Enterprise operation risk assessment method based on artificial intelligence analysis

Abstract

The invention discloses an enterprise operation risk assessment method based on artificial intelligence analysis, and relates to the technical field of data analysis, and the method comprises the steps: obtaining a public information data set and an enterprise feature data set of an enterprise; obtaining daily attendance rate, energy consumption, production rhythm fluctuation rate and equipment use efficiency of the enterprise based on the enterprise feature data set; the public information data set comprises industrial and commercial information, enterprise change information, bidding and tendering information, enterprise registration capital, enterprise annual report, administrative penalty information, operation information, intellectual property information and legal litigation information; establishing a correlation model according to the daily attendance rate and the energy consumption to obtain a correlation value; calculating the matching degree of the production rhythm fluctuation rate and the equipment use efficiency to obtain a target matching value; performing comprehensive calculation on the association value and the target matching value to obtain a risk value; the enterprise operation risk level is determined according to the risk value, a visual risk prompt is provided, the potential operation risk is identified, and the effect of monitoring the enterprise operation risk in real time is achieved.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to an enterprise operation risk assessment method based on artificial intelligence analysis. Background Art

[0002] Financial institutions have increased their loan support for small and medium-sized enterprises, but the business information of small and medium-sized enterprises is not open and transparent, which brings challenges to financial institutions in assessing credit and mastering business information. In order to ensure the safety of loans, financial institutions obtain corporate information through multiple channels, including financial reports, tax bills, bank statements, electricity bills, public information and field visits. However, these channels have problems such as delayed, inaccurate, untrue and untimely information, which limits the ability of account managers to manage post-loan customers, increases labor costs, and once the business operation is at risk, it is difficult for financial institutions to obtain information in time, resulting in huge loan losses. It is worth noting that corporate electricity consumption information is an important indicator that can reflect the business status of the enterprise. By monitoring and analyzing the electricity consumption of enterprises, it can help financial institutions to grasp the business status of enterprises more timely and accurately and reduce loan risks.

[0003] The Chinese invention patent with application number 202110395829.6 provides a business risk assessment method and system based on artificial intelligence analysis. The electricity consumption characteristic data, surveillance video information of the office area and personnel clock-in information are input into a pre-trained business risk analysis model to obtain a risk prediction value; a corresponding report signal is issued based on the risk prediction value.

[0004] However, in the prior art, when conducting risk prediction, although various data are obtained for risk analysis, the information content on the surface of the data is used during the analysis, the evaluation dimension is single, and the intrinsic connection between multiple data is not obtained in the risk assessment model, which reduces the accuracy of risk prediction and the sophistication of risk assessment. Summary of the invention

[0005] This application solves the problem in the prior art that the single assessment dimension reduces the accuracy of risk prediction by providing a business risk assessment method based on artificial intelligence analysis, and achieves the technical effect of providing accurate risk prediction.

[0006] The present application provides a method for assessing business operation risks based on artificial intelligence analysis, the method comprising: S100: Obtaining a public information data set and an enterprise characteristic data set of an enterprise; obtaining the daily attendance rate, energy consumption, production rhythm fluctuation rate and equipment utilization efficiency of the enterprise based on the enterprise characteristic data set; S200: establishing a correlation model according to the daily attendance rate and energy consumption to obtain a correlation value; Calculate the matching degree between production rhythm fluctuation rate and equipment utilization efficiency to obtain the target matching value; The risk value is obtained by comprehensively calculating the association value and the target matching value; S300: Determine the enterprise operation risk level according to the risk value, and obtain the public risk value-added according to the public information data set; adjust the enterprise operation risk level according to the public risk value-added, and determine a corresponding reporting signal according to the adjusted enterprise operation risk level.

[0007] Further, in step S200, a correlation value is obtained, and a calculation formula for calculating the correlation value based on the attendance rate decrease rate and the energy consumption change is as follows:

[0008] in, is the associated value, represents the correlation coefficient between attendance rate and energy consumption, is the attendance decline rate, is the change in energy consumption; Get the target matching value, including:

[0009] in, represents the matching value at time t, represents the rate of change of equipment efficiency at time t, It represents the rate of change of production rhythm at time t; The formula for calculating the target match value is as follows:

[0010] in, is the target match value, is the total number of selected moments, Represents the matching value at the i-th time t.

[0011] Furthermore, the enterprise characteristic data set includes power consumption characteristic data, equipment production data, monitoring video data and personnel clock-in data; energy consumption is calculated based on the power consumption characteristic data; The equipment production data includes equipment operation data and production task allocation data, and the original production rhythm and original equipment utilization rate are calculated based on the equipment operation data and the production task allocation data; According to the monitoring video data, the time series data of personnel flow and equipment usage status are obtained to generate the personnel flow feature sequence and equipment usage feature sequence; Based on the personnel flow characteristic sequence and equipment use characteristic sequence, the original production rhythm and original equipment use rate are adjusted to obtain the production rhythm fluctuation rate and equipment use efficiency; Based on the personnel punch-in data, the punch-in records are obtained and the daily attendance rate is calculated.

[0012] Furthermore, the method further comprises: S210: Obtain production process data, and divide the production cycle into multiple production stages according to key event nodes and fixed time windows; S220: Inputting characteristic parameters of each production stage into the stage performance evaluation model to obtain corresponding stage state values; S230: Based on the risk value and the stage status value, generate a deviation sequence, and then obtain a deviation impact factor; based on the deviation impact factor, correct the risk value to obtain a corrected risk value; and execute step S300 according to the corrected risk value.

[0013] Among them, the production process data includes: equipment operating status, order switching records, quality inspection completion time, personnel attendance records and equipment energy consumption data; the key events include: order switching, equipment startup, equipment shutdown and quality inspection completion; the characteristic parameters of each production stage include: equipment idling rate, personnel configuration deviation, equipment utilization rate, production quality rate and energy consumption.

[0014] Furthermore, the stage performance evaluation model includes: collecting and organizing characteristic parameter data in the production process, preprocessing the data, dividing the collected data into a training set, a validation set and a test set, selecting a linear kernel function, and setting a penalty parameter; using the training set data for training, and adjusting the model parameters through a sequence minimum optimization algorithm; using the validation set for verification, evaluating the performance, and adjusting the model parameters according to the verification results; using the test set data for testing, optimizing according to the test results, and obtaining the stage performance evaluation model after the optimization is completed; The preprocessing includes data cleaning, processing of missing values ​​and outliers, and data standardization.

