain target structured data and target unstructured data; and preprocessing the target structure data and the target non-structure data to obtain standardized data, and performing multi-dimensional feature extraction on the standardized data to respectively obtain financial time sequence features, index association features and industry reference features. Compared with manual operation, the data acquisition efficiency can be improved, and human errors are reduced.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based financial performance assessment and supervision method and system. Background Art

[0002] In today's digital and intelligent world, financial performance appraisals are crucial for a company's stable development and competitiveness. Accurate and scientific financial performance appraisals can provide strong support for strategic decision-making, rationally allocate resources, motivate employees, and ultimately help companies achieve their long-term goals.

[0003] Currently, traditional financial performance appraisal and supervision methods rely heavily on the manual collection and organization of financial data, which not only consumes significant manpower, material resources, and time, but is also prone to human error. For example, when compiling financial statement data, manual entry errors can lead to data bias, which in turn affects the accuracy of performance appraisals. Furthermore, when faced with comprehensive analysis of massive and complex financial data, traditional methods can only perform simple assessments based on limited data indicators, failing to deeply explore the correlations between data from different sources. This makes it impossible to fully reflect the overall financial status of the company and the overall performance of employees in their financial work, further affecting the accuracy of the performance appraisal and supervision process. Summary of the Invention

[0004] The present invention provides an artificial intelligence-based financial performance appraisal and supervision method and system, which are used to improve data collection efficiency, reduce human errors, avoid the time cost of manual data entry, and comprehensively consider financial performance from multiple angles, make full use of data feature information, and more accurately evaluate the company's financial status and employee work results, avoiding the problem that traditional methods are unable to fully reflect the overall financial picture and employee performance due to insufficient evaluation dimensions.

[0005] In a first aspect, the present invention provides a financial performance assessment and supervision method based on artificial intelligence, comprising:

[0006] Performing a primary verification on the structured data and unstructured data of the acquired financial system based on data consistency to obtain a first verification result, and performing a secondary verification on the structured data and the unstructured data based on a preset baseline model to obtain a second verification result;

[0007] Based on the first verification result and the second verification result, screening and removing the structured data and the unstructured data to obtain target structured data and target unstructured data;

[0008] Preprocessing the target structured data and the target unstructured data to obtain standardized data, and performing multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features;

[0009] Assess and evaluate financial performance based on the financial time series characteristics, the indicator correlation characteristics, and the industry benchmark characteristics to obtain an assessment result;

[0010] Conduct supervision and analysis on the assessment results, and determine response strategies based on the supervision and analysis results.

[0011] In a second aspect, the present invention further provides an artificial intelligence-based financial performance assessment and supervision system, which is applied to the artificial intelligence-based financial performance assessment and supervision method as described in the first aspect; the artificial intelligence-based financial performance assessment and supervision system includes:

[0012] An acquisition verification module is configured to perform a primary verification on the structured data and unstructured data of the acquired financial system based on data consistency to obtain a first verification result, and perform a secondary verification on the structured data and the unstructured data based on a preset baseline model to obtain a second verification result;

[0013] a verification and removal module, configured to screen and remove the structured data and the unstructured data based on the first verification result and the second verification result to obtain target structured data and target unstructured data;

[0014] A feature extraction module is used to preprocess the target structured data and the target unstructured data to obtain standardized data, and perform multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features;

[0015] An assessment and evaluation module, configured to assess and evaluate financial performance based on the financial time series characteristics, the indicator correlation characteristics, and the industry benchmark characteristics, and obtain an assessment and evaluation result;

[0016] The supervision and response module is used to supervise and analyze the assessment results and determine the response strategy based on the supervision and analysis results.

[0017] In a third aspect, the present invention also provides an electronic device comprising: a memory for storing a computer software program; and a processor for reading and executing the computer software program, thereby implementing any of the above-mentioned artificial intelligence-based financial performance appraisal and supervision methods.

[0018] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium, wherein the storage medium stores a computer software program, and when the computer software program is executed by a processor, it implements any of the above-mentioned artificial intelligence-based financial performance appraisal and supervision methods.

[0019] In a fifth aspect, the present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-mentioned artificial intelligence-based financial performance appraisal and supervision methods.

[0020] The artificial intelligence-based financial performance appraisal supervision method provided by the embodiment of the present invention automatically obtains structured data and unstructured data and performs multiple verifications. Compared with manual operations, it can not only improve data collection efficiency and reduce human errors, but also avoid the time cost of manual entry, and effectively eliminate erroneous and inconsistent data, improve data accuracy, and provide a reliable data basis for subsequent performance appraisals; in addition, through the extraction of multi-dimensional features and the construction of multi-dimensional evaluation models, it breaks through the limitations of traditional manual methods, can deeply explore the correlation between data from different sources, comprehensively reflect the financial status of the enterprise and the comprehensive performance of employees, and can comprehensively consider financial performance from multiple angles, make full use of data feature information, and more accurately evaluate the financial status of the enterprise and the work results of employees, avoiding the problem that traditional methods cannot fully reflect the overall financial picture and employee performance due to insufficient evaluation dimensions, and improve the accuracy of the performance appraisal supervision process. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flowchart of an artificial intelligence-based financial performance assessment and supervision method provided by an embodiment of the present invention;

[0022] Figure 2 This is a schematic diagram of the structure of an artificial intelligence-based financial performance assessment and supervision system provided by an embodiment of the present invention;

[0023] Figure 3 An embodiment diagram of an electronic device provided by an embodiment of the present invention;

[0024] Figure 4 A diagram of an embodiment of a computer-readable storage medium provided for an embodiment of the present invention. DETAILED DESCRIPTION

[0025] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 those skilled in the art without making any creative efforts shall fall within the scope of protection of the present invention.

[0026] In the description of the present invention, the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly specify the number of technical features indicated. Therefore, features specified as "first" or "second" may explicitly or implicitly include one or more features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0027] In the description of the present invention, the term "for example" is used to mean "used as an example, illustration or explanation". Any embodiment of the present invention described as "for example" is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is given to enable any person skilled in the art to implement and use the present invention. In the following description, details are listed for the purpose of explanation. It should be understood that a person of ordinary skill in the art can recognize that the present invention can be implemented without using these specific details. In other examples, well-known structures and processes are not elaborated in detail to avoid obscuring the description of the present invention with unnecessary details. Therefore, the present invention is not intended to be limited to the embodiments shown, but is consistent with the widest scope consistent with the principles and features disclosed herein.

