Enterprise risk dynamic assessment method, system and device based on multi-dimensional data

Abstract

The invention discloses an enterprise risk dynamic assessment method, system and device based on multi-dimensional data, and the method comprises the steps: obtaining multi-dimensional enterprise related data, and obtaining a preprocessing assessment data set through preprocessing; performing exception extraction and feature extraction on the electric energy standard data and the productivity standard data to obtain an electric energy evaluation data set and a productivity evaluation data set; an enterprise risk assessment pre-training model is established, and the enterprise risk assessment pre-training model comprises at least three risk assessment pre-training sub-models and a total output layer; training the risk assessment pre-training sub-model to obtain a corresponding risk sub-assessment model and form an enterprise risk assessment model; and analyzing the to-be-predicted enterprise risk data to obtain an electric energy risk assessment probability, a productivity risk assessment probability and a production risk assessment probability, thereby obtaining risk prediction data. According to the invention, the multi-dimensional data of the enterprise risk is analyzed and predicted, so that the accuracy of enterprise risk assessment is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of risk assessment, and in particular to a method, system and device for dynamic enterprise risk assessment based on multidimensional data. Background Art

[0002] In the current field of enterprise management, accurate assessment of enterprise risks is conducive to the stable development of enterprises. At present, there are two commonly used enterprise risk assessment methods: one is to analyze the enterprise's electricity consumption and historical transaction records; the other is to obtain some relevant data through manual market research and then analyze them. These methods basically monitor and analyze the enterprise's electricity consumption and transaction records over a period of time, and use statistical or machine learning methods to analyze the change trends and cyclical characteristics, and then infer whether the enterprise's production activities are stable and whether there are operating risks.

[0003] Although the current risk assessment methods can identify potential risky customers to a certain extent, they have the following defects: they ignore the relationship between production capacity and energy consumption. In the production process of an enterprise, energy consumption and production capacity are directly related to the operational efficiency and cost control of the enterprise. The existing assessment methods ignore this link, resulting in inaccurate risk assessment results; at the same time, the existing methods have single assessment data, do not analyze videos such as the production process of the enterprise and the operating status of processing equipment, lack the integration of video surveillance technology, and it is difficult to accurately assess the stability of the enterprise's operations and its ability to produce safely; they cannot meet the high dynamics and high flexibility of changes in the market environment, customer needs and internal conditions of the enterprise, resulting in enterprise risk assessment often lagging behind actual market changes. Summary of the invention

[0004] In view of the shortcomings of the prior art, the present invention provides a method, system and device for dynamically evaluating enterprise risks based on multidimensional data.

[0005] In order to solve the above technical problems, the present invention is solved by the following technical solutions: A method for dynamic enterprise risk assessment based on multidimensional data, comprising the following steps: Acquire multi-dimensional enterprise-related data and perform preprocessing to obtain a preprocessing evaluation data set, wherein the multi-dimensional enterprise-related data includes power data, capacity data, and production data, and the preprocessing evaluation data includes power standard data, capacity standard data, and production standard data; Perform abnormal data analysis and feature extraction on the electric energy standard data to obtain an electric energy evaluation data set; perform abnormal data analysis and feature extraction on the production capacity standard data to obtain a production capacity evaluation data set, wherein the electric energy evaluation data includes electric energy abnormal data and electric energy feature data, and the production capacity evaluation data includes production capacity abnormal data and production capacity feature data; Establish an enterprise risk assessment pre-training model, the enterprise risk assessment pre-training model includes at least three risk assessment pre-training sub-models and a total output layer, each risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer, the data receiving layer receives an electric energy assessment data set, a production capacity assessment data set and a production standard data set respectively, the data processing layer performs corresponding processing on the electric energy assessment data, the production capacity assessment data and the production standard data respectively, the probability output layer obtains the electric energy risk assessment probability, the production capacity risk assessment probability and the production risk assessment probability respectively, and performs a comprehensive assessment through the total output layer to obtain risk prediction data; The risk assessment pre-trained sub-model is trained respectively through the electric energy assessment data set, the capacity assessment data set and the production standard data set to obtain the corresponding risk assessment sub-model, and the enterprise risk assessment model is formed through the corresponding risk assessment sub-model; Based on the enterprise risk assessment model, the risk data of the enterprise to be predicted is analyzed to obtain the power risk assessment results, production capacity risk assessment results and production risk assessment results, and then the risk prediction results are obtained.

[0006] As an implementable method, the obtaining of multi-dimensional enterprise-related data includes the following steps: According to the energy consumption data of the enterprise, the meter readings of the day and the previous day are obtained, and the electric energy data is obtained based on the collection rate data of different regions, which is expressed as follows:

[0007] Obtain enterprise production management data and analyze it to obtain capacity data, where the capacity data includes order production data; Collect the video data produced by the enterprise to obtain production data, and form multi-dimensional enterprise-related data based on power data, capacity data and production data; in, Indicates Daily power data, Indicates the meter reading for the day. Indicates the meter reading of the previous day. Indicates the acquisition rate data. Indicates today.

[0008] As an implementable method, the preprocessing comprises the following steps: Traverse the electric energy data and production capacity data, remove invalid and erroneous data, and obtain the removed electric energy data and production capacity data; The electric energy data after elimination is analyzed, and based on the maximum and minimum values ​​of the electric energy data after elimination, the electric energy data is standardized to obtain the electric energy standard data, which is expressed as follows:

[0009] The capacity data after elimination is analyzed, and based on the maximum and minimum values ​​of the capacity data after elimination, the capacity data is standardized to obtain the capacity standard data, which is expressed as follows:

[0010] Perform video segmentation on production data, remove erroneous data and duplicate data from the split production data, and annotate abnormal behaviors to obtain a production standard data set; in, Represents electrical energy data, Represents production capacity data, Indicates the standard data of electric energy. Indicates the minimum value of electric energy data. Indicates the maximum value of the electric energy data. Indicates the production capacity standard data, Indicates the minimum value of the capacity data. Indicates the maximum value of the capacity data.