[0015] Further, in step S230, it includes: calculating the difference between the risk value and the stage status value as a deviation; arranging the deviations of each stage in chronological order to form a deviation sequence; calculating the mean and standard deviation of the deviation sequence; calculating the relative deviation of each deviation from the mean of the deviation sequence; obtaining the deviation influence factor based on the relative deviation and the standard deviation; setting a corresponding correction coefficient based on the deviation influence factor, and obtaining a corrected risk value based on the risk value and the correction coefficient.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages: By analyzing multi-dimensional data, establishing a correlation model between attendance rate and energy consumption, and calculating the matching degree between production rhythm and equipment utilization efficiency, the effect of identifying potential operating risks is achieved; by comprehensively calculating the correlation value and matching value, the risk value is obtained, and the enterprise's operating risk level is determined based on the risk value, providing intuitive risk warnings, and achieving the effect of real-time monitoring of enterprise operating risks. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flowchart of a method for assessing business risks based on artificial intelligence analysis in an embodiment of the present invention. DETAILED DESCRIPTION

[0018] To facilitate the understanding of the present invention, the present application will be described more comprehensively below with reference to the relevant drawings; the drawings show preferred embodiments of the present invention, but the present invention can be implemented in many different forms and is not limited to the embodiments described herein; on the contrary, the purpose of providing these embodiments is to enable a more thorough and comprehensive understanding of the disclosed content of the present invention.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by technicians in the technical field to which the present invention belongs; the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention; the term "and / or" used herein includes any and all combinations of one or more related listed items.

[0020] Embodiment 1: Figure 1 As shown, a method for assessing business operation risks based on artificial intelligence analysis comprises: S100: Obtaining a public information data set and an enterprise characteristic data set of an enterprise; obtaining the daily attendance rate, energy consumption, production rhythm fluctuation rate and equipment utilization efficiency of the enterprise based on the enterprise characteristic data set; Specifically, the power consumption waveform and the number of connected devices in the power consumption characteristic data are obtained, the key characteristic points are extracted, and a power consumption characteristic sequence is formed, and the energy consumption is further calculated; the energy consumption requires obtaining the power and usage time of the corresponding equipment, and the energy consumption is obtained according to the product of power and time.

[0021] The equipment production data includes equipment operation data and production task allocation data, and the original production rhythm and original equipment utilization rate are calculated based on the equipment operation data and the production task allocation data; Based on the monitoring video data, the time series data of personnel flow and equipment usage status are obtained to generate personnel flow feature sequence and equipment usage feature sequence; based on the personnel flow feature sequence, the actual working time records and corresponding attendance records of the personnel are obtained, and based on the equipment usage feature sequence, the actual usage time records of the equipment are obtained.

[0022] The original production rhythm and the original equipment utilization rate are adjusted based on the personnel flow characteristic sequence and the equipment usage characteristic sequence to obtain the production rhythm fluctuation rate and the equipment utilization efficiency; the original production rhythm and the original equipment utilization rate are calculated based on the equipment operation data and the production task allocation data to obtain the expected production rhythm and equipment utilization rate, while the actual production rhythm and equipment utilization rate are obtained based on the personnel flow characteristic sequence and the equipment usage characteristic sequence. The adjustment is to obtain the corresponding fluctuation rate and utilization efficiency according to the actual and expected work arrangements, which requires data statistical comparison and analysis and calculation based on actual conditions. This application will not go into details here.

[0023] Based on the personnel punch-in data, the punch-in records are obtained and the daily attendance rate is calculated.

[0024] In some embodiments, in step S100, the power consumption data is obtained based on the data acquisition device to generate a waveform. The waveform is marked with time points and key feature points are extracted, and the key feature points include but are not limited to: power value, current value and voltage value. The number of connected devices is determined according to the power value, and the frequency value is calculated in combination with the current value and the voltage value. The waveform is analyzed by Fourier transform to obtain the harmonic value. The Fourier transform analysis method is a well-known prior art, and this application will not be described in detail here. The phase angle change is determined according to the harmonic value, and the characteristic points in the waveform are identified. The time point, power value, current value, voltage value, frequency value, harmonic value, and phase angle of the characteristic point are normalized. The normalized characteristic points are classified by a clustering algorithm to obtain the device operation mode. A power consumption feature sequence is established according to the device operation mode, and the sequence contains time points, power values, current values, voltage values, frequency values, harmonic values, and phase angles. The power consumption feature sequence is modeled by a time series analysis method to obtain a device power consumption prediction model. The power consumption feature sequence is input into the prediction model to output the device power consumption trend. According to the power consumption trend of the equipment, the power consumption anomaly is judged, and the power consumption feature sequence is analyzed by the association rule mining algorithm to obtain the power consumption behavior pattern of the equipment and establish the power consumption behavior feature library. The energy consumption is calculated based on the power and time in the power consumption feature data.

[0025] Specifically, when obtaining power consumption characteristic data, the user's power consumption waveform data is first collected through the smart meter. For example, the power consumption waveform of a user on a typical working day shows that the power consumption is 5 kWh from 8 am to 10 am, 8 kWh from 2 pm to 4 pm, and 2 kWh from 8 pm to 10 pm. Through time series analysis, these data can extract key characteristic points such as peak, valley and average power consumption. Next, the user's power consumption devices are connected through the Internet of Things technology, and the number of connected devices is counted. For example, a household has 15 devices, including air conditioners, refrigerators, washing machines, etc. By analyzing the power consumption patterns of these devices, the power consumption characteristics of different devices can be further identified. For example, the power consumption of air conditioners in summer is significantly higher than that in other seasons, while the power consumption of refrigerators is relatively stable. Based on these data, clustering algorithms (such as K-means clustering) are used to correlate the power consumption waveform and the number of devices to form a power consumption characteristic sequence. For example, by clustering the user's power consumption waveform data with the number of devices, it is found that the power consumption waveform of users with less than 10 devices is relatively flat, while the power consumption waveform of users with more than 15 devices shows obvious fluctuations. Through these analyses, the user's power consumption feature sequence can be generated to provide data support for subsequent power consumption behavior prediction and load management.

[0026] In some embodiments, for surveillance videos, a target detection algorithm is used to identify personnel and equipment in video frames to obtain personnel location and equipment status information. According to the timestamp marking of the personnel location and equipment status of each frame, time series data of personnel flow and equipment use are generated. The time series data is divided according to the preset time window, and the number of personnel, equipment status, flow direction, regional location, equipment type, flow speed, residence time, frequency of use, personnel density, equipment energy consumption, and flow volume characteristics in each time window are counted. The statistical features are screened using a feature selection algorithm to retain features with high correlation with personnel flow and equipment use. Based on the screened features, a clustering algorithm is used to classify the personnel flow pattern to determine the feature distribution of different personnel flow patterns. According to the equipment use status characteristics, a classification algorithm is used to determine the normal use, abnormal use, and fault state of the equipment. Combined with the classification results of the personnel flow pattern and the judgment results of the equipment use status, a feature sequence of personnel flow and equipment use is generated.

[0027] Get the personnel list and schedule, extract the personnel information and work arrangements; at the same time, calculate the attendance rate and absence number at each time point based on the punch-in records and attendance sheet. Combine the task table and allocation table to determine the task volume and completion volume of each person; based on statistical analysis, calculate the correlation between attendance rate and task completion volume, and group the personnel according to attendance rate and task completion volume to obtain the personnel classification results. Use regression analysis to predict the task allocation of personnel in different categories and generate an optimized allocation table. Adjust the scheduling table and work arrangements based on the optimized allocation table and actual completion volume.