[0028] See Figure 1 , Figure 1 This is a flow chart of the artificial intelligence-based financial performance assessment and supervision method provided by the present invention. In the embodiment of the present invention, the execution subject of the artificial intelligence-based financial performance assessment and supervision method is the assessment and supervision system. Therefore, the artificial intelligence-based financial performance assessment and supervision method includes:

[0029] Step 10: Verify the structured data and unstructured data of the acquired financial system based on data consistency to obtain a first verification result, and perform a second verification on the structured data and unstructured data based on a preset baseline model to obtain a second verification result.

[0030] Optionally, the assessment and supervision system first automatically captures structured data from data sources such as the ERP system, electronic invoice system, and bank reconciliation system through the API interface in accordance with the preset data transmission protocol. Then, for unstructured data in the financial system, such as paper vouchers, OCR image recognition technology is used to identify and scan text, data and other information on the paper vouchers to achieve the acquisition of unstructured data. At the same time, distributed crawler technology can also be used to obtain structured and unstructured data in the financial system, thereby achieving fast and stable data collection from various systems without manual copying and pasting.

[0031] Furthermore, the performance monitoring system processes the acquired structured and unstructured data. Since data accuracy and reliability are crucial for financial performance monitoring, after acquiring structured data (such as numerical data in financial statements, stored in a relational database as two-dimensional tables, such as the amounts of each asset and liability item in a balance sheet) and unstructured data (such as text descriptions in financial reports and meeting minutes) from the financial system, it first performs a data consistency verification. Specifically, blockchain hash value verification technology is used to verify the consistency of data collected across systems. After collection, each piece of data is assigned a unique hash value, similar to a data "fingerprint." When data is transmitted and stored between different systems, hash value comparisons can be used to determine whether it has been tampered with or corrupted. If the hash values ​​are consistent, the data is complete and accurate; if they are inconsistent, there may be issues with the data, requiring further investigation and correction, as described in steps 1011 through 1014.

[0032] Furthermore, after the assessment and supervision system completes the first verification using hash value verification technology, a second verification is performed on the structured and unstructured data based on fuzzy sets and a pre-set baseline model. The baseline model can be a machine learning model trained with a large amount of historical financial data, such as a linear regression model or decision tree model used to predict financial indicators, or a text classification model based on deep learning, as described in steps 1021-1024.

[0033] Furthermore, the embodiments of the present invention can quickly identify obvious data errors and inconsistencies through a single verification, saving time and resources for subsequent more in-depth analysis, ensuring the quality of basic data, providing a relatively reliable data foundation for subsequent secondary verification, and improving the accuracy and efficiency of overall data verification. The secondary verification is beneficial to the learning ability of the model. Compared with the simple rule-based inspection of the first verification, the baseline model can comprehensively consider the complex relationships between multiple factors, thereby discovering more hidden data anomalies, providing more comprehensive and in-depth verification of the data, enhancing the credibility of the data, and providing a more solid guarantee for subsequent screening and analysis.

[0034] Step 20: Based on the first verification result and the second verification result, the structured data and the unstructured data are screened and removed to obtain target structured data and target unstructured data.

[0035] Optionally, the assessment and supervision system filters and removes structured data and unstructured data based on the results of the first and second verifications. For structured and unstructured data, corresponding screening rules are set based on the integrity, consistency, and rationality issues found in the first verification, as well as the deviations in the model predictions in the second verification. For example, if the first verification finds that the customer name in a sales record is empty (integrity issue), or if the second verification finds that the cost data of a product deviates by more than 20% from the cost predicted based on historical data and market conditions (rationality issue), these data records will be marked for removal. This ensures the quality of data used for subsequent analysis and improves the accuracy of assessment and supervision.

[0036] Step 30: pre-process the target structured data and the target unstructured data to obtain standardized data, and perform multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features.

[0037] Optionally, the assessment and supervision system preprocesses the filtered target structure data, including data cleaning (such as removing duplicate records), data conversion (such as unifying the date format) and data normalization. Specifically, during data normalization, the minimum-maximum scaling method is used to map the data to the [0, 1] interval, and the formula can be used: Where X0 represents the original data, X max 、X min Represents the minimum and maximum values ​​of the original data respectively. For example, for sales amount data, assuming the minimum value is 1,000 yuan and the maximum value is 100,000 yuan, if the sales amount of a record is 5,000 yuan, the normalized value is: For the target unstructured data, natural language processing techniques are used to perform operations such as text cleaning (removing stop words and special symbols), word segmentation (segmenting the text into individual words), and word vector conversion (converting the text into a computer-processable vector form, such as using models such as Word2Vec and GloVe). This ultimately results in preprocessed, standardized data. Data preprocessing unifies the data format and scope, eliminating the impact of differences in data dimensions and formats on subsequent analysis, allowing different types of data to be compared and analyzed on the same basis. This improves the readability and processability of the data, providing a higher-quality data foundation for multi-dimensional feature extraction and enhancing the accuracy and effectiveness of feature extraction.

[0038] Furthermore, the assessment and supervision system extracts multi-dimensional features from the pre-processed standardized data, including financial time series feature extraction, indicator correlation feature extraction, and industry benchmark feature extraction. Specifically, for financial time series feature extraction, feature extraction is performed by analyzing the changing trends of financial data over time. For example, for monthly sales data, its monthly growth rate can be calculated using the following formula:

[0039]

[0040] Among them, Sales cur Represents the sales of the current month, Sales pre It is expressed as the sales of the previous month. In addition, you can also calculate the moving average, such as the 5-month moving average, the formula is:

[0041] Sales preri It is represented by the sales of the previous i months. By analyzing these time series characteristics, we can more clearly understand the dynamic changes in the company's financial status.

[0042] For the extraction of indicator correlation features, we can mine the correlation between different financial indicators. For example, by calculating the debt-to-asset ratio ( Total lia Expressed as the total liabilities of the enterprise, Total las Expressed as the total assets of the enterprise), current ratio ( Among them, Curent ass Expressed as the enterprise's current assets, Current lia The Pearson correlation coefficient can be used to measure the linear correlation between indicators. The formula is: where x i and y i are the data values ​​of the two indicators, and are the average values ​​of the two indicator data, and n is the number of data points.