[0011] As an implementable method, the abnormal data analysis and feature extraction of electric energy standard data includes the following steps: Preset power thresholds and obtain power average and power standard deviation based on historical power standard data; Based on the electric energy threshold, electric energy average value, electric energy standard deviation and current electric energy standard data, the electric energy analysis data is obtained, which is expressed as follows:

[0012] If the power analysis data does not meet the power threshold, the current power standard data is abnormal power data; The electric energy standard data is analyzed to obtain daily average characteristic data of electric energy and monthly electric energy characteristic data. Based on the monthly electric energy characteristic data, the monthly average characteristic data of electric energy is obtained, which is expressed as follows:

[0013]

[0014] Based on the analysis of the current daily energy characteristic data and the previous day's energy characteristic data, the daily energy growth characteristic data is obtained, which is expressed as follows:

[0015] Based on daily average characteristic data of electric energy, monthly average characteristic data of electric energy and daily growth characteristic data of electric energy, electric energy characteristic data is formed; in, Indicates the power analysis data, Represents the average value of electric energy. represents the standard deviation of electric energy, Represents the average characteristic data of daily energy, Indicates the standard data of electric energy. Indicates that day, Indicates the total number of days. Indicates the average characteristic data of monthly electric energy, Indicates Monthly power characteristic data for the month, Indicates the total number of months, Indicates daily energy growth characteristic data, Indicates The standard data of electricity for the day, Indicates Daily electricity standard data.

[0016] As an implementable method, the abnormal data analysis and feature extraction of the production capacity standard data includes the following steps: Preset capacity thresholds and obtain the capacity average and capacity standard deviation based on historical capacity standard data; Based on the capacity threshold, capacity average, capacity standard deviation and current capacity standard data, the capacity analysis data is obtained, which is expressed as follows:

[0017] If the capacity analysis data does not meet the capacity threshold, the current capacity standard data is capacity abnormal data; Obtain and analyze the capacity standard data for multiple days to obtain the capacity characteristic data, which is expressed as follows:

[0018] in, Indicates abnormal capacity data. Indicates that day, represents the average capacity. represents the standard deviation of production capacity, Represents capacity characteristic data, Indicates the total number of days. Indicates Daily production capacity standard data, Indicates capacity standard data.

[0019] As an implementable embodiment, the data receiving layer receives the windowed power evaluation data; The corresponding data processing layer includes a forget gate, an input gate, and an output gate. The forget gate obtains the power evaluation data and obtains the discarded information data through analysis, which is expressed as follows:

[0020] The input gate analyzes the power evaluation data to obtain the updated information data and the current information status, which are expressed as follows:

[0021]

[0022] Get the long-term information status of the previous moment, combine the discarded information status, updated information status and current information status, and get the long-term information status of the current moment, as shown below:

[0023] The output gate analyzes the power evaluation data to obtain the power output data. By analyzing the power output data and the long-term information state at the current moment, the sequence relationship data is obtained, which is expressed as follows:

[0024]

[0025] The probability output layer evaluates the sequence relationship data to obtain the probability of power risk assessment; in, Indicates discarding information data. represents the sigmoid function, represents the forget gate weight, Represents the sequence relationship data of the previous moment, Indicates the current power evaluation data, represents the forget gate bias, Indicates update information data. represents the input gate weight, represents the input gate bias, Indicates the current information status. represents the weight of the current state unit, represents the bias of the current state unit, Indicates the current long-term information status. Indicates the long-term information status of the previous moment. Indicates the power output data, represents the output gate weight, represents the output gate bias, Represents sequence relationship data.

[0026] As an implementable embodiment, the data receiving layer receives capacity assessment data; The corresponding data processing layer includes an update gate and a reset gate. The capacity evaluation data is analyzed by updating the gate weight and resetting the gate weight to obtain the updated evaluation data and the reset evaluation data, which are expressed as follows:

[0027]

[0028] By resetting the evaluation data and the capacity dependency data of the previous moment for analysis, the candidate hidden state of the current moment is obtained, which is expressed as follows:

[0029] Based on the updated evaluation data, the capacity dependency data at the previous moment and the candidate hidden state at the current moment, the capacity dependency data at the current moment is obtained, which is expressed as follows:

[0030] The probability output layer analyzes the capacity dependency data to obtain the probability of capacity risk assessment; in, Indicates updating the evaluation data. represents the update gate weight, Indicates the capacity dependency data at the previous moment, Represents capacity assessment data, Indicates resetting the evaluation data. Resets the gate weights. represents the sigmoid function, represents the candidate hidden state at the current moment, represents the weight matrix, Indicates the capacity dependency data at the current moment, Indicates the current moment.

[0031] As an implementable method, the data receiving layer receives a production standard data set and divides it into a training set, a test set and a validation set; The corresponding data processing layer includes a backbone network layer and a neck network layer. The backbone network layer extracts features of the production standard data, including edge features, texture features and shape features, and reduces the complexity and parameter amount of feature calculation through a convolution module and a residual module; the neck network layer obtains image features at different stages of the backbone network and fuses them to obtain fused image features, and enhances the fused image features to obtain enhanced image features; The probability output layer performs analysis and evaluation based on the enhanced image features to obtain the production risk assessment probability.

[0032] As an implementable method, the comprehensive assessment is performed to obtain risk prediction data, specifically: Set the model weight parameters, analyze the power risk assessment probability, capacity risk assessment probability and production risk assessment probability based on the model weight parameters, and obtain risk prediction data, which is expressed as follows:

[0033] in, represents risk prediction data, represents the model weight parameters of the risk assessment pre-trained sub-model, represents the probability of electric energy risk assessment, the probability of capacity risk assessment, or the probability of production risk assessment, Represents the number of risk assessment pre-trained sub-models.

[0034] As an implementable method, it also includes cross-validation and evaluation of the risk assessment sub-model, specifically: Get the evaluation indicators of the risk assessment sub-model, where the evaluation indicators include accuracy, recall, F1 score and AUC-ROC, as shown below:

[0035]

[0036]

[0037]

[0038] Model subsets are established through the power evaluation data set, capacity evaluation data set and production standard data set respectively, and the risk assessment sub-model is cross-validated based on the model subset and evaluation indicators, as shown below:

[0039] in, Indicates accuracy, represents true positive data, represents true negative data, represents false positive data, represents false negative data, represents the recall rate, express Score value, express Curve area data, represents the cross validation score, represents the number of model subsets, Represents the risk assessment sub-model The evaluation indicators are Indicates that through The risk assessment sub-model obtained by training the model subsets is Indicates the true proportion, represents the false positive proportion.