[0028] For example, suppose an employee's clock-in record on October 1, 2024 shows that he clocked in at 8 a.m. and clocked out at 5 p.m., while the scheduling data shows that the employee's work schedule is from 8 a.m. to 5 p.m. By comparison, it can be confirmed that the employee's attendance on that day was normal. Next, calculate the daily attendance rate. Assuming that there are 50 employees in the department on that day, 45 of them are normal, then the attendance rate is 90%.

[0029] S200: Establish a correlation model based on the daily attendance rate and energy consumption to obtain a correlation value; calculate the matching degree between the production rhythm fluctuation rate and the equipment utilization efficiency to obtain a target matching value; and comprehensively calculate the correlation value and the target matching value to obtain a risk value.

[0030] In some embodiments, in step S200, the attendance rate and energy consumption in the historical data are obtained. A model for associating the attendance rate with the energy consumption is established by a linear regression algorithm. The decline rate between the current attendance rate and the historical average attendance rate is calculated. The change between the current energy consumption and the historical average energy consumption is obtained. If the attendance rate decline rate is greater than a preset threshold and the energy consumption change is not less than zero, it is determined that there is an operating risk. The association value is calculated based on the product of the attendance rate decline rate and the energy consumption change.

[0031] For example, the monthly attendance rate and energy consumption of a factory in the past year. The monthly attendance rate of the factory fluctuates between 80% and 95%, and the energy consumption varies between 5000 and 8000 kWh. Through linear regression analysis, the relationship model between attendance rate and energy consumption can be fitted. Assuming that the fitting result is energy consumption = 1000×attendance rate + 3000, the R² value is 85, indicating that the model has a high explanatory power. Set the risk judgment standard according to the actual situation. If the attendance rate drops by more than 5% and the energy consumption remains unchanged or increases, it is judged that there is an operating risk. For example, the attendance rate drops from 90% to 83% in a certain month, and the energy consumption is still 7000 kWh. According to the model prediction, the energy consumption should be 8300 kWh. The actual energy consumption is lower than the predicted value, but because the attendance rate drops by more than 5% and the energy consumption does not decrease, the month is automatically marked as having an operating risk. Through this association model, real-time monitoring of operating conditions can be achieved and timely measures can be taken to reduce risks.

[0032] In some embodiments, the loss function for building the association model is set to:

[0033] Where L is the loss function; and are the linear regression coefficients, representing the intercept and slope; is the attendance rate of the i-th sample; is the energy consumption of the i-th sample; is the total number of samples. Samples refer to the attendance rate and energy consumption obtained within the set time interval. For example, if the time interval is set to 24 hours, the attendance rate and energy consumption are counted every 24 hours to form the sample of experimental statistics; The linear relationship model between attendance rate and energy consumption is established as follows:

[0034] Solving by least squares method and , so that the loss function Minimize, the derivation formula is:

[0035] In some embodiments, historical data is collected, including attendance And the corresponding energy consumption , calculate the historical average attendance rate and historical average energy consumption :

[0036] In some embodiments, the calculation formula for calculating the attendance drop is as follows:

[0037] in, is the attendance decline rate, which is used to quantify the degree of attendance abnormality; It is the historical average attendance rate, reflecting the attendance level during normal operation of the enterprise; is the current attendance rate, which needs to be monitored and obtained in real time; Based on current attendance , calculate the expected energy consumption , calculated using the linear relationship model as follows:

[0038] The calculation formula for calculating the change in energy consumption is as follows:

[0039] in, is the change in energy consumption, is the historical average energy consumption (kW h), is the current energy consumption (kW·h); In some embodiments, the correlation coefficient between attendance rate and energy consumption is calculated as follows:

[0040] in, represents the correlation coefficient between attendance rate and energy consumption, represents the attendance rate of the i-th sample, represents the energy consumption of the i-th sample, is the historical average attendance rate, is the historical average energy consumption, and n represents the total number of samples.

[0041] In some embodiments, the calculation formula for calculating the correlation value based on the attendance rate decrease rate and the energy consumption change amount is as follows:

[0042] in, is the associated value, is the correlation coefficient.

[0043] In some embodiments, in step S200, production task allocation data is obtained, and equipment information and production task volume in task allocation are extracted. Based on the equipment information and production task volume, the actual efficiency of the equipment within the preset time is calculated. By comparing the actual efficiency of the equipment with the preset threshold, it is determined whether the equipment utilization efficiency is lower than the preset threshold. If the equipment utilization efficiency is lower than the preset threshold, the production rhythm fluctuation rate is marked as abnormal, and an abnormality report is generated. According to the abnormality report, the matching degree between the production rhythm fluctuation rate and the equipment utilization efficiency is analyzed to obtain a matching value. For the unmatched links, an optimization algorithm is used to reallocate production tasks and adjust the equipment utilization rhythm. The equipment utilization efficiency is recalculated through the adjusted production task allocation data to confirm whether it reaches the preset threshold.

[0044] Specifically, in the process of production task allocation, the production capacity of each device is first determined through data analysis. For example, device A can produce 100 products per hour, and device B can produce 80 products per hour. According to the urgency of the production task and the type of product, the system uses a weighted algorithm to allocate tasks, and allocates urgent tasks to device A first, and ordinary tasks to device B. The monitoring of production rhythm fluctuation rate collects equipment operation data in real time. For example, the actual production speed of device A is 95 pieces / hour, and the actual production speed of device B is 75 pieces / hour. The system presets the equipment utilization efficiency threshold to 90%, that is, the minimum production speed of device A should be 90 pieces / hour, and the minimum production speed of device B should be 72 pieces / hour. By calculating that the utilization efficiency of device A is 95% and the utilization efficiency of device B is 975%, both of which are higher than the preset threshold, the production rhythm fluctuation rate is normal. If the actual production speed of device B drops to 70 pieces / hour, its utilization efficiency is 85%, which is lower than the preset threshold. The system will determine that the production rhythm fluctuation rate is abnormal, and automatically adjust the task allocation, transferring some tasks to device A to ensure overall production efficiency.

[0045] In some embodiments, the calculation formula for equipment utilization efficiency is as follows:

[0046] in, Indicates the efficiency of equipment use. Indicates the actual production task volume, Indicates rated production capacity, It represents the preset time, and the length of the preset time must be greater than the length of the preset time interval i.

[0047] In some embodiments, the calculation formula of the production rhythm fluctuation rate is as follows:

[0048] in, represents the production rhythm fluctuation rate, represents the production time of the jth task, represents the average production time, Indicates the total task volume. Production time refers to the time required to complete a single task, which is used to calculate the production inoculation fluctuation rate. The production time of each task is different and needs to be counted based on the actual task volume.