[0043] To extract industry benchmark characteristics, a company's financial data is compared and analyzed with that of other companies in the same industry to extract industry benchmark characteristics. For example, the average gross profit margin, asset turnover rate, and other indicators of companies in the same industry can be used as industry benchmarks. The difference between the company and the industry benchmark, such as the gross profit margin difference, can then be calculated. This difference analysis can help understand the company's competitiveness and position in the industry.

[0044] The embodiments of this invention comprehensively depict a company's financial status from multiple perspectives, providing rich, comprehensive feature information for financial performance evaluation, making the evaluation results more accurate and objective. Financial time series features can be used to understand a company's development trends, indicator correlation features can reveal the relationships between internal financial indicators, and industry benchmark features can clarify a company's position within the industry. Combining these features allows for a more comprehensive assessment of a company's financial performance.

[0045] Step 40: assess and evaluate the financial performance based on the financial time series characteristics, indicator correlation characteristics, and industry benchmark characteristics to obtain an assessment result.

[0046] Optionally, after obtaining the extracted financial time series features, indicator correlation features, and industry benchmark features, the assessment and supervision system utilizes deep learning models and data processing technology to perform a fusion analysis of the extracted multi-dimensional features, and then utilizes the trained multi-dimensional evaluation model to analyze the fused features to obtain the corresponding assessment and evaluation results, as specifically described in steps 401 to 405, thereby achieving a comprehensive evaluation of performance appraisals and comprehensively considering features of multiple dimensions, avoiding the one-sidedness of single indicator evaluation, and being able to more comprehensively and accurately reflect the financial performance of the enterprise. Compared to using only a single indicator such as net profit to evaluate corporate performance, a comprehensive evaluation of multi-dimensional features can cover multiple aspects such as the company's development trends, internal financial relationships, and industry competitiveness, making the evaluation results more valuable for reference, and providing companies with detailed financial performance analysis reports to help them identify their own strengths and weaknesses, and provide a clear direction for improving and enhancing financial performance.

[0047] Step 50: Conduct supervision and analysis on the assessment results, and determine a response strategy based on the supervision and analysis results.

[0048] Optionally, after determining the assessment results, the assessment and supervision system first determines the dynamic indicator threshold for the current time period based on the assessment indicators in the historical financial performance data. The system then compares the performance assessment indicators in the assessment results with the dynamic indicator threshold to determine the degree of deviation from the current financial performance and formulates a response strategy, as described in steps 501 to 504. This allows for timely identification of financial issues and the implementation of measures to prevent the expansion of financial risks.

[0049] The embodiment of the present invention automatically acquires structured data and unstructured data and performs multiple verifications. Compared with manual operations, it can not only improve data collection efficiency and reduce human errors, but also avoid the time cost of manual entry, and effectively eliminate erroneous and inconsistent data, improve data accuracy, and provide a reliable data basis for subsequent performance appraisals; in addition, through the extraction of multi-dimensional features and the construction of multi-dimensional evaluation models, it breaks through the limitations of traditional manual methods, can deeply explore the correlation between data from different sources, comprehensively reflect the financial status of the enterprise and the comprehensive performance of employees, and can comprehensively consider financial performance from multiple angles, make full use of data feature information, and more accurately evaluate the financial status of the enterprise and the work results of employees, avoiding the problem that traditional methods cannot fully reflect the overall financial picture and employee performance due to insufficient evaluation dimensions, and improve the accuracy of the performance appraisal supervision process.

[0050] In one embodiment, steps 1011 to 1014 are described as follows:

[0051] Step 1011 : assign identifiers and timestamps to the structured data and the unstructured data respectively to obtain assignment results.

[0052] Optionally, when acquiring structured and unstructured data from the financial system, the assessment and supervision system first assigns unique identifiers to structured data (such as figures and account details in financial statements) and unstructured data (such as financial report documents and invoice images). For example, each invoice is assigned a specific number and each account is assigned a unique code. Simultaneously, the system captures the timestamp of the data's creation or last modification, accurate to the millisecond level, to facilitate subsequent accurate identification and tracking of the data.

[0053] Furthermore, when assigning unique identifiers to structured data, the assessment and supervision system can also utilize the database's primary key generation mechanism. For example, in a MySQL database, an auto-incrementing primary key can be set to automatically generate a unique identifier for each new structured data record inserted. For unstructured data, a hash algorithm (such as MD5 or SHA-256) can be used in conjunction with the data's unique characteristics (such as the file path or a partial summary of the file content) to generate a unique identifier.

[0054] Step 1012: Perform hash processing on the allocation result with the timestamp and identifier to obtain an initial hash value.

[0055] Optionally, the assessment and supervision system uses a strong cryptographic hash algorithm, such as SHA-512, based on the identifier and timestamp assigned in the previous step. The data with the timestamp and identifier is used as input and hashed to generate an initial hash value. This strong encryption ensures that data is difficult to tamper with or forge, and even small changes are detectable, enabling timely detection of data anomalies and ensuring data authenticity.

[0056] Step 1013: Combine the initial hash value and the hash value of the previous adjacent data based on the time sequence or the business logic sequence to obtain a hash chain, and store the hash chain in the distributed node.

[0057] Optionally, the assessment and supervision system combines the initial hash value of the current data with the hash value of the data generated at the previous time point in chronological order or business logic order, such as in the chronological order of data generation. For example, assuming there are three data, whose initial hash values ​​are H1, H2, and H3, respectively, in chronological order, first combine H1 and H2 into a new string S1=H1+H2, and then perform hash calculation on S1 to obtain H12=HashFunction(S1). Then combine H12 and H3 into S2=H12+H3, and then perform hash calculation on S2 to obtain H123=HashFunction(S2), where H123 represents the final hash value of the hash chain formed by these three data based on chronological order. Afterwards, in a distributed system, each intermediate hash value and the final hash value can be stored on different distributed nodes, and storage management is performed through a distributed storage protocol (such as Ceph, GlusterFS, etc.).

[0058] Step 1014: compare the hash chain of the current data with the stored hash chain to obtain a first verification result.