[0040] A dynamic enterprise risk assessment system based on multidimensional data, comprising a data preprocessing module, a data extraction module, a model building module, a model training module and a risk assessment module; The data preprocessing module acquires and preprocesses multi-dimensional enterprise-related data to obtain a preprocessed evaluation data set, wherein the multi-dimensional enterprise-related data includes power data, capacity data, and production data, and the preprocessed evaluation data includes power standard data, capacity standard data, and production standard data; The data extraction module performs abnormal data analysis and feature extraction on the electric energy standard data to obtain an electric energy evaluation data set; performs abnormal data analysis and feature extraction on the production capacity standard data to obtain a production capacity evaluation data set, wherein the electric energy evaluation data includes electric energy abnormal data and electric energy feature data, and the production capacity evaluation data includes production capacity abnormal data and production capacity feature data; The model building module builds an enterprise risk assessment pre-training model, which includes at least three risk assessment pre-training sub-models and a total output layer. Each risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer. The data receiving layer receives an electric energy assessment data set, a production capacity assessment data set and a production standard data set respectively. The data processing layer processes the electric energy assessment data, the production capacity assessment data and the production standard data respectively. The probability output layer obtains the electric energy risk assessment probability, the production capacity risk assessment probability and the production risk assessment probability respectively, and performs a comprehensive assessment through the total output layer to obtain risk prediction data; The model training module trains the risk assessment pre-training sub-model through the electric energy assessment data set, the capacity assessment data set and the production standard data set to obtain the corresponding risk assessment sub-model, and forms the enterprise risk assessment model through the corresponding risk assessment sub-model; The risk assessment module analyzes the risk data of the enterprise to be predicted based on the enterprise risk assessment model, obtains the power risk assessment results, the production capacity risk assessment results and the production risk assessment results, and then obtains the risk prediction results.

[0041] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the following method is implemented: Acquire multi-dimensional enterprise-related data and perform preprocessing to obtain a preprocessing evaluation data set, wherein the multi-dimensional enterprise-related data includes power data, capacity data, and production data, and the preprocessing evaluation data includes power standard data, capacity standard data, and production standard data; Perform abnormal data analysis and feature extraction on the electric energy standard data to obtain an electric energy evaluation data set; perform abnormal data analysis and feature extraction on the production capacity standard data to obtain a production capacity evaluation data set, wherein the electric energy evaluation data includes electric energy abnormal data and electric energy feature data, and the production capacity evaluation data includes production capacity abnormal data and production capacity feature data; Establish an enterprise risk assessment pre-training model, the enterprise risk assessment pre-training model includes at least three risk assessment pre-training sub-models and a total output layer, each risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer, the data receiving layer receives an electric energy assessment data set, a production capacity assessment data set and a production standard data set respectively, the data processing layer performs corresponding processing on the electric energy assessment data, the production capacity assessment data and the production standard data respectively, the probability output layer obtains the electric energy risk assessment probability, the production capacity risk assessment probability and the production risk assessment probability respectively, and performs a comprehensive assessment through the total output layer to obtain risk prediction data; The risk assessment pre-trained sub-model is trained respectively through the electric energy assessment data set, the capacity assessment data set and the production standard data set to obtain the corresponding risk assessment sub-model, and the enterprise risk assessment model is formed through the corresponding risk assessment sub-model; Based on the enterprise risk assessment model, the risk data of the enterprise to be predicted is analyzed to obtain the power risk assessment results, production capacity risk assessment results and production risk assessment results, and then the risk prediction results are obtained.

[0042] A device for dynamically assessing enterprise risks based on multidimensional data comprises a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the following method when executing the computer program: Acquire multi-dimensional enterprise-related data and perform preprocessing to obtain a preprocessing evaluation data set, wherein the multi-dimensional enterprise-related data includes power data, capacity data, and production data, and the preprocessing evaluation data includes power standard data, capacity standard data, and production standard data; Perform abnormal data analysis and feature extraction on the electric energy standard data to obtain an electric energy evaluation data set; perform abnormal data analysis and feature extraction on the production capacity standard data to obtain a production capacity evaluation data set, wherein the electric energy evaluation data includes electric energy abnormal data and electric energy feature data, and the production capacity evaluation data includes production capacity abnormal data and production capacity feature data; Establish an enterprise risk assessment pre-training model, the enterprise risk assessment pre-training model includes at least three risk assessment pre-training sub-models and a total output layer, each risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer, the data receiving layer receives an electric energy assessment data set, a production capacity assessment data set and a production standard data set respectively, the data processing layer performs corresponding processing on the electric energy assessment data, the production capacity assessment data and the production standard data respectively, the probability output layer obtains the electric energy risk assessment probability, the production capacity risk assessment probability and the production risk assessment probability respectively, and performs a comprehensive assessment through the total output layer to obtain risk prediction data; The risk assessment pre-trained sub-model is trained respectively through the electric energy assessment data set, the capacity assessment data set and the production standard data set to obtain the corresponding risk assessment sub-model, and the enterprise risk assessment model is formed through the corresponding risk assessment sub-model; Based on the enterprise risk assessment model, the risk data of the enterprise to be predicted is analyzed to obtain the power risk assessment results, production capacity risk assessment results and production risk assessment results, and then the risk prediction results are obtained.

[0043] The present invention has significant technical effects due to the adoption of the above technical solution: The present invention integrates multiple data sources to form multi-dimensional enterprise-related data, including power data, production capacity data, and video surveillance data of production sites. By performing abnormal analysis and feature extraction on the data, the present invention combines time series algorithms and abnormal detection algorithms to achieve a comprehensive, real-time, and dynamic assessment of enterprise risks. This improves the accuracy and timeliness of enterprise risk assessment, helps enterprises better identify and manage potential risks, and enhances market competitiveness. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0045] Figure 1 It is a schematic flow diagram of the method of the present invention; Figure 2 It is an overall schematic diagram of the system of the present invention; Figure 3 It is a schematic diagram of the model structure of the present invention. DETAILED DESCRIPTION

[0046] The present invention is further described in detail below in conjunction with embodiments. The following embodiments are for explanation of the present invention but the present invention is not limited to the following embodiments.