[0049] In some embodiments, the calculation formula of the matching value is as follows:

[0050] in, represents the matching value at time t, represents the rate of change of equipment efficiency at time t, Indicates the rate of change of production rhythm at time t; where time t refers to the start or end time point of the production task, or it can be any moment within the preset time, and is an instantaneous time point used to calculate the matching degree.

[0051] In the T time, multiple matching values ​​corresponding to time t will be obtained, and the average matching value will be calculated as the target matching value. The selected time should be greater than 5. The calculation formula for calculating the target matching value is as follows:

[0052] in, is the target match value, is the total number of selected moments, Represents the matching value at the i-th time t.

[0053] In some embodiments, the associated value and the target matching value are comprehensively calculated to obtain the risk value; the risk value is calculated using a weighted average method according to the weight and standardized value of each indicator, and the calculated associated value and target matching value are first standardized to be in the same dimension; the weight distribution is determined according to the importance of each indicator in the risk assessment, and the specific method includes pre-setting according to field experts or using a hierarchical analysis method, by building a hierarchical model, using a comparison matrix to determine the relative importance of each indicator, and then calculating the weight. The weight of each indicator can also be directly given based on historical data or experimental data. The corresponding method is selected according to the actual situation, and this application does not make specific restrictions here. According to the standardized associated value and target matching value and the corresponding weight, a weighted average is calculated to obtain a risk value.

[0054] In some embodiments, S300: determine the enterprise operating risk level according to the risk value, and obtain a public risk value-added according to the public information data set; adjust the enterprise operating risk level according to the public risk value-added, and determine a corresponding reporting signal according to the adjusted enterprise operating risk level.

[0055] In some embodiments, the public information dataset includes but is not limited to: Industrial and commercial information: enterprise registration information, business scope, shareholder structure, legal representative and branches, etc.; Enterprise change information: equity change, legal representative change and business scope change, etc. At the same time, detailed records of changes in key information such as enterprise name, registered address, business scope, etc., including the time of change, specific content before and after the change and the reason for the change, will help analyze the historical evolution and stability of the enterprise; Bidding information: bidding projects participated by the enterprise, winning bids and project amounts, etc., to obtain the reasons why the enterprise did not win the bid, the list of participating enterprises and the advantages of the winning enterprise, so as to more comprehensively evaluate the market competitiveness and project execution capabilities of the enterprise; Enterprise registered capital: the initial registered capital of the enterprise and the subsequent capital increase, plus the enterprise's financing information, including financing rounds, financing amount, investors and investment conditions, etc., as well as the enterprise's external investment, such as investment amount, investment objects and investment fields, etc., which helps to understand the financial strength and future development potential of the enterprise; Annual report of the enterprise: annual operating conditions and financial status, etc.; Administrative penalty information: records of administrative penalties imposed on the enterprise; Business information: import and export credit, recruitment information, tax rating, general taxpayer qualification, financing information, enterprise supplier information, enterprise customer information, qualification certification, administrative license and WeChat public account. Import and export credit records the international trade reputation of the enterprise, recruitment information reflects the human resource needs and recruitment situation of the enterprise, and tax rating reflects the tax compliance of the enterprise, which can more comprehensively reflect the market reputation, human resource status and tax compliance of the enterprise; Intellectual property information: patent information, software copyright, trademark information, copyright and domain name information; Legal litigation information: litigation cases involving the company and the results of the judgments.

[0056] The public information data set also includes tax data, invoice data, provident fund and social security data, and multiple loan data. Specifically, tax data includes the company's tax declaration records, tax payment status, and tax audit results; invoice data includes the company's invoicing records, invoice amount, invoice type (such as special value-added tax invoices, ordinary invoices, etc.) and invoicing objects; provident fund and social security data includes the amount, proportion, and payment record of provident funds and social security paid by the company for its employees; multiple loan data reflects the company's borrowing and lending situation in multiple financial institutions, including the loan amount, loan term, and repayment status.

[0057] Among them, tax data can reflect the tax compliance, financial health and profitability of an enterprise. Abnormal tax data may indicate that the enterprise has engaged in tax evasion or tax avoidance, or faces greater tax risks, which in turn affects the enterprise's reputation and financial status. Invoice data is an important record of a company's financial activities, which can reflect the company's transaction scale, transaction objects and transaction frequency. By analyzing invoice data, we can evaluate the authenticity of the company's business, the compliance of revenue recognition and potential tax risks. Provident fund and social security data can reflect the human resource management status and employee welfare benefits of an enterprise. Reasonable provident fund and social security payment records show that an enterprise complies with labor laws and regulations and cares about the welfare of employees, which helps to enhance the corporate image and employee loyalty. On the contrary, defaulting on or underpaying provident fund and social security may lead to labor disputes and increase the legal risks and financial burden of the enterprise. Multiple borrowing data can reveal a company's financing activities and debt status. Frequent multiple borrowing may indicate that the company is facing greater financial pressure or financial risks, and excessive borrowing may cause the company to fall into a debt crisis. In addition, multiple borrowing may also affect the company's credit rating and financing capabilities, and increase future financing costs. By analyzing multiple borrowing data, the company's debt repayment ability, financing strategy and risk tolerance can be evaluated.

[0058] For the public information data set, we should focus on the business risks and judicial risks faced by enterprises. The business risks include: serious violations of law and trust, environmental penalties, equity freezes, administrative penalties, major tax violations, tax arrears information, abnormal operations, equity pledges and movable property mortgages; the judicial risks include: judicial auctions, restrictions on high consumption, dishonest debtors, debtors, final cases, bankruptcy cases, case filing information, court opening announcements and court announcements.

[0059] The public information dataset can be obtained through official channels or third-party data service platforms, and this application does not impose any specific restrictions on this.

[0060] Conduct risk analysis on the obtained public information data set, including: Business information analysis: Check whether the business scope is consistent with the actual business, whether the shareholder structure is stable, whether the legal representative has a bad record, etc. Change information analysis: Frequent equity changes and legal representative changes are used to discover the possibility of unstable business operations or potential risks; Bidding information analysis: Frequent participation in bidding but low bid winning rate may indicate that the company is at a disadvantage in market competition; A sharp drop in the amount of the winning project may mean that the company's business is shrinking; Registered capital analysis: Too low registered capital may affect the company's reputation and business development capabilities; A sharp reduction in registered capital may indicate that the company is short of funds; Annual report analysis: Poor financial conditions and operating losses may indicate that the company has financial risks; Administrative penalties and legal proceedings analysis: Administrative penalties and legal proceedings may directly affect the company's reputation and normal operations. Set corresponding risk ratios for different analysis contents, such as serious, medium, and general risk ratios, and combine all analysis contents and corresponding risk ratios to form a comprehensive public risk value-added. When conducting risk analysis, you can use preset risk analysis rules or models for automatic analysis, or you can let assessors perform manual analysis. The specific needs need to be adjusted and set according to actual conditions, and this application does not make specific restrictions here.