[0059] Optionally, when performing a comparison, the assessment and supervision system first obtains the final hash value of the hash chain generated by the current data according to the above steps. Then, the final hash value of the corresponding hash chain previously stored is obtained from the distributed node. The two hash values ​​are accurately compared. If they are exactly the same, it means that the current data is consistent with the previously stored data throughout the entire process, from the assignment of identifiers and timestamps to the generation of the hash chain, and no data tampering or errors have occurred. If the hash values ​​are different, further analysis is required. Starting from each link in the hash chain, each intermediate hash value can be compared step by step to determine at which data processing step the difference occurred. For example, first compare the second-to-last intermediate hash value. If they are the same, then the difference may have occurred in the hash calculation link of the last data; if they are different, continue to trace back and compare.

[0060] This embodiment of the present invention ensures the reliability of financial data at multiple levels through a series of steps, including identifier and timestamp assignment, hash processing, and hash chain comparison. Both structured and unstructured data can be effectively verified and managed within this system. Furthermore, the construction of hash chains provides strong data traceability. From the final hash chain hash value, each source data hash value, along with its corresponding identifier and timestamp, can be traced back layer by layer, providing powerful support for financial audits and data problem troubleshooting.

[0061] In one embodiment, steps 1021 to 1024 are described as follows:

[0062] Step 1021 : extract data features from the structured data and the unstructured data to obtain feature values.

[0063] Optionally, the assessment and supervision system extracts key features from the attributes of the structured financial data based on the data itself. In terms of numerical features, such as the amount in the financial statements and the quantity in the purchase order, these values ​​directly reflect the scale and extent of the business; time features include the time when the transaction occurred, the time of financial accounting, etc., which can reflect the chronological order and timeliness of the business; category features cover account categories (such as assets and liabilities) and business types (such as procurement and sales), which are used to distinguish financial activities of different natures. For unstructured data, natural language processing technology is used to mine key semantic features from text information. For example, extract key terms in contract documents, such as payment terms, delivery date, etc.; extract information such as supplier name, product or service name in the text content of the invoice image. This makes it easier to understand and process the data.

[0064] Step 1022: Convert the eigenvalues ​​into fuzzy sets and determine the membership of each eigenvalue in the fuzzy set.

[0065] Optionally, the assessment and supervision system can use fuzzy logic theory to divide numerical features, such as the amount in financial data, into "low amount", "medium amount" and "high amount" fuzzy sets based on business scenarios and historical data, taking the amount as an example. If most of the small transaction amounts in the historical data are between 0 and 1,000 yuan, this interval can be set as the "low amount" fuzzy set range. However, it should be noted that the fuzzy set range is not an absolute boundary, and membership functions (such as triangular membership functions and trapezoidal membership functions) can be used to determine the membership of the current amount value in each fuzzy set. For example, for the triangular membership function, for the "low amount" set, the membership is 1 in the 0-500 yuan range, and the membership drops linearly to 0 in the 500-1,000 yuan range.

[0066] Furthermore, the assessment and supervision system divides time characteristics into "near-term," "medium-term," and "long-term" fuzzy sets. Assuming the financial reporting cycle is used as a reference, the last month is "near-term," 1-3 months is "medium-term," and more than 3 months is "long-term." Similarly, a membership function is used to determine the membership of a transaction time within these fuzzy sets. For categorical characteristics, taking accounting subject categories as an example, an account may not belong to a single category. For example, "advance payments" have both liability attributes and are related to sales operations to a certain extent. Rules can be set to determine its membership within different fuzzy category sets.

[0067] Furthermore, in one embodiment, for a sales transaction amount of RMB 800, a triangular membership function is used to determine its membership in the "low amount" fuzzy set. Suppose the triangular membership function of the "low amount" fuzzy set is:

[0068]

[0069] Substituting x=800, we get

[0070] Similarly, for a transaction that is two months from now, determine its membership in the "medium term" fuzzy set. Assume that the trapezoidal membership function of the "medium term" fuzzy set is:

[0071]

[0072] Substituting x=2, we can get μ mid-term (2)=1.

[0073] Step 1023 : updating the preset baseline model based on the current business environment parameters, and determining the verification threshold of each fuzzy set.

[0074] Optionally, the preset baseline model in the assessment and supervision system is built based on historical financial data, reflecting the normal distribution pattern of the company's financial data over a period of time. Current business environment parameters include market volatility (e.g., stock market fluctuations affecting investment return data) and business expansion or contraction (e.g., increased business volume and changes in cost structure during expansion). Therefore, the preset baseline model is updated based on the current business environment parameters. For example, during periods of significant market volatility, the fluctuation range of a company's investment return will increase. The validation threshold corresponding to the "high return" fuzzy set needs to be adjusted, such as upward, based on the degree of market volatility and changes in historical investment return data. If a company's business expands, the normal range of data such as purchase volume and sales volume will change, requiring recalculation of the validation thresholds for fuzzy sets such as "high purchase amount" and "high sales amount." Therefore, by statistically analyzing the distribution shift of historical data under the current business environment, new validation thresholds can be determined using statistical methods (e.g., adjusting the mean and standard deviation) or machine learning algorithms (e.g., predicting new threshold ranges through regression analysis). This allows dynamic validation thresholds to adapt to changing business scenarios, avoid misjudgments caused by fixed standards, and improve validation accuracy and adaptability.

[0075] Step 1024: Compare the membership degree corresponding to each eigenvalue with the verification threshold to obtain a second verification result.

[0076] Optionally, the assessment and supervision system compares the membership of each data feature value in the fuzzy set with the verification threshold of the corresponding fuzzy set one by one. Specifically, for multiple feature values ​​of a financial data (such as amount feature, time feature, etc.), judgments are made separately. If the membership of the amount feature of a certain data in the "high amount" fuzzy set is higher than the upper limit of the verification threshold of the "high amount" fuzzy set, and other related features also show abnormal trends (such as the time feature has an abnormally high membership in the "recent" fuzzy set, which is unreasonable in combination with the business logic), the data is marked as having problems. If the membership of all feature values ​​is within the verification threshold range of their respective fuzzy sets and conforms to the normal pattern learned based on the baseline model, the data is considered to have passed the secondary verification.

[0077] This embodiment of the present invention fully considers the ambiguity and uncertainty of financial data and utilizes fuzzy logic to process data characteristics, making the verification process more tailored to the actual financial data. It can effectively handle complex data scenarios that traditional precise methods struggle to address. Furthermore, by updating the baseline model and verification thresholds based on current business environment parameters, the verification system can adapt to market and business changes, maintaining the ability to effectively identify abnormal data and preventing verification failures due to environmental changes.