[0047] Embodiment 1: A dynamic enterprise risk assessment method based on multidimensional data, such as Figure 1 The method comprises the following steps: S100, acquiring multi-dimensional enterprise-related data and preprocessing it to obtain a preprocessed evaluation data set, wherein the multi-dimensional enterprise-related data includes power data, capacity data and production data, and the preprocessed evaluation data includes power standard data, capacity standard data and production standard data; S200, performing abnormal data analysis and feature extraction on the electric energy standard data to obtain an electric energy evaluation data set; performing abnormal data analysis and feature extraction on the production capacity standard data to obtain a production capacity evaluation data set, wherein the electric energy evaluation data includes electric energy abnormal data and electric energy feature data, and the production capacity evaluation data includes production capacity abnormal data and production capacity feature data; S300, establishing an enterprise risk assessment pre-training model, the enterprise risk assessment pre-training model includes at least three risk assessment pre-training sub-models and a total output layer, each risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer, the data receiving layer receives an electric energy assessment data set, a production capacity assessment data set and a production standard data set respectively, the data processing layer processes the electric energy assessment data, the production capacity assessment data and the production standard data respectively, the probability output layer obtains the electric energy risk assessment probability, the production capacity risk assessment probability and the production risk assessment probability respectively, and performs a comprehensive assessment through the total output layer to obtain risk prediction data; S400, respectively training the risk assessment pre-training sub-model through the electric energy assessment data set, the capacity assessment data set and the production standard data set to obtain the corresponding risk assessment sub-model, and forming the enterprise risk assessment model through the corresponding risk assessment sub-model; S500, analyzing the risk data of the enterprise to be predicted based on the enterprise risk assessment model to obtain the electric energy risk assessment results, the production capacity risk assessment results and the production risk assessment results, and then obtaining the risk prediction results.

[0048] The present invention collects video data produced by an enterprise to obtain production data, combines electric energy data and production capacity data to form multi-dimensional enterprise-related data, and forms an electric energy assessment data set, a production capacity assessment data set and a production standard data set through data analysis, constructs an enterprise risk assessment pre-training model, and obtains an enterprise risk assessment model through training, and then performs reasoning analysis based on the enterprise risk assessment model to obtain risk prediction results. The present invention performs a comprehensive and accurate dynamic assessment of enterprise risks based on multi-dimensional data, and realizes real-time, accurate and dynamic assessment of enterprise customer risks. It overcomes the problems of single data, neglect of the relationship between data and lack of video data fusion in existing methods, and realizes a comprehensive, accurate and timely assessment of enterprise risks. Compared with existing methods, it improves the accuracy and timeliness of risk assessment, helps enterprises better identify and manage risks, and enhances the market competitiveness of enterprises.

[0049] The assessment and management of enterprise risks is of great significance to the stable development of enterprises. In the current field of enterprise risk management, a variety of technical solutions have been proposed and used in practice for enterprise risk management. A typical method is to monitor and analyze the electricity consumption data of enterprises, collect the electricity consumption data of enterprises over a period of time, identify the changing trend and periodic characteristics of the electricity consumption of enterprises through statistical or machine learning methods, and then infer whether the production activities of enterprises are stable, so as to achieve the purpose of predicting the risk level of enterprises. In addition, traditional enterprise risk analysis can be achieved through market research and other means, but such methods are easily limited by the authenticity and timeliness of data, and manual research has the problems of high cost and low efficiency. In the production process of enterprises, the energy consumption and production capacity of enterprises are directly related to the operating efficiency and cost control ability of enterprises. The video surveillance data of the production site of enterprises, such as the production process monitoring data and the operating status of processing equipment, are valuable resources for evaluating the stability of enterprise operations and safe production capabilities. However, existing methods lack multi-dimensional data considerations when conducting data analysis, ignore the relationship between electricity data, capacity data and production data, and cannot effectively use production video information for risk prediction. It is difficult to fully reflect the real risks of the enterprise and cannot meet the high dynamics and flexibility required by the market environment, customer needs and internal changes of the enterprise. Traditional risk assessments often lag behind the actual changes of the enterprise.

[0050] This embodiment takes a glass production enterprise as an example, and realizes data collection by combining smart sensors with Internet of Things technology, and using data collection reporting and cloud storage. By installing smart sensors in the key production links of the glass production enterprise's production line, the electric energy consumption data used by the glass processing enterprise for production is collected in real time to obtain electric energy data; by combining the enterprise's production relationship system with the production capacity monitoring system, the data is stored in the cloud platform, and the data changes of glass products from ordering to production and processing are recorded in real time to obtain production capacity data; high-definition cameras are used to conduct all-weather video monitoring of the production site of the glass processing enterprise to obtain production data. Multi-dimensional enterprise-related data is formed through electric energy data, production capacity data and production data. The specific process includes the following steps: Step 1: In this embodiment, the energy consumption benchmark value is set according to the production scale of the enterprise, the debugging power information generated at the initial installation of the power equipment is eliminated, the data collection error is reduced, and the power sensor is used to collect data to obtain the meter readings of the day and the meter readings of the previous day; Step 2: Based on the collection rate data of different regions, analyze the meter readings of the day and the meter readings of the previous day to obtain the electric energy data, which is expressed as follows:

[0051] Step 3: Combine the enterprise production management system with the capacity monitoring system, and store the data on the cloud platform to record the data changes in the glass production process in real time, including the production quantity and production cycle of glass processing and other capacity indicators, and record the order volume and the growth and change of various production indicators, so as to analyze the change trend of the enterprise capacity data and obtain the capacity data, among which the capacity data includes order production data; Step 4: Use high-definition cameras to monitor the production site of the glass production enterprise around the clock. Through video acquisition and irregular extraction of images of different frequencies, production data is obtained, and combined with power data and production capacity data to form multi-dimensional enterprise-related data; in, Indicates Daily power data, Indicates the meter reading for the day. Indicates the meter reading of the previous day. Indicates the acquisition rate data. Indicates today.

[0052] To facilitate subsequent data analysis, in this embodiment, multi-dimensional enterprise-related data is pre-processed to ensure data consistency and comparability through format conversion and standardization, including the following steps: Step 1: To ensure the integrity and accuracy of the power data and the production capacity data, in this embodiment, the power data and the production capacity data are traversed, and invalid and erroneous data are eliminated to obtain the power data and production capacity data after elimination; Step 2: Standardize the electric energy data after elimination. Based on the maximum and minimum values ​​of the electric energy data after elimination, analyze the electric energy data to obtain the electric energy standard data, which is expressed as follows:

[0053] Step 3: Standardize the capacity data after elimination. Based on the maximum and minimum values ​​of the capacity data after elimination, standardize the capacity data to obtain the capacity standard data, which is expressed as follows:

[0054] Step 4: Perform video segmentation on the production data, remove the erroneous data and duplicate data in the production data after video segmentation, and mark abnormal behaviors to obtain a production standard data set; in, Represents electrical energy data, Represents production capacity data, Indicates the standard data of electric energy. Indicates the minimum value of electric energy data. Indicates the maximum value of the electric energy data. Indicates the production capacity standard data, Indicates the minimum value of the capacity data. Indicates the maximum value of the capacity data.