[0061] The enterprise operating risk level is adjusted according to the public risk appreciation, specifically including pre-setting the mapping relationship between the public risk appreciation and the enterprise operating risk level, setting multiple threshold ranges, and each threshold range corresponds to a reference coefficient, for example: when the public risk appreciation is between [0,10), the corresponding reference coefficient is 1; when the public risk appreciation is between [10,20), the corresponding reference coefficient is 1.2; when the public risk appreciation is between [20,30), the corresponding reference coefficient is 1.5; when the public risk appreciation is greater than 30, the corresponding reference coefficient is 2. Compare the actual calculated public risk appreciation with the preset threshold range, determine the range it belongs to, and then obtain the corresponding reference coefficient.

[0062] The original business risk calculated by the risk assessment method described above is adjusted using the reference coefficient. For example, if the original risk level is medium risk and the reference coefficient is 1.2, the adjusted risk level is high risk. The corresponding reporting signal is determined based on the adjusted business risk level.

[0063] In some embodiments, the risk level classification is to classify the risk into different levels, such as low risk, medium risk, high risk, etc., based on the calculated risk value. Determine the classification criteria for each risk level based on business needs and risk tolerance; compare the calculated risk value with the level standard to determine the risk level; formulate corresponding risk prevention and control measures for different risk levels.

[0064] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: This application analyzes multi-dimensional data, establishes a correlation model between attendance rate and energy consumption, and calculates the matching degree between production rhythm and equipment utilization efficiency, thereby achieving the effect of identifying potential business risks; by comprehensively calculating the correlation value and matching value, the risk value is obtained, and the business risk level of the enterprise is determined according to the risk value, providing intuitive risk warnings and achieving the effect of real-time monitoring of business risks.

[0065] Embodiment 2: In the above embodiment, the risk value is calculated by the association model and the matching model, but the impact of the performance changes at each stage of the production process on the risk value is not considered. This embodiment further modifies the risk value.

[0066] In some embodiments, the method further comprises: S210: Obtain production process data, and divide the production cycle into multiple production stages according to key event nodes and fixed time windows.

[0067] In some embodiments, production process data can be obtained from the enterprise's production management system and equipment monitoring system, including but not limited to equipment operating status, order switching records, quality inspection completion time, personnel attendance records, equipment energy consumption data, etc.

[0068] Identify key time nodes, and identify key events such as order switching, equipment startup, equipment shutdown, and quality inspection completion according to the production process; pre-set every 30 minutes as a time window as the basis for the division of production stages. If a key event node is detected within a fixed time window, the division of the time window is readjusted based on the time of event occurrence; for example, if a certain device is started at 10:15, and the current time window is 10:00-10:30, the time window is adjusted to two windows of 10:00-10:15 and 10:15-10:30. Combine event triggering with time windows to dynamically generate production stage boundaries in a hybrid mode. Suppose a production line starts production at 8:00, equipment A starts at 10:15, quality inspection is completed at 11:00, and equipment A is shut down at 12:00. The production stage can be divided into: 8:00-10:15 (before equipment A is started), 10:15-11:00 (before quality inspection is completed), and 11:00-12:00 (before equipment A is shut down).

[0069] S220: Input the characteristic parameters of each production stage into the stage performance evaluation model to obtain the corresponding stage state value.

[0070] In some embodiments, the characteristic parameters of each production stage include but are not limited to: equipment idling rate, which refers to the ratio of the time when the equipment is not producing but consuming electricity to the total time of the stage, and the idling time is calculated based on the equipment operating status data in the above embodiments; personnel allocation deviation, which refers to the difference between the actual number of people on duty and the theoretical number of people required, and is calculated based on the daily attendance rate and the personnel flow characteristic sequence; equipment utilization rate, which refers to the ratio of the actual working time of the equipment to the total time of the stage; production quality rate, which refers to the ratio of the number of products that pass the quality inspection within the stage to the total number of quality inspections; energy consumption efficiency, which refers to the amount of energy consumed to produce a unit product within the stage.

[0071] The stage performance evaluation model includes: collecting and collating characteristic parameter data in the production process, preprocessing the data, dividing the collected data into a training set, a validation set and a test set, selecting a linear kernel function, and setting a penalty parameter; using the training set data for training, adjusting the model parameters through a sequence minimum optimization algorithm; using the validation set for verification, evaluating the performance, and adjusting the model parameters according to the verification results; using the test set data for testing, optimizing according to the test results, and obtaining the stage performance evaluation model after the optimization is completed; The preprocessing includes data cleaning, processing of missing values ​​and outliers, and data standardization.

[0072] In some embodiments, the characteristic parameters in the new production data are input into the trained stage efficiency evaluation model, and prediction is performed based on the input characteristic parameters to output the stage state value. The stage state value can be a continuous value (such as a real number between 0 and 1), and the calculated stage state value is analyzed to evaluate the efficiency level of the production stage.

[0073] The stage performance evaluation model maps the input feature parameters to a high-dimensional space through a kernel function. In this high-dimensional space, the linear separability between data points is stronger, which is convenient for classification or regression. In the high-dimensional space, SVM looks for an optimal hyperplane to separate data points of different categories. This hyperplane can be solved by an optimization algorithm (such as a sequence minimum optimization algorithm). For a given input feature parameter, the stage performance evaluation model maps it to a high-dimensional space and calculates its distance to the optimal hyperplane. This distance can be used to represent the stage state value. If the distance is positive, it means that the production stage belongs to the efficient category; if the distance is negative, it means that the production stage belongs to the inefficient or invalid category; if the distance is close to zero, it means that the production stage is in a boundary state. This embodiment only provides a method for model training. In practical applications, the functions and corresponding parameters of the model need to be adjusted according to actual conditions, and this application does not make specific restrictions here.

[0074] S230: Based on the risk value and the stage status value, generate a deviation sequence, and then obtain a deviation impact factor; based on the deviation impact factor, correct the risk value to obtain a corrected risk value; and execute step S300 according to the corrected risk value.

[0075] In some embodiments, for each production stage, its risk value and stage status value are calculated; the difference between the risk value and the stage status value is calculated as the deviation; the deviations of each stage are arranged in chronological order to form a deviation sequence. The mean and standard deviation of the deviation sequence are calculated; for each deviation, the relative deviation of each deviation from the mean of the deviation sequence is calculated; based on the relative deviation and standard deviation, the deviation impact factor is obtained, and the calculation formula is as follows:

[0076] in, represents the deviation influencing factor; is the absolute value of the relative deviation, Indicates deviation, represents the mean value of the deviation series; Expresses the standard deviation of the deviation.