[0078] In one embodiment, steps 401 to 405 are described as follows:

[0079] Step 401: Mining the financial changes in the financial time series features based on the long short-term memory network to obtain the financial hidden state of each time step.

[0080] Optionally, the assessment and supervision system mines financial changes in financial time series features using a long short-term memory (LSTM) network. In the financial field, financial time series features include financial data that changes over time, such as monthly income, expenditure, and profit. The LSTM is a special type of recurrent neural network (RNN) that excels at handling long-term dependencies in sequential data. Therefore, the financial time series data is input into the LSTM network in chronological order. The memory units in the network are able to learn and remember financial change information at different time steps. At each time step, the LSTM controls the inflow, retention, and outflow of information through input, forget, and output gates. Based on this, through calculations over multiple time steps, the LSTM can mine the changing patterns of financial data over time and obtain the financial hidden states corresponding to each time step. These hidden states contain rich information about financial changes.

[0081] Step 402: Analyze the causal relationships between different financial indicators in the indicator association features based on the causal convolutional neural network, and construct a causal relationship strength matrix based on the analysis results.

[0082] Optionally, when the assessment and supervision system analyzes the causal relationship between different financial indicators in the indicator correlation characteristics, since in the financial field, there are complex causal relationships between different financial indicators, for example, an increase in sales revenue may lead to an increase in profits, and an increase in costs may compress profit margins, etc., the causal convolutional neural network analyzes the indicator correlation characteristics, and its convolution operation is performed in chronological order, which ensures the rationality of the causal relationship, that is, the output of the current time step only depends on the input of the previous time step and will not be affected by future information. By performing convolution operations on the time series data of multiple financial indicators, the network can learn the strength of the causal relationship between different indicators, and then construct a causal relationship strength matrix. The elements in the matrix represent the strength of the causal relationship between different financial indicators. For example, the larger the value of an element in a row or column in the matrix, the stronger the causal relationship between the corresponding two financial indicators.

[0083] Step 403 : performing difference processing on the target unstructured data, the target structured data and the industry average data based on the industry benchmark characteristics, and performing dimensionality reduction on the difference processing results to obtain the industry benchmark difference characteristics.

[0084] Optionally, the target unstructured data (such as financial report texts, contract texts, etc.) and target structured data (such as numerical data in financial statements) obtained by the assessment and supervision system contain the company's own financial information, while the industry average data represents the general financial level of companies in the same industry, and the industry benchmark characteristics reflect the average financial status within the industry. Therefore, for structured data, the numerical difference is directly calculated; for unstructured data, key semantic information is first extracted and quantified through technologies such as natural language processing, and then the difference is calculated with the corresponding industry average information; and because the processed data dimension may be high and contain redundant information, dimensionality reduction technology (such as principal component analysis PCA, linear discriminant analysis LDA, etc.) is used to reduce the data dimension while retaining the main information of the data as much as possible to obtain industry benchmark difference characteristics.

[0085] Furthermore, in one embodiment, consider a manufacturing company whose target structured data includes five financial indicators: debt-to-asset ratio, current ratio, and net profit margin. The key financial indicators extracted from the target unstructured data include accounts receivable turnover and inventory turnover. The average data for these seven indicators across the same industry are obtained, and the difference between the company's own indicators and the industry average is calculated, resulting in a seven-dimensional difference value vector. Principal component analysis (PCA) is then used to reduce the dimensionality of this seven-dimensional vector, retaining the first three principal components and converting the seven-dimensional vector into a three-dimensional industry-benchmark difference feature vector.

[0086] Step 404 , performing feature fusion on the financial hidden state, the causal correlation strength matrix, and the industry benchmark difference feature to obtain a target fusion feature.

[0087] Optionally, the assessment and supervision system fuses the previously obtained financial hidden state, causal correlation strength matrix and industry benchmark difference features. It can fuse the above features by constructing a topological graph and combining it with a graph convolutional neural network to finally obtain the target fusion features, as described in steps 4041 to 4044.

[0088] In step 405, the target fusion features are input into the multi-dimensional evaluation model to obtain the assessment results output by the multi-dimensional evaluation model; the multi-dimensional evaluation model is trained based on the sample fusion features and their corresponding assessment label results.

[0089] Optionally, the multi-dimensional evaluation model built into the assessment and supervision system is a pre-trained model based on a large number of sample fusion features and their corresponding assessment and evaluation label results. The sample fusion features come from the fusion features obtained by processing the financial data of many companies in the previous steps, and the assessment and evaluation label results are the evaluation conclusions (such as excellent, good, medium, poor, etc.) obtained by these companies through professional financial analysis or actual operating results. The target fusion features are input into the trained multi-dimensional evaluation model. The model processes and maps the input features through internal parameters and structures (such as neural network weights, activation functions, etc.), and ultimately outputs the assessment and evaluation results of the target company, which represents the company's assessment level or specific score in terms of financial performance or the score corresponding to each performance indicator.

[0090] The embodiment of the present invention analyzes and extracts features from financial data from three different perspectives: financial time series features, indicator correlation features, and industry benchmark features. These features are then integrated to comprehensively cover important information such as the temporal changes in financial data, indicator relationships, and industry comparisons, making the assessment results more comprehensive and reliable.

[0091] In one embodiment, steps 4041 to 4044 are described as follows:

[0092] Step 4041 , constructing a topological graph with the financial hidden state, causal correlation strength matrix, and industry benchmark difference features as node sets, the elements in each node set as node attributes, and the logical relationships between features as edges.

[0093] Optionally, the assessment and supervision system treats the financial hidden state, causal strength matrix, and industry benchmark differential features as a collection of nodes in a graph. In the financial hidden state, the hidden state at each time step is a node, and nodes at adjacent time steps are connected due to the temporal order and consistency of financial changes. In the causal strength matrix, nodes corresponding to strongly correlated financial indicators are connected, such as the sales revenue and sales cost nodes, due to their close business causal relationship. In the industry benchmark differential features, nodes related to the same financial indicator are connected, such as the company's own profit margin and the industry average profit margin nodes. The attributes of each node are the element values ​​in the corresponding feature matrix, such as the specific value of the financial hidden state at a specific time step. By constructing edges and nodes in this way, a topological graph is formed, integrating the three types of feature information, intuitively presenting the complex relationships between features, and facilitating the understanding and processing of complex features.

[0094] Step 4042: Update the features of each node in the topology graph based on the graph neural network to obtain updated enhanced features.