[0055] In this embodiment, abnormal data analysis is performed on the electric energy standard data and the production capacity standard data to obtain abnormal electric energy data and abnormal production capacity data, and feature extraction is performed on the electric energy standard data and the production capacity standard data to obtain electric energy characteristic data and production capacity characteristic data, thereby forming an electric energy evaluation data set and a production capacity evaluation data set. The abnormal data analysis and feature extraction of the electric energy standard data include the following steps: Step 1: Preset the power threshold, obtain the historical power standard data and analyze it to obtain the power average value and power standard deviation, and analyze it in combination with the current power standard data to obtain power analysis data, which is expressed as follows:

[0056] Step 2: Analyze based on the power threshold and power analysis data. If the power analysis data does not meet the power threshold, in this embodiment, it is set to or , then the current electric energy standard data is abnormal electric energy data; Step 3: Analyze the electric energy standard data to obtain daily average characteristic data and monthly electric energy characteristic data. Based on the monthly electric energy characteristic data, obtain the monthly average characteristic data. Based on the current daily energy characteristic data and the previous day's daily energy characteristic data, analyze to obtain daily energy growth characteristic data, which is expressed as follows:

[0057]

[0058]

[0059] Step 4: Based on the daily energy average characteristic data, the monthly energy average characteristic data and the daily energy growth characteristic data, the electric energy characteristic data is formed; in, Indicates the power analysis data, Represents the average value of electric energy. represents the standard deviation of electric energy, Represents the average characteristic data of daily energy, Indicates the standard data of electric energy. Indicates that day, Indicates the total number of days. Indicates the average characteristic data of monthly electric energy, Indicates Monthly power characteristic data for the month, Indicates the total number of months, Indicates daily energy growth characteristic data, Indicates The standard data of electricity for the day, Indicates Daily electricity standard data.

[0060] Abnormal data analysis and feature extraction of production capacity standard data include the following steps: Step 1: Preset the capacity threshold, obtain and analyze the historical capacity standard data, obtain the capacity average and capacity standard deviation, and analyze the capacity standard data to obtain the capacity analysis data, which is expressed as follows:

[0061] Step 2: Analyze based on the capacity threshold and capacity analysis data. If the capacity analysis data does not meet the capacity threshold, in this embodiment, it is set to or , then the current capacity standard data is capacity abnormal data; Step 3: Obtain and analyze the capacity standard data for multiple days to obtain the capacity characteristic data, which is expressed as follows:

[0062] in, Indicates abnormal capacity data. Indicates that day, represents the average capacity. represents the standard deviation of production capacity, Represents capacity characteristic data, Indicates the total number of days. Indicates Daily production capacity standard data, Indicates capacity standard data.

[0063] In this embodiment, for the electric energy assessment data set, the capacity assessment data set and the production standard data set, an enterprise risk assessment pre-training model including at least three risk assessment pre-training sub-models and a total output layer is established, such as Figure 3 As shown, the risk assessment pre-training sub-model is used to analyze the power assessment data set, the capacity assessment data set and the production standard data set respectively, and the power risk assessment probability, the capacity risk assessment probability and the production risk assessment probability are obtained. Among them, the risk assessment pre-training sub-model for analyzing the power assessment data set is implemented by a recursive neural network, which solves the gradient vanishing problem of ordinary recursive neural networks when processing long sequence data, including a data receiving layer, a data processing layer and a probability output layer. The specific steps are as follows: Step 1: Window the power assessment data and input it into the data receiving layer of the risk assessment pre-training sub-model; Step 2: The corresponding data processing layer includes a forget gate, an input gate, and an output gate. The forget gate determines which information needs to be retained and which information needs to be discarded, and maps it to between 0 and 1 through the sigmoid activation function to obtain the discarded information data. The discarded information data close to 0 is forgotten, and the discarded information data close to 1 is retained. The discarded information data is represented as follows:

[0064] Step 3: The input gate controls how much information in the current input data will be updated to the memory unit. Similarly, the update ratio is obtained through the sigmoid function to obtain the updated information data, and then transformed through the tanh function to obtain the current information data, which is expressed as follows:

[0065]

[0066] Step 4: Get the long-term information status of the previous moment, combine the discarded information status, updated information status and current information status, and get the long-term information status of the current moment, which is expressed as follows:

[0067] Step 5: The output gate determines which information in the memory unit will be output as the hidden state at the current moment, and the power output data is obtained through the sigmoid function. The long-term information state at the current moment is analyzed through the tanh function and the sequence relationship data is obtained based on the power output function. The power output data and the sequence relationship data are expressed as follows:

[0068]

[0069] Step 6: The probability output layer performs risk assessment based on the sequence relationship data to obtain the probability of power risk assessment; in, Indicates discarding information data. represents the sigmoid function, represents the forget gate weight, Represents the sequence relationship data of the previous moment, Indicates the current power evaluation data, represents the forget gate bias, Indicates update information data. represents the input gate weight, represents the input gate bias, Indicates the current information status. represents the weight of the current state unit, represents the bias of the current state unit, Indicates the current long-term information status. Indicates the long-term information status of the previous moment. Indicates the power output data, represents the output gate weight, represents the output gate bias, Represents sequence relationship data.

[0070] In this embodiment, the risk assessment pre-training sub-model for analyzing the capacity assessment data set is implemented by a recurrent neural network, which avoids the gradient vanishing and gradient exploding problems that occur when processing sequence data, and retains valid information and discards useless information through a gating mechanism, thereby improving the computational efficiency of data processing. The risk assessment pre-training sub-model includes a data receiving layer, a data processing layer, and a probability output layer, and specifically includes the following steps: Step 1: The data receiving layer receives the input capacity assessment data; Step 2: The data processing layer includes an update gate and a reset gate. The update gate determines how much historical information the model should pass to the future, and how much information from the previous time and the current time needs to be continued to be passed. That is, the capacity evaluation data is analyzed through the sigmoid function, the update gate weight, and the reset gate weight to obtain the updated evaluation data and the reset evaluation data, and then the information that the network needs to continue to pass is analyzed, which is expressed as follows:

[0071]

[0072] Step 3: Obtain the capacity dependency data of the previous moment and analyze it in combination with the reset evaluation data to obtain the candidate hidden state of the current moment. Combine the updated evaluation data and the candidate hidden state of the current moment to obtain the capacity dependency data of the current moment. The candidate hidden state and capacity dependency data of the current moment are expressed as follows:

[0073]

[0074] Step 4: The probability output layer performs risk assessment analysis on the capacity dependency data to obtain the capacity risk assessment probability; in, Indicates updating the evaluation data. represents the update gate weight, Indicates the capacity dependency data at the previous moment, Represents capacity assessment data, Indicates resetting the evaluation data. Resets the gate weights. represents the sigmoid function, represents the candidate hidden state at the current moment, represents the weight matrix, Indicates the capacity dependency data at the current moment, Indicates the current moment.