[0077] In some embodiments, a corresponding correction coefficient is set according to the deviation impact factor, and a corrected risk value is obtained based on the risk value and the correction coefficient. Specifically, the range of the correction coefficient should be set according to business needs and risk tolerance, and historical risk event data, including risk values, deviation impact factors, etc., are collected. The data is analyzed to calculate statistical indicators such as the mean, standard deviation, maximum, and minimum values ​​of the deviation impact factor. According to the statistical indicators of the deviation impact factor, correction coefficients of different ranges are set, and multiple threshold ranges are set, including [a, b), [b, c), where a represents the difference between the mean and the standard deviation, b represents the mean, and c represents the sum of the mean and the standard deviation. For example: when the deviation impact factor is less than a, the correction coefficient is set to 0.8 (indicating that the risk is low and the risk value is appropriately reduced); When the deviation impact factor is between [a, b), the correction factor is 1 (indicating that the risk is moderate and the risk value is not adjusted); When the deviation impact factor is between [b, c), the correction factor is 1.2 (indicating a higher risk, and the risk value should be appropriately increased); When the deviation impact factor is not less than c, the correction factor is 1.5 (indicating that the risk is very high and the risk value is significantly increased). The above correction factor is only an example and should be adjusted according to the specific situation in actual application. The original risk value is corrected using the set correction factor to obtain the corrected risk value. The corrected risk value is the product of the risk value and the correction factor.

[0078] In this embodiment, a deviation sequence is generated based on the risk value and the stage status value, and a deviation impact factor is obtained according to the deviation sequence. The risk value generated in Example 1 is corrected according to the deviation impact factor to obtain a corrected risk value; the enterprise operating risk level is determined according to the corrected risk value, and a corresponding report signal is issued according to the enterprise operating risk level.

[0079] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: This application further corrects the risk value by introducing production stage division, stage efficiency evaluation and deviation impact factors, improves the accuracy and reliability of risk assessment, and achieves the effect of further accurately understanding the business risk status of the enterprise.

[0080] Embodiment 3: In the above embodiment, the business risk of the enterprise is evaluated only based on the current production situation. When accidental events occur, it will affect the correct evaluation and analysis and reduce the accuracy of the evaluation. This embodiment makes further improvements on the basis of the above embodiment.

[0081] Continue to refer Figure 1 , the method further comprises: S400: Obtaining the modified risk values ​​and corresponding stage status values ​​at different times to form time series data, and then constructing a time series model; Based on the time series model, draw a risk evolution map, and set the corresponding benchmark efficiency threshold for each production stage based on the risk evolution map; S500: Based on the benchmark performance threshold, record the number of performance deviations in each production stage, formulate preventive measures based on the number of occurrences and the difference handling mechanism, and then determine the cause of the risk; comprehensively determine the enterprise's operating risk level based on the risk cause and the modified risk value.

[0082] In some embodiments, according to the above embodiments, a large number of modified risk values ​​and corresponding stage status values ​​corresponding to different times are obtained to form time series data, and the time series data is preprocessed, including cleaning and standardizing all data to ensure the consistency and comparability of the data.

[0083] In some embodiments, the time series model construction process is as follows: use LSTM time series analysis technology to model time series data and predict future risk trends. Specifically, based on the time series data, create time series features and generate stage state codes, and the time series features are characteristics of the time series data, including lag features and moving average features. The stage state code converts the stage state value into a numerical value or a unique hot code to facilitate model processing, which is a well-known technical means and will not be described in detail in this application.

[0084] Specifically, the lag feature refers to the use of historical information of time series data to create new features. In time series analysis, the data point at the current moment is often affected by its previous data point. Therefore, by introducing lag features to assist the model in capturing historical dependencies in the data, we must first determine the order of the created lag features. The lag order represents the time unit of the lag. For example, lag 1 means using the data of the previous time point, lag 2 means using the data of the previous two time points, and so on. The generated lag features are added as new columns to the original time series data to form new time series data.

[0085] The moving average feature is a method of smoothing data by calculating the average value of time series data within a certain window, which can reduce the noise in the data and highlight the long-term trend of the data. The window size of the moving average feature needs to be adjusted according to specific needs. A larger window can capture longer-term trends, but may ignore some short-term fluctuations.

[0086] In some embodiments, the model structure is defined, the input layer is used to receive time series data; the LSTM layer sets the number of LSTM units to determine the complexity of the model; the fully connected layer is used to output the predicted value; the loss function, such as the mean square error, is determined; the time series data is divided into a training set and a test set, the training set data is used to train the time series model, and the hyperparameters are adjusted to optimize the model performance; the test set data is used to verify the model effect and evaluate the prediction accuracy. The new time series data is input into the trained time series model to obtain the future risk value and stage status value predicted by the model.

[0087] In some embodiments, the prediction results are visualized to form a risk evolution map, which shows the change of risk value over time and the performance status of different production stages. Define the X-axis as time and the Y-axis as risk value; draw a line graph of the change of risk value over time based on the prediction data; and use different colors or lines to mark the change of status at different stages.

[0088] In some embodiments, new risk value and stage status value data are collected regularly, and the new prediction results are combined with the existing data to form a complete time series data; the risk evolution map is redrawn using the updated data, and the risk evolution map is updated regularly to maintain its timeliness and accuracy.

[0089] In some embodiments, a corresponding benchmark performance threshold is set for each production stage based on the risk evolution map. Specifically, the historical risk value and stage status value are analyzed to determine the performance level range of each production stage under normal operating conditions. The benchmark performance threshold is set for each production stage based on the historical data analysis results and the generated risk evolution map. Preferably, internal changes in the enterprise are also considered, and the benchmark performance threshold is regularly evaluated and adjusted to ensure that it is adapted to current production conditions. The stage status value of each production stage is tracked in real time, and the stage status value is compared with the benchmark performance threshold to identify performance deviations.

[0090] In some embodiments, based on the monitoring system, the number of occurrences of performance deviations in each production stage is recorded, and preventive measures are formed according to the number of occurrences and the difference handling mechanism. The difference handling mechanism refers to taking corresponding corrective and preventive measures for the number and nature of performance deviations in the production stage under a set of predefined rules and processes to ensure that production efficiency is stable within a preset benchmark threshold range. This mechanism is designed to quickly respond to performance deviations, reduce production losses, and improve overall operational efficiency. For example, for one deviation, the preventive measures are set to: automatically adjust equipment parameters, optimize personnel allocation, or adjust production plans to quickly restore performance levels to within the benchmark threshold; for three deviations, the preventive measures are set to: trigger an in-depth detection mechanism, or set up a special team to conduct a root cause analysis of the cause of the deviation, and formulate short-term corrective measures and long-term improvement plans; for five or more deviations, the preventive measures are set to: initiate a comprehensive review mechanism and implement major improvement projects to fundamentally solve the performance deviation problem.

[0091] In some embodiments, the cause of the risk is determined based on the prevention result; the enterprise's operating risk level is determined comprehensively based on the risk cause and the modified risk value, and a corresponding report signal is issued based on the enterprise's operating risk level.