[0095] Optionally, the assessment and supervision system uses a message passing mechanism of a graph neural network, allowing each node to receive feature information from the nodes connected to it. For example, in a set of financial hidden state nodes, a node at a certain time step will receive feature information from the node at the previous time step. The node will fuse the received information with its own features. The fusion method can be weighted summation, which assigns different weights according to the importance of the connected nodes and then sums them up; or it can use aggregation functions such as average pooling, which averages the received information and its own features, or maximum pooling, which takes the maximum value. In this way, the features of each node are updated to include information from more neighboring nodes, making the node features richer and simulating the propagation and interaction process of information in reality. This enhances the richness of node features, simulates real-world information interaction, and enables the features to more comprehensively reflect information related to financial performance.

[0096] Step 4043: Perform graph convolution processing on the updated enhanced features to obtain target node features.

[0097] Optionally, the assessment and supervision system performs graph convolution operations on the updated node features, which is achieved by sliding the convolution kernel on the graph structure. In the graph, the convolution kernel performs convolution operations with the feature matrix of the node and its neighborhood to extract local structural features and high-order correlation information between nodes. For example, in a causal relationship graph of financial indicators, the convolution kernel can capture the local structural features composed of multiple related financial indicator nodes, as well as more complex high-order correlations between these indicators (not only direct correlations, but also indirect correlations, etc.). Through the convolution operation, a new feature representation, namely the target node feature, is obtained, which further explores the deep connection between features and improves the accuracy and depth of the feature reflection on financial performance.

[0098] In step 4044, global average aggregation or global maximum aggregation is performed on the target node features to obtain the target fusion features.

[0099] Optionally, the assessment and supervision system performs global average aggregation or global maximum aggregation on the target node features after graph convolution processing. Global average aggregation is to calculate the average value of all node feature values ​​to form a fixed-length vector; global maximum aggregation is to take the maximum value of all node feature values ​​to form a vector. These two aggregation methods integrate all node information and condense it into a fusion result that can fully reflect the financial performance characteristics. A fusion feature that can fully reflect the financial performance characteristics is formed. For example, global average aggregation can reflect the average level of all node features and comprehensively reflect the overall financial situation; global maximum aggregation highlights the most significant feature information. The target fusion feature obtained in this way can be used for subsequent financial performance evaluation, providing a concise and comprehensive input for the evaluation model.

[0100] The embodiment of the present invention utilizes graph neural networks to process complex relationships between features. By constructing topological graphs, message passing, graph convolution and other operations, it effectively mines deep correlations and high-order information between features. Compared with traditional methods, it has more advantages in processing multi-feature fusion. From node attribute construction and feature update to graph convolution and aggregation, it gradually enriches and integrates feature information, comprehensively integrates financial hidden states, causal correlation strength matrices and industry benchmark difference features, so that the final target fusion features can fully reflect various aspects of financial performance information and improve evaluation accuracy.

[0101] In one embodiment, steps 501 to 503 are described as follows:

[0102] Step 501 : Decompose the financial performance indicators in the historical financial performance data, and determine the indicator threshold of each financial indicator based on the decomposition result.

[0103] Optionally, the assessment and supervision system obtains historical financial performance data, such as financial performance data from the past 3-5 years, and uses the time-varying sequences of multiple financial performance indicators contained in the historical financial performance data, such as revenue, profit, debt-to-asset ratio, etc., to decompose the financial performance indicators in the historical financial performance data, such as Holt-Winters seasonal decomposition. Decompose the financial performance indicators in the historical financial performance data, and decompose each financial performance indicator into trend items, seasonal items, and random items. The trend item reflects the overall direction of change of the financial indicator over a longer period of time, such as whether the company's revenue increases or decreases year by year; the seasonal item reflects the cyclical fluctuations of the data due to seasonal factors, such as the significant increase in revenue in certain industries in a specific quarter; the random item is the random fluctuation remaining after removing the trend and seasonal factors.

[0104] Furthermore, the assessment and supervision system uses the exponentially weighted moving average (EWMA) algorithm to calculate the dynamic benchmark value of each indicator based on the decomposed trend items. The formula is D t =αY t +(1-α)D t-1 , where D t Expressed as the dynamic benchmark value of period t, Y t It is expressed as the actual value of the t period, α is the smoothing coefficient and (0<α<1). The smoothing coefficient α determines the importance attached to new data. The larger α is, the higher the weight of recent data is. After obtaining the dynamic benchmark value, the dynamic standard deviation is calculated. Determine the dynamic threshold interval [D t -kσ t ,D t +kσ t ], where k represents a coefficient set based on the business risk appetite. When risk appetite is high, the k value can be appropriately reduced, while when risk appetite is low, the k value can be increased. The resulting indicator threshold can adapt to the dynamic changes in financial data. Dynamically determining indicator thresholds, taking into account the trend, seasonal, and random characteristics of financial data, more accurately reflects the normal fluctuation range of financial performance indicators and avoids misjudgments caused by static thresholds.

[0105] Furthermore, let's take the company's monthly sales revenue data for the past five years as an example. Using the Holt-Winters seasonal decomposition algorithm, the sales revenue series is decomposed into trend terms, seasonal terms, and random terms. Assume that by analyzing the trend term, using the EWMA algorithm, with α = 0.3, the dynamic benchmark value D for month t is calculated. t Then calculate the dynamic standard deviation σ of the past 12 months (n=12) t , according to the enterprise risk preference, set k = 2, and determine the dynamic threshold interval [D t -2σ t ,D t +2σ t ].

[0106] Step 502 : Compare each financial performance indicator value in the assessment result with the indicator threshold value to obtain a comparison result.

[0107] Optionally, the assessment and supervision system compares the various financial performance indicator values ​​in the assessment results, such as the actual profit rate and asset turnover rate, with the indicator threshold range determined in the previous step. For indicator values ​​that are higher than the upper threshold, the deviation is calculated. This deviation can reflect the extent to which the indicator value is higher than the upper limit of normal fluctuation; for indicator values ​​below the lower limit of the threshold, the deviation is calculated. This indicator reflects the extent to which the indicator value falls below the lower limit of normal fluctuation. This calculation accurately quantifies the deviation of each financial performance indicator value from the normal range. This allows for quick determination of whether a financial performance indicator is within the normal range, providing intuitive indicator status information to corporate financial managers, enabling them to keep abreast of the company's financial situation.