[0075] In this embodiment, when the production standard data set is analyzed by the risk assessment pre-training sub-model, the risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer. First, the production standard data set is received through the data receiving layer, and is divided into a training set, a test set and a validation set according to the proportion; the corresponding data processing layer first extracts features of the received production standard data through the backbone network layer, wherein the extracted features include edge features, texture features and shape features, and then analyzes the features through the convolution module and the residual module, and reduces the complexity of feature calculation and the number of network parameters; then, the image features obtained by the backbone network at different stages are obtained through the neck network layer and feature fusion is performed to obtain fused image features, and the fused image features are further enhanced to obtain enhanced image features; the probability output layer obtains the production risk assessment probability by analyzing and evaluating the enhanced image features.

[0076] After obtaining the probability of electric energy risk assessment, the probability of production capacity risk assessment and the probability of production risk assessment, this embodiment uses the total output layer of the enterprise risk assessment pre-training model to comprehensively analyze the probability of electric energy risk assessment, the probability of production capacity risk assessment and the probability of production risk assessment by using the model fusion and integration method to improve the accuracy of risk assessment and enhance the generalization ability and prediction performance of the risk assessment pre-training sub-model. For example, the probability of electric energy risk assessment, the probability of production capacity risk assessment and the probability of production risk assessment are weighted or voted to make decisions, and the advantages of each risk assessment sub-model are combined to reduce the prediction error of a single sub-model. Specifically: Set the model weight parameters, analyze the power risk assessment probability, capacity risk assessment probability and production risk assessment probability based on the model weight parameters, and obtain risk prediction data, which is expressed as follows:

[0077] in, represents risk prediction data, represents the model weight parameters of the risk assessment pre-trained sub-model, represents the probability of electric energy risk assessment, the probability of capacity risk assessment, or the probability of production risk assessment, Represents the number of risk assessment pre-trained sub-models.

[0078] At the same time, in the process of training the risk assessment pre-trained sub-model through the electric energy assessment data set, the production capacity assessment data set and the production standard data set, the network hyperparameters of the risk assessment pre-trained sub-model are fine-tuned in this embodiment. Through the grid search method, all possible network hyperparameter combinations are traversed. Through the random search method, the network hyperparameter combination is selected to improve the model parameter search efficiency and dynamically adjust the network hyperparameters of the risk assessment pre-trained sub-model to obtain the optimal model parameter combination of the risk assessment pre-trained sub-model and obtain the optimal network hyperparameters. Based on the optimal network hyperparameters, the corresponding risk assessment sub-model is obtained, thereby forming an enterprise risk assessment model.

[0079] In this embodiment, during the training of the risk assessment pre-training sub-model, the evaluation index of the risk assessment sub-model is obtained and the model is optimized based on the evaluation index, and then convergence is achieved through iteration to obtain the risk assessment sub-model. Among them, the evaluation indicators include accuracy, recall, F1 score and AUC-ROC. The accuracy is used to measure the proportion of samples correctly predicted by the model, the recall is used to measure the proportion of the number of correctly identified positive samples to the number of all actual positive samples, the F1 score takes into account the accuracy and recall of the model when evaluating the model, and the AUC-ROC is used to evaluate the performance of the model, and can intuitively understand the different performances of the model under different network parameters, and then select the best model. The specific calculation methods of accuracy, recall, F1 score and AUC-ROC are expressed as follows:

[0080]

[0081]

[0082]

[0083] in, Indicates accuracy, represents true positive data, represents true negative data, represents false positive data, represents false negative data, represents the recall rate, express Score value, express Curve area data, Indicates the true proportion, represents the false positive proportion.

[0084] In addition, in the process of training the model, this embodiment also includes establishing a model subset through the electric energy assessment data set, the capacity assessment data set and the production standard data set, and cross-validating the risk assessment sub-model based on the model subset and the assessment index, specifically:

[0085] in, represents the cross validation score, represents the number of model subsets, Represents the risk assessment sub-model The evaluation indicators are Indicates that through The risk assessment sub-model is obtained by training the model subsets.

[0086] Embodiment 2: A dynamic enterprise risk assessment system based on multidimensional data, such as Figure 2 As shown, it includes a data preprocessing module 100, a data extraction module 200, a model building module 300, a model training module 400 and a risk assessment module 500; The data preprocessing module 100 acquires and preprocesses multi-dimensional enterprise-related data to obtain a preprocessed evaluation data set, wherein the multi-dimensional enterprise-related data includes power data, capacity data and production data, and the preprocessed evaluation data includes power standard data, capacity standard data and production standard data; The data extraction module 200 performs abnormal data analysis and feature extraction on the electric energy standard data to obtain an electric energy evaluation data set; performs abnormal data analysis and feature extraction on the production capacity standard data to obtain a production capacity evaluation data set, wherein the electric energy evaluation data includes electric energy abnormal data and electric energy feature data, and the production capacity evaluation data includes production capacity abnormal data and production capacity feature data; The model building module 300 builds an enterprise risk assessment pre-training model, which includes at least three risk assessment pre-training sub-models and a total output layer. Each risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer. The data receiving layer receives an electric energy assessment data set, a production capacity assessment data set and a production standard data set respectively. The data processing layer processes the electric energy assessment data, the production capacity assessment data and the production standard data respectively. The probability output layer obtains the electric energy risk assessment probability, the production capacity risk assessment probability and the production risk assessment probability respectively, and performs a comprehensive assessment through the total output layer to obtain risk prediction data; The model training module 400 trains the risk assessment pre-training sub-model through the electric energy assessment data set, the capacity assessment data set and the production standard data set to obtain the corresponding risk assessment sub-model, and forms the enterprise risk assessment model through the corresponding risk assessment sub-model; The risk assessment module 500 analyzes the risk data of the enterprise to be predicted based on the enterprise risk assessment model to obtain the power risk assessment result, the production capacity risk assessment result and the production risk assessment result, and then obtains the risk prediction result.