[0092] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: This application constructs a time series model and a risk evolution map to more accurately predict and display the changing trend of business risk over time and the performance status of different production stages; by setting a benchmark performance threshold and monitoring performance deviation, it can achieve real-time monitoring and respond to production efficiency fluctuations; establish a difference processing mechanism to achieve more standardized deviation correction, which helps to quickly find the cause of the risk; comprehensively consider the risk cause and the revised risk value to determine the business risk level of the enterprise, and achieve a more comprehensive and accurate risk assessment.

[0093] Embodiment 4: This embodiment is a further improvement on the basis of the above embodiment.

[0094] When performance deviations occur continuously in a certain production stage, historical data of the same operating conditions are automatically traced back, and the root cause type of abnormal fluctuations is identified by comparing auxiliary factors such as environmental parameters and equipment maintenance records.

[0095] A self-correction mechanism for the deviation sequence is established. In the event of continuous efficiency deviations in the production stage, it is determined whether the deviation sequence generated by the production process data exceeds the preset range. If it exceeds the range, a dynamic alarm mechanism based on the deviation influencing factor is triggered. For example, when the unit energy consumption output ratio of a process deviates from the mean by 15% for three consecutive cycles, the system automatically marks the abnormal status of the process and injects the auxiliary correction coefficient into the risk value calculation process.

[0096] The method further comprises: S510: If the number of occurrences is greater than three times, search the data warehouse for historical data of the same working condition according to the working condition identification of the current production stage; perform correlation analysis based on the current performance deviation data and historical data of the same working condition to identify auxiliary factors; S520: Identify the change range of the current deviation sequence and compare it with the pre-set range, and obtain the correction factor according to the comparison result; S520: Determine the cause of the risk based on the auxiliary factors and the corrective factors.

[0097] The data warehouse is used to store historical data of all production stages, including: environmental parameters, equipment maintenance records, production efficiency indicators and production batches; the working condition identifier is a unique identification code for each production stage; the auxiliary factor refers to an important influencing factor reflecting efficiency deviation; the correction factor is used to reflect the degree of deviation of the deviation sequence.

[0098] In some embodiments, the data warehouse is used to store historical data of all production stages, including but not limited to: environmental parameters, equipment maintenance records, production efficiency indicators, and production batches, etc. The data in the data warehouse is set to be indexed according to timestamps and working condition identifiers; a unique and unique working condition identifier is established for each production stage or process. The working condition identifier is a unique identification code for each production stage, including equipment ID, production formula ID, operator ID, etc., to ensure that the same or similar working conditions can be accurately identified. Set the search time. To ensure the availability of historical data, the search time should not exceed one year, and historical working condition data within one year should be searched. For example, when a certain equipment on a production line has continuous performance deviations, the production data of the equipment under the same formula in the past three months is automatically searched in the data warehouse according to the equipment ID and production formula ID, including production capacity, energy consumption, equipment failure rate, etc., and analyzed.

[0099] In some embodiments, correlation analysis is performed based on current performance deviation data and historical data of the same operating condition, auxiliary factors are identified, and the correlation between current performance deviation data and historical data of the same operating condition is discovered based on an association rule mining (Apriori) algorithm.

[0100] Specifically, according to the current performance deviation data and historical data of the same working condition, including environmental parameters, equipment maintenance records, operator skill level and other related information; the data is converted into Boolean or numerical data format, and the minimum support and minimum confidence are set. The minimum support represents the minimum frequency of an item set appearing in the data set, which is used to filter frequent item sets. The minimum confidence represents the strength of the association rule, that is, if the antecedent (such as a certain auxiliary factor) appears, the probability that the consequent (such as performance deviation) also appears; Use the Apriori algorithm to generate frequent item sets layer by layer, starting from 1 item set, generate k item sets through combination, and calculate the support of each item set, retain the item sets with support greater than or equal to the minimum support as frequent item sets; repeat the above process until no new frequent item sets can be generated. Generate association rules from frequent item sets, calculate the confidence of each rule, retain the rules with confidence greater than or equal to the minimum confidence, analyze the generated association rules, and find out the auxiliary factors that are significantly related to performance deviation. These auxiliary factors may be single items (such as too high temperature) or item sets (such as too high temperature and abnormal humidity), which need to be determined according to actual conditions. Auxiliary factors include but are not limited to: environmental parameters: such as too high or too low temperature, abnormal humidity, etc.; equipment maintenance records: such as insufficient maintenance frequency, failure to replace key components in time, etc.; operator skill level: such as insufficient training for new operators, poor operating habits, etc.

[0101] In some embodiments, the correction factor is used to reflect the degree of deviation between the change amplitude of the performance deviation sequence and the preset range. It is necessary to quantitatively analyze the deviation sequence, calculate its starting value, maximum value, duration and other key indicators, and compare them with the preset range.

[0102] Specifically, extract the data of the current performance deviation sequence, including the deviation value at each time point, calculate the starting value, maximum value, minimum value, average value, standard deviation and other statistics of the deviation sequence, and calculate the duration of the deviation sequence, that is, the length of time from the beginning to the end of the deviation. According to historical data and experimental data, pre-set the preset range of performance deviation, including normal deviation range, warning deviation range and severe deviation range. Each range can be a specific numerical interval or a threshold based on statistics (such as the mean plus or minus the standard deviation), which is set according to the actual situation. Compare the key indicators of the deviation sequence (such as maximum value, duration) with the preset range, and calculate the relative deviation.

[0103] Determine the preset range, and set different weights in different preset ranges. For example, if it is in the normal deviation range, the weight is 0.5; if it is in the warning deviation range, the weight is 1; if it is in the serious deviation range, the weight is 2; The correction factor is determined based on the weight and relative deviation, and the degree of deviation is determined based on the value of the correction factor and the pre-set correction threshold. The correction threshold needs to be dynamically set based on historical data and experimental data.

[0104] For example, a performance deviation sequence has a maximum value of 10, a duration of 2 hours, a preset normal deviation range of [0, 5], a warning deviation range of (5, 10], a severe deviation range of (10, +∞), and a correction threshold of 0.6.

[0105] Calculate the degree of deviation: The maximum deviation is (10-5) / 5=1 (relative deviation). If it exceeds a certain time threshold (such as 1 hour), it is considered a deviation. The weight is 1, the relative deviation is 1, and the correction factor is 2, which is much larger than the correction threshold. At this time, the degree of deviation is high.

[0106] In some embodiments, the risk cause is determined based on auxiliary factors and corrective factors. Specifically, a combined analysis is performed based on the auxiliary factors and the corrective factors. For example, poor operating habits are combined with high values ​​of the corrective factors to determine that poor operating habits are one of the main risk causes of performance deviations. If other auxiliary factors (such as excessive temperature and insufficient maintenance frequency) also appear frequently in the association rules and correspond to higher values ​​of the corrective factors, they may also be risk causes.