[0108] Step 503 : Classify the comparison results to determine the financial deviation level, and determine a response strategy based on the financial deviation level.

[0109] Optionally, the assessment and supervision system can categorize financial deviations based on the magnitude of the calculated deviation. These deviations are categorized as mild, moderate, and severe. A deviation greater than 20% is considered severe, indicating a significant deviation from the normal range and potentially a serious issue. A deviation between 10% and 20% is considered moderate, indicating a significant deviation requiring attention. A deviation less than 10% is considered mild, indicating a relatively minor but minor deviation. Different response strategies are implemented for each deviation level. Specifically, for mild deviations, internal reminders are issued and relevant personnel are organized to conduct a business review to identify potential minor issues in business processes and promptly correct them. For moderate deviations, a special investigation is initiated to conduct in-depth analysis of the specific business processes that led to the deviation, such as excessive procurement costs or poor sales channels. Short-term improvement plans are then developed to address the issues in a targeted manner. For severe deviations, a rectification team is established to comprehensively review and optimize the relevant business processes and, if necessary, adjust business strategies to fundamentally address the issues that led to the significant financial performance anomaly.

[0110] The embodiments of the present invention can adapt to changes in financial data over time by decomposing historical financial performance data and dynamically calculating thresholds. Compared with static thresholds, it more accurately reflects the normal fluctuation range of financial indicators and effectively captures financial performance anomalies. It determines the response strategy based on the classification of comparison results, realizes the precise management of financial risks, neither overreacts nor ignores problems, improves the company's sensitivity to financial performance anomalies and the timeliness of response, reduces financial risks, and ensures the financial health of the company.

[0111] Furthermore, the artificial intelligence-based financial performance appraisal and supervision system provided by the present invention is described below. The artificial intelligence-based financial performance appraisal and supervision system described below and the artificial intelligence-based financial performance appraisal and supervision method described above can refer to each other.

[0112] Optional, see Figure 2 , Figure 2 This is a schematic diagram of the structure of the artificial intelligence-based financial performance assessment and supervision system provided by the present invention. The artificial intelligence-based financial performance assessment and supervision system includes:

[0113] An acquisition and verification module 210 is configured to perform a primary verification on the structured data and unstructured data of the acquired financial system based on data consistency to obtain a first verification result, and perform a secondary verification on the structured data and unstructured data based on a preset baseline model to obtain a second verification result;

[0114] A verification and removal module 220 is configured to filter and remove the structured data and the unstructured data based on the first verification result and the second verification result to obtain target structured data and target unstructured data;

[0115] Feature extraction module 230 is used to preprocess the target structured data and target unstructured data to obtain standardized data, and perform multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features;

[0116] An assessment module 240 is used to assess and evaluate financial performance based on financial time series characteristics, indicator correlation characteristics, and industry benchmark characteristics to obtain assessment results;

[0117] The supervision and response module 250 is used to supervise and analyze the assessment results and determine the response strategy based on the supervision and analysis results.

[0118] The embodiment of the present invention automatically acquires structured data and unstructured data and performs multiple verifications. Compared with manual operations, it can not only improve data collection efficiency and reduce human errors, but also avoid the time cost of manual entry, and effectively eliminate erroneous and inconsistent data, improve data accuracy, and provide a reliable data basis for subsequent performance appraisals; in addition, through the extraction of multi-dimensional features and the construction of multi-dimensional evaluation models, it breaks through the limitations of traditional manual methods, can deeply explore the correlation between data from different sources, comprehensively reflect the financial status of the enterprise and the comprehensive performance of employees, and can comprehensively consider financial performance from multiple angles, make full use of data feature information, and more accurately evaluate the financial status of the enterprise and the work results of employees, avoiding the problem that traditional methods cannot fully reflect the overall financial picture and employee performance due to insufficient evaluation dimensions, and improve the accuracy of the performance appraisal supervision process.

[0119] See also Figure 3 , Figure 3 This is a diagram of an embodiment of an electronic device provided by an embodiment of the present invention. Figure 3 As shown, an embodiment of the present invention provides an electronic device 300, including a memory 310, a processor 320, and a computer program 311 stored in the memory 310 and executable on the processor 320. When the processor 320 executes the computer program 311, the following steps are implemented:

[0120] The structured data and unstructured data of the acquired financial system are verified once based on the consistency of the data to obtain a first verification result, and the structured data and unstructured data are verified again based on a preset baseline model to obtain a second verification result;

[0121] Based on the first verification result and the second verification result, the structured data and the unstructured data are screened and removed to obtain target structured data and target unstructured data;

[0122] Preprocess the target structured data and target unstructured data to obtain standardized data, and perform multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features;

[0123] Evaluate financial performance based on financial time series characteristics, indicator correlation characteristics, and industry benchmark characteristics to obtain evaluation results;

[0124] Conduct supervision and analysis on the assessment results, and determine the response strategy based on the results of supervision and analysis.

[0125] See also Figure 4 , Figure 4 Detailed description of an embodiment of a computer-readable storage medium provided by an embodiment of the present invention. Figure 4 As shown, this embodiment provides a computer-readable storage medium 400 on which a computer program 311 is stored. When the computer program 311 is executed by a processor, the following steps are implemented:

[0126] The structured data and unstructured data of the acquired financial system are verified once based on the consistency of the data to obtain a first verification result, and the structured data and unstructured data are verified again based on a preset baseline model to obtain a second verification result;

[0127] Based on the first verification result and the second verification result, the structured data and the unstructured data are screened and removed to obtain target structured data and target unstructured data;

[0128] Preprocess the target structured data and target unstructured data to obtain standardized data, and perform multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features;

[0129] Evaluate financial performance based on financial time series characteristics, indicator correlation characteristics, and industry benchmark characteristics to obtain evaluation results;

[0130] Conduct supervision and analysis on the assessment results, and determine the response strategy based on the results of supervision and analysis.