[0087] Various changes and modifications can be made without departing from the spirit and scope of the present invention, and all equivalent technical solutions also belong to the scope of the present invention.

[0088] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0089] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, devices, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0090] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0091] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0092] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0093] It should be noted that: The "one embodiment" or "embodiment" mentioned in the specification means that the specific features, structures or characteristics described in conjunction with the embodiment are included in at least one embodiment of the present invention. Therefore, the phrases "one embodiment" or "embodiment" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0094] In addition, it should be noted that the shapes and names of the parts and components of the specific embodiments described in this specification may be different. Any equivalent or simple changes made based on the structure, features and principles described in the patent concept of the present invention are included in the protection scope of the patent of the present invention. The technicians in the technical field of the present invention can make various modifications or supplements to the specific embodiments described or replace them in a similar manner, as long as they do not deviate from the structure of the present invention or exceed the scope defined by the claims, they should all fall within the protection scope of the present invention.

Claims

1. A method for dynamic enterprise risk assessment based on multidimensional data, characterized in that: The following steps are involved: Acquire multi-dimensional enterprise-related data and perform preprocessing to obtain a preprocessing evaluation data set, wherein the multi-dimensional enterprise-related data includes power data, capacity data, and production data, and the preprocessing evaluation data includes power standard data, capacity standard data, and production standard data; Perform abnormal data analysis and feature extraction on the electric energy standard data to obtain an electric energy evaluation data set; perform abnormal data analysis and feature extraction on the production capacity standard data to obtain a production capacity evaluation data set, wherein the electric energy evaluation data includes electric energy abnormal data and electric energy feature data, and the production capacity evaluation data includes production capacity abnormal data and production capacity feature data; Establish an enterprise risk assessment pre-training model, the enterprise risk assessment pre-training model includes at least three risk assessment pre-training sub-models and a total output layer, each risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer, the data receiving layer receives an electric energy assessment data set, a production capacity assessment data set and a production standard data set respectively, the data processing layer performs corresponding processing on the electric energy assessment data, the production capacity assessment data and the production standard data respectively, the probability output layer obtains the electric energy risk assessment probability, the production capacity risk assessment probability and the production risk assessment probability respectively, and performs a comprehensive assessment through the total output layer to obtain risk prediction data; The risk assessment pre-trained sub-model is trained respectively through the electric energy assessment data set, the capacity assessment data set and the production standard data set to obtain the corresponding risk assessment sub-model, and the enterprise risk assessment model is formed through the corresponding risk assessment sub-model; Based on the enterprise risk assessment model, the risk data of the enterprise to be predicted is analyzed to obtain the power risk assessment results, production capacity risk assessment results and production risk assessment results, and then the risk prediction results are obtained.

2. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: The step of obtaining multi-dimensional enterprise-related data includes the following steps: According to the energy consumption data of the enterprise, the meter readings of the day and the previous day are obtained, and the electric energy data is obtained based on the collection rate data of different regions, which is expressed as follows: Obtain enterprise production management data and analyze it to obtain capacity data, where the capacity data includes order production data; Collect the video data produced by the enterprise to obtain production data, and form multi-dimensional enterprise-related data based on power data, capacity data and production data; in, Indicates Daily power data, Indicates the meter reading for the day. Indicates the meter reading of the previous day. Indicates the acquisition rate data. Indicates today.

3. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: The pre-processing comprises the following steps: Traverse the electric energy data and production capacity data, remove invalid and erroneous data, and obtain the removed electric energy data and production capacity data; The electric energy data after elimination is analyzed, and based on the maximum and minimum values ​​of the electric energy data after elimination, the electric energy data is standardized to obtain the electric energy standard data, which is expressed as follows: The capacity data after elimination is analyzed, and based on the maximum and minimum values ​​of the capacity data after elimination, the capacity data is standardized to obtain the capacity standard data, which is expressed as follows: Perform video segmentation on production data, remove erroneous data and duplicate data from the split production data, and annotate abnormal behaviors to obtain a production standard data set; in, Represents electrical energy data, Represents production capacity data, Indicates the standard data of electric energy. Indicates the minimum value of electric energy data. Indicates the maximum value of the electric energy data. Indicates the production capacity standard data, Indicates the minimum value of the capacity data, Indicates the maximum value of the capacity data.

4. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: The abnormal data analysis and feature extraction of the electric energy standard data includes the following steps: Preset power thresholds and obtain power average and power standard deviation based on historical power standard data; Based on the electric energy threshold, electric energy average value, electric energy standard deviation and current electric energy standard data, the electric energy analysis data is obtained, which is expressed as follows: If the power analysis data does not meet the power threshold, the current power standard data is abnormal power data; The electric energy standard data is analyzed to obtain daily average characteristic data of electric energy and monthly characteristic data of electric energy. Based on the monthly characteristic data of electric energy, the monthly average characteristic data of electric energy is obtained, which is expressed as follows: Based on the analysis of the current daily energy characteristic data and the previous day's energy characteristic data, the daily energy growth characteristic data is obtained, which is expressed as follows: Based on daily average energy characteristic data, monthly average energy characteristic data and daily energy growth characteristic data, electric energy characteristic data is formed; in, Indicates the power analysis data, Represents the average value of electric energy. represents the standard deviation of electric energy, Represents the average characteristic data of daily energy, Indicates the standard data of electric energy. Indicates that day, Indicates the total number of days. Indicates the average characteristic data of monthly electric energy, Indicates Monthly power characteristic data for the month, Indicates the total number of months, Indicates daily energy growth characteristic data, Indicates The standard data of electricity for the day, Indicates Daily electricity standard data.

5. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: The abnormal data analysis and feature extraction of the production capacity standard data includes the following steps: Preset capacity thresholds and obtain the capacity average and capacity standard deviation based on historical capacity standard data; Based on the capacity threshold, capacity average, capacity standard deviation and current capacity standard data, the capacity analysis data is obtained, which is expressed as follows: If the capacity analysis data does not meet the capacity threshold, the current capacity standard data is capacity abnormal data; Obtain and analyze the capacity standard data for multiple days to obtain the capacity characteristic data, which is expressed as follows: in, Indicates abnormal capacity data. Indicates that day, represents the average capacity, represents the standard deviation of production capacity, Represents capacity characteristic data, Indicates the total number of days. Indicates Daily production capacity standard data, Indicates capacity standard data.

6. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: The data receiving layer receives the electric energy evaluation data after windowing; The corresponding data processing layer includes a forget gate, an input gate, and an output gate. The forget gate obtains the power evaluation data and obtains the discarded information data through analysis, which is expressed as follows: The input gate analyzes the power evaluation data to obtain the updated information data and the current information status, which are expressed as follows: Get the long-term information status of the previous moment, combine the discarded information status, updated information status and current information status, and get the long-term information status of the current moment, as shown below: The output gate analyzes the power evaluation data to obtain the power output data. By analyzing the power output data and the long-term information state at the current moment, the sequence relationship data is obtained, which is expressed as follows: The probability output layer evaluates the sequence relationship data to obtain the probability of power risk assessment; in, Indicates discarding information data. represents the sigmoid function, represents the forget gate weight, Represents the sequence relationship data of the previous moment, Indicates the current power evaluation data, represents the forget gate bias, Indicates update information data. represents the input gate weight, represents the input gate bias, Indicates the current information status. represents the weight of the current state unit, represents the bias of the current state unit, Indicates the current long-term information status. Indicates the long-term information status of the previous moment. Indicates the power output data, represents the output gate weight, represents the output gate bias, Represents sequence relationship data.

7. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: The data receiving layer receives the capacity assessment data; The corresponding data processing layer includes an update gate and a reset gate. The capacity evaluation data is analyzed by updating the gate weight and resetting the gate weight to obtain the updated evaluation data and the reset evaluation data, which are expressed as follows: By resetting the evaluation data and the capacity dependency data of the previous moment for analysis, the candidate hidden state of the current moment is obtained, which is expressed as follows: Based on the updated evaluation data, the capacity dependency data at the previous moment and the candidate hidden state at the current moment, the capacity dependency data at the current moment is obtained, which is expressed as follows: The probability output layer analyzes the capacity dependency data to obtain the probability of capacity risk assessment; in, Indicates updating the evaluation data. represents the update gate weight, Indicates the capacity dependency data at the previous moment, Represents capacity assessment data, Indicates resetting the evaluation data. Resets the gate weights. represents the sigmoid function, represents the candidate hidden state at the current moment, represents the weight matrix, Indicates the capacity dependency data at the current moment, Indicates the current moment.

8. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: The data receiving layer receives the production standard data set and divides it into a training set, a test set and a validation set; The corresponding data processing layer includes a backbone network layer and a neck network layer. The backbone network layer extracts features of the production standard data, including edge features, texture features and shape features, and reduces the complexity and parameter amount of feature calculation through a convolution module and a residual module; the neck network layer obtains image features at different stages of the backbone network and fuses them to obtain fused image features, and enhances the fused image features to obtain enhanced image features; The probability output layer performs analysis and evaluation based on the enhanced image features to obtain the production risk assessment probability.

9. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: The comprehensive assessment is performed to obtain risk prediction data, specifically: Set the model weight parameters, analyze the power risk assessment probability, capacity risk assessment probability and production risk assessment probability based on the model weight parameters, and obtain risk prediction data, which is expressed as follows: in, represents risk prediction data, represents the model weight parameters of the risk assessment pre-trained sub-model, represents the probability of electric energy risk assessment, the probability of capacity risk assessment, or the probability of production risk assessment, Represents the number of risk assessment pre-trained sub-models.

10. The enterprise risk dynamic assessment method based on multidimensional data according to claim 1 is characterized in that: It also includes cross-validation and evaluation of the risk assessment sub-model, specifically: Get the evaluation indicators of the risk assessment sub-model, where the evaluation indicators include accuracy, recall, F1 score and AUC-ROC, as shown below: Model subsets are established through the power evaluation data set, capacity evaluation data set and production standard data set respectively, and the risk assessment sub-model is cross-validated based on the model subset and evaluation indicators, as shown below: in, Indicates accuracy, represents true positive data, represents true negative data, represents false positive data, represents false negative data, represents the recall rate, express Score value, express Curve area data, represents the cross validation score, represents the number of model subsets, Represents the risk assessment sub-model The evaluation indicators are Indicates that through The risk assessment sub-model obtained by training the model subsets is Indicates the true proportion, represents the false positive proportion.

11. A dynamic enterprise risk assessment system based on multidimensional data, characterized in that: It includes data preprocessing module, data extraction module, model building module, model training module and risk assessment module; The data preprocessing module acquires and preprocesses multi-dimensional enterprise-related data to obtain a preprocessed evaluation data set, wherein the multi-dimensional enterprise-related data includes power data, capacity data, and production data, and the preprocessed evaluation data includes power standard data, capacity standard data, and production standard data; The data extraction module performs abnormal data analysis and feature extraction on the electric energy standard data to obtain an electric energy evaluation data set; performs abnormal data analysis and feature extraction on the production capacity standard data to obtain a production capacity evaluation data set, wherein the electric energy evaluation data includes electric energy abnormal data and electric energy feature data, and the production capacity evaluation data includes production capacity abnormal data and production capacity feature data; The model building module builds an enterprise risk assessment pre-training model, which includes at least three risk assessment pre-training sub-models and a total output layer. Each risk assessment pre-training sub-model includes a data receiving layer, a data processing layer and a probability output layer. The data receiving layer receives an electric energy assessment data set, a production capacity assessment data set and a production standard data set respectively. The data processing layer processes the electric energy assessment data, the production capacity assessment data and the production standard data respectively. The probability output layer obtains the electric energy risk assessment probability, the production capacity risk assessment probability and the production risk assessment probability respectively, and performs a comprehensive assessment through the total output layer to obtain risk prediction data; The model training module trains the risk assessment pre-training sub-model through the electric energy assessment data set, the capacity assessment data set and the production standard data set to obtain the corresponding risk assessment sub-model, and forms the enterprise risk assessment model through the corresponding risk assessment sub-model; The risk assessment module analyzes the risk data of the enterprise to be predicted based on the enterprise risk assessment model, obtains the power risk assessment results, the production capacity risk assessment results and the production risk assessment results, and then obtains the risk prediction results.

12. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 10 is implemented.

13. A device for dynamic enterprise risk assessment based on multidimensional data, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 10 is implemented.

Images