[0107] Based on the above analysis, poor operating habits, high temperature, insufficient maintenance frequency, etc. are identified as possible risk causes. The practical significance and industry knowledge of these auxiliary factors are considered to further verify and confirm the risk causes. For example, poor operating habits may lead to improper equipment operation, high temperature may lead to equipment overheating, and insufficient maintenance frequency may lead to increased equipment wear.

[0108] Based on the risk causes obtained, we can further determine whether the risks are accidental or a persistent problem, and then judge the operating risks of an enterprise as a whole. This solves the problem of determining the operating risks of an enterprise through a single analysis, which leads to the inability to accurately judge the actual operating risks of the enterprise.

[0109] The technical solutions in the above embodiments of the present application have at least the following technical effects or advantages: This application can more accurately identify the root cause type of performance deviation through tracing and correlation analysis of historical data of the same operating conditions; the self-correction mechanism and dynamic alarm mechanism of the deviation sequence can promptly detect and respond to abnormal fluctuations in the production process; and through the combined analysis of corrective factors and auxiliary factors, the cause of the risk can be determined more comprehensively.

[0110] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for assessing business risk based on artificial intelligence analysis, characterized in that: The method comprises: S100: Obtaining a public information data set and an enterprise characteristic data set of an enterprise; obtaining the daily attendance rate, energy consumption, production rhythm fluctuation rate and equipment utilization efficiency of the enterprise based on the enterprise characteristic data set; S200: establishing a correlation model according to the daily attendance rate and energy consumption to obtain a correlation value; Calculate the matching degree between production rhythm fluctuation rate and equipment utilization efficiency to obtain the target matching value; The risk value is obtained by comprehensively calculating the association value and the target matching value; S300: Determine the enterprise operation risk level according to the risk value, and obtain the public risk value-added according to the public information data set; adjust the enterprise operation risk level according to the public risk value-added, and determine a corresponding reporting signal according to the adjusted enterprise operation risk level.

2. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 1, characterized in that: In step S200, a correlation value is obtained. The calculation formula for calculating the correlation value based on the attendance rate decrease rate and the energy consumption change is as follows: in, is the associated value, represents the correlation coefficient between attendance rate and energy consumption, is the attendance decline rate, is the change in energy consumption; Get the target matching value, including: in, represents the matching value at time t, represents the rate of change of equipment efficiency at time t, It represents the rate of change of production rhythm at time t; The formula for calculating the target match value is as follows: in, is the target match value, is the total number of selected moments, Represents the matching value at the i-th time t.

3. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 1, characterized in that: The enterprise characteristic data set includes power consumption characteristic data, equipment production data, monitoring video data and personnel punch-in data; energy consumption is calculated based on the power consumption characteristic data; The equipment production data includes equipment operation data and production task allocation data, and the original production rhythm and original equipment utilization rate are calculated based on the equipment operation data and the production task allocation data; According to the monitoring video data, the time series data of personnel flow and equipment usage status are obtained to generate the personnel flow feature sequence and equipment usage feature sequence; Based on the personnel flow characteristic sequence and equipment use characteristic sequence, the original production rhythm and original equipment use rate are adjusted to obtain the production rhythm fluctuation rate and equipment use efficiency; Based on the personnel punch-in data, the punch-in records are obtained and the daily attendance rate is calculated.

4. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 3, characterized in that: The method further comprises: S210: Obtain production process data, and divide the production cycle into multiple production stages according to key event nodes and fixed time windows; S220: Inputting characteristic parameters of each production stage into the stage performance evaluation model to obtain corresponding stage state values; S230: Based on the risk value and the stage status value, generate a deviation sequence, and then obtain a deviation impact factor; based on the deviation impact factor, modify the risk value to obtain a modified risk value; and execute step S300 according to the modified risk value; Among them, the production process data includes: equipment operating status, order switching records, quality inspection completion time, personnel attendance records and equipment energy consumption data; the key events include: order switching, equipment startup, equipment shutdown and quality inspection completion; the characteristic parameters of each production stage include: equipment idling rate, personnel configuration deviation, equipment utilization rate, production quality rate and energy consumption.

5. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 4, characterized in that: The stage performance evaluation model includes: collecting and collating characteristic parameter data in the production process, preprocessing the data, dividing the collected data into a training set, a validation set and a test set, selecting a linear kernel function, and setting a penalty parameter; using the training set data for training, adjusting the model parameters through a sequence minimum optimization algorithm; using the validation set for verification, evaluating the performance, and adjusting the model parameters according to the verification results; using the test set data for testing, optimizing according to the test results, and obtaining the stage performance evaluation model after the optimization is completed; The preprocessing includes data cleaning, processing of missing values ​​and outliers, and data standardization.

6. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 4, characterized in that: In step S230, it includes: calculating the difference between the risk value and the stage status value as a deviation; arranging the deviations of each stage in chronological order to form a deviation sequence; calculating the mean and standard deviation of the deviation sequence; calculating the relative deviation of each deviation from the mean of the deviation sequence; obtaining the deviation influence factor based on the relative deviation and the standard deviation; setting the corresponding correction coefficient based on the deviation influence factor, and obtaining the corrected risk value based on the risk value and the correction coefficient.

7. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 4, characterized in that: The method further comprises: S400: Obtaining the modified risk values ​​and corresponding stage status values ​​at different times to form time series data, and then constructing a time series model; Based on the time series model, draw a risk evolution map, and set the corresponding benchmark efficiency threshold for each production stage based on the risk evolution map; S500: Based on the benchmark performance threshold, record the number of performance deviations in each production stage, formulate preventive measures based on the number of occurrences and the difference handling mechanism, and then determine the cause of the risk; comprehensively determine the enterprise's operating risk level based on the risk cause and the modified risk value.

8. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 7, characterized in that: The time series model is constructed as follows: Create time series features and stage state codes based on time series data; Define the model structure, including the input layer, LSTM layer, and fully connected layer; Divide the time series data into training set and test set, use the training set data to train the time series model; use the test set data to verify the model effect; input the new time series data into the trained time series model to obtain the predicted future risk value and stage status value; The time series features include lag features and moving average features, and the stage state encoding is to convert the stage state value into a one-hot encoding.

9. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 7, characterized in that: Step S500 also includes: S510: If the number of occurrences is greater than three times, search the data warehouse for historical data of the same working condition according to the working condition identification of the current production stage; perform correlation analysis based on the current performance deviation data and historical data of the same working condition to identify auxiliary factors; S520: Identify the change range of the current deviation sequence and compare it with the pre-set range, and obtain the correction factor according to the comparison result; S520: Determine the cause of the risk based on the auxiliary factors and the corrective factors.

10. The enterprise operation risk assessment method based on artificial intelligence analysis as claimed in claim 9, characterized in that: The data warehouse is used to store historical data of all production stages, including: environmental parameters, equipment maintenance records, production efficiency indicators and production batches; the working condition identifier is a unique identification code for each production stage; the auxiliary factor refers to an important influencing factor reflecting efficiency deviation; the correction factor is used to reflect the degree of deviation of the deviation sequence.

Images