[0131] On the other hand, the present invention further provides a computer program product, which includes a computer program. The computer program can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the artificial intelligence-based financial performance assessment and supervision method provided by the above methods, which includes:

[0132] The structured data and unstructured data of the acquired financial system are verified once based on the consistency of the data to obtain a first verification result, and the structured data and unstructured data are verified again based on a preset baseline model to obtain a second verification result;

[0133] Based on the first verification result and the second verification result, the structured data and the unstructured data are screened and removed to obtain target structured data and target unstructured data;

[0134] Preprocess the target structured data and target unstructured data to obtain standardized data, and perform multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features;

[0135] Evaluate financial performance based on financial time series characteristics, indicator correlation characteristics, and industry benchmark characteristics to obtain evaluation results;

[0136] Conduct supervision and analysis on the assessment results, and determine the response strategy based on the results of supervision and analysis.

[0137] The system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units. That is, they may be located in one place or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of this embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0138] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus the necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or certain parts of the embodiment.

[0139] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A financial performance assessment and supervision method based on artificial intelligence, characterized in that: include: Performing a primary verification on the structured data and unstructured data of the acquired financial system based on data consistency to obtain a first verification result, and performing a secondary verification on the structured data and the unstructured data based on a preset baseline model to obtain a second verification result; Based on the first verification result and the second verification result, screening and removing the structured data and the unstructured data to obtain target structured data and target unstructured data; Preprocessing the target structured data and the target unstructured data to obtain standardized data, and performing multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features; Assess and evaluate financial performance based on the financial time series characteristics, the indicator correlation characteristics, and the industry benchmark characteristics to obtain an assessment result; Conduct supervision and analysis on the assessment results, and determine response strategies based on the supervision and analysis results.

2. The artificial intelligence-based financial performance assessment and supervision method according to claim 1 is characterized in that: The financial performance is assessed and evaluated based on the financial time series characteristics, the indicator correlation characteristics, and the industry benchmark characteristics to obtain assessment results, including: Mining the financial changes in the financial time series features based on the long short-term memory network to obtain the financial hidden state of each time step; Analyze the causal relationships between different financial indicators in the indicator correlation features based on a causal convolutional neural network, and construct a causal relationship strength matrix based on the analysis results; Based on the industry benchmark features, difference processing is performed on the target unstructured data, the target structured data and the industry average data, and dimension reduction is performed on the difference processing results to obtain industry benchmark difference features; Performing feature fusion on the financial hidden state, the causal correlation strength matrix, and the industry benchmark difference feature to obtain a target fusion feature; The target fusion features are input into a multidimensional evaluation model to obtain an assessment result output by the multidimensional evaluation model; the multidimensional evaluation model is trained based on the sample fusion features and their corresponding assessment label results.

3. The artificial intelligence-based financial performance assessment and supervision method according to claim 2 is characterized in that: The feature fusion of the financial hidden state, the causal correlation strength matrix, and the industry benchmark difference feature to obtain a target fusion feature includes: A topological graph is constructed by using the financial hidden state, the causal association strength matrix, and the industry benchmark difference feature as node sets, the elements in each node set as node attributes, and the logical relationships between the features as edges; Based on the graph neural network, the features of each node in the topological graph are updated to obtain updated enhanced features; Performing graph convolution processing on the updated enhanced features to obtain target node features; Perform global average aggregation or global maximum aggregation on the target node features to obtain the target fusion features.

4. The artificial intelligence-based financial performance assessment and supervision method according to claim 1 is characterized in that: The structured data and unstructured data of the acquired financial system are verified based on the consistency of the data to obtain a first verification result, including: assigning identifiers and timestamps to the structured data and the unstructured data respectively, and obtaining assignment results; Performing hash processing on the allocation result with the timestamp and identifier to obtain an initial hash value; Performing hash processing on the initial hash value and the hash value of the previous adjacent data based on a time sequence or a business logic sequence to obtain a hash chain, and storing the hash chain in a distributed node; The hash chain of the current data is compared with the stored hash chain to obtain the first verification result.

5. The artificial intelligence-based financial performance assessment and supervision method according to claim 1 is characterized in that: The second verification of the structured data and the unstructured data based on the preset baseline model to obtain a second verification result includes: Extracting data features from the structured data and the unstructured data to obtain feature values; Converting the eigenvalues ​​into fuzzy sets, and determining the membership of each eigenvalue in the fuzzy sets; Updating the preset baseline model based on the current business environment parameters and determining the verification threshold of each of the fuzzy sets; The degree of membership corresponding to each eigenvalue is compared with the verification threshold to obtain the second verification result.

6. The artificial intelligence-based financial performance assessment and supervision method according to claim 1 is characterized in that: The supervisory analysis of the assessment results and the determination of response strategies based on the supervisory analysis results include: Decompose the financial performance indicators in the historical financial performance data and determine the indicator thresholds of each financial indicator based on the decomposition results; Comparing each financial performance indicator value in the assessment result with the indicator threshold to obtain a comparison result; The comparison results are graded to determine a financial deviation level, and the response strategy is determined based on the financial deviation level.

7. The artificial intelligence-based financial performance assessment and supervision method according to claim 6 is characterized in that: The financial deviation levels include slight deviation, moderate deviation and high deviation.

8. A financial performance assessment and supervision system based on artificial intelligence, characterized by: Applicable to the artificial intelligence-based financial performance assessment and supervision method according to any one of claims 1 to 7; the artificial intelligence-based financial performance assessment and supervision system comprises: An acquisition verification module is configured to perform a primary verification on the structured data and unstructured data of the acquired financial system based on data consistency to obtain a first verification result, and perform a secondary verification on the structured data and the unstructured data based on a preset baseline model to obtain a second verification result; a verification and removal module, configured to screen and remove the structured data and the unstructured data based on the first verification result and the second verification result to obtain target structured data and target unstructured data; A feature extraction module is used to preprocess the target structured data and the target unstructured data to obtain standardized data, and perform multi-dimensional feature extraction on the standardized data to obtain financial time series features, indicator correlation features, and industry benchmark features; An assessment and evaluation module, configured to assess and evaluate financial performance based on the financial time series characteristics, the indicator correlation characteristics, and the industry benchmark characteristics, and obtain an assessment and evaluation result; The supervision and response module is used to supervise and analyze the assessment results and determine the response strategy based on the supervision and analysis results.

9. An electronic device comprising: Memory for storing computer software programs; A processor for reading and executing the computer software program, characterized in that when the processor executes the computer software program, it implements the artificial intelligence-based financial performance appraisal and supervision method as described in any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing a computer software program, wherein: When the computer software program is executed by a processor, the artificial intelligence-based financial performance appraisal and supervision method as described in any one of claims 1 to 7 is implemented.

Images