CN115936243A - Enterprise stability evaluation method, device, equipment and storage medium - Google Patents
Enterprise stability evaluation method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN115936243A CN115936243A CN202211676264.XA CN202211676264A CN115936243A CN 115936243 A CN115936243 A CN 115936243A CN 202211676264 A CN202211676264 A CN 202211676264A CN 115936243 A CN115936243 A CN 115936243A
- Authority
- CN
- China
- Prior art keywords
- enterprise
- data
- model
- target
- basic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses an enterprise stability evaluation method, device, equipment and storage medium, wherein power utilization data of a target enterprise to be evaluated are input into a plurality of basic models and are simultaneously processed to obtain an evaluation result output by each basic model and aiming at the target enterprise, and the evaluation results are subjected to weighted calculation to obtain a target evaluation result of the target enterprise. The basic model is obtained based on power consumption data of a large number of enterprises and enterprise stability data training, the target enterprise stability can be accurately predicted, and the evaluation results of all the target enterprises obtained by the basic model are subjected to weighting processing, so that the evaluation results are optimized, and the most accurate target evaluation result corresponding to the target enterprise is obtained.
Description
Technical Field
The present application relates to the field of data processing technologies, and in particular, to a method, an apparatus, a device, and a storage medium for evaluating enterprise stability.
Background
The method has an important role in assessing the operation stability of the enterprises before and after the financial wind-controlled loan is controlled, and in joint modeling and financial wind-controlled loan control. Most of the traditional enterprise stability judgment models adopt machine learning models or deep learning models such as linear regression, logistic regression, decision trees, RNN (probabilistic neural network) and the like, but the traditional machine learning models have high data requirement and low prediction accuracy, and the final enterprise stability prediction results obtained by the same sample data are different due to the fact that different machine learning models or deep learning models have different defects, so that the accuracy is greatly reduced.
Disclosure of Invention
In view of this, the present application provides an enterprise stability evaluation method, apparatus, device and storage medium, which are used to solve the problems of single model and inaccurate evaluation in the existing enterprise stability evaluation mode.
In order to achieve the above object, the following solutions are proposed:
an enterprise stability evaluation method comprises the following steps:
collecting power consumption data of a target enterprise to be evaluated;
respectively inputting the electricity utilization data into at least two basic models for processing to obtain an evaluation result output by each basic model and aiming at the target enterprise, wherein the training target values of the basic models are the same, and the model structures of the basic models are different;
and performing weighted calculation on the evaluation result aiming at the target enterprise and output by each basic model to obtain a target evaluation result corresponding to the target enterprise.
Preferably, before the inputting the electricity data into at least two basic models respectively for processing to obtain the evaluation result output by each basic model for the target other enterprises, the method further includes:
selecting a model structure of the basic model according to the functional characteristics corresponding to the basic model to obtain at least two blank basic models;
and modeling based on the at least two blank basic models to obtain at least two basic models, wherein the basic models are models obtained by training based on the electricity utilization data of enterprises as characteristic values and the stability evaluation data of the enterprises as target values.
Preferably, the modeling based on the at least two blank base models to obtain at least two base models includes:
acquiring power utilization data and operation data of a plurality of enterprises;
performing data preprocessing on the electricity utilization data and the operation data to obtain a data set corresponding to each enterprise, wherein the data set comprises electricity fee fluctuation data and enterprise operation stability evaluation data corresponding to the enterprises;
randomly dividing the data set into a training set and a testing set;
and respectively training and analyzing each blank basic model based on the training set and the test set to obtain a basic model for outputting enterprise stability evaluation data based on the electric charge fluctuation data of the enterprise, wherein model structures corresponding to the blank basic model are at least two of a nonlinear autoregressive network with external input, long-term and short-term memory and a threshold cycle unit.
Preferably, the performing data preprocessing on the electricity consumption data and the operation data to obtain data corresponding to each enterprise includes:
extracting the electric charge fluctuation condition of the enterprise from the electricity consumption data;
extracting enterprise stability evaluation data of the enterprise from the business data;
and combining the electric charge fluctuation condition and the enterprise stability evaluation data corresponding to each enterprise to obtain a data set corresponding to each enterprise.
Preferably, the method further comprises the following steps:
inputting the electricity utilization data of the multiple enterprises into the trained basic model to obtain the evaluation result output by the basic model and corresponding to each enterprise;
combining the evaluation results output by each basic model of each enterprise to obtain a combined evaluation value corresponding to each enterprise;
modeling based on a Markov strategy process to obtain a weight optimization model;
inputting the combined evaluation value into the weight optimization model to perform weight optimization to obtain a weight matrix corresponding to each enterprise;
determining a target weight matrix from the weight matrix corresponding to each enterprise;
and modeling based on the target weight matrix to obtain a comprehensive calculation model for performing weighted calculation on the evaluation result output by each basic model and aiming at the target enterprise.
Preferably, the performing weighted calculation on the evaluation result output by each base model for the target enterprise includes:
inputting the evaluation result output by each basic model and aiming at the target enterprise into a comprehensive calculation model;
and carrying out weighted summation processing on the evaluation results based on the calculation parameters of the comprehensive calculation model to obtain target evaluation results corresponding to the target enterprises.
An enterprise stability evaluation device comprising:
the data acquisition unit is used for acquiring the electricity utilization data of a target enterprise to be evaluated;
the data model processing unit is used for respectively inputting the power utilization data into at least two basic models to be processed to obtain an evaluation result output by each basic model and aiming at the target enterprise, the training target values of the basic models are the same, and the model structures of the basic models are different;
and the evaluation result acquisition unit is used for performing weighted calculation on the evaluation result output by each basic model and aiming at the target enterprise to obtain a target evaluation result corresponding to the target enterprise.
Preferably, before executing the data model processing unit, the method further comprises:
the basic model selection unit is used for selecting the model result of the basic model according to the functional characteristics corresponding to the basic model to obtain at least two blank basic models;
and the basic model establishing unit is used for modeling based on the at least two blank basic models to obtain at least two basic models, wherein the basic models are models obtained by training based on the electricity utilization data of enterprises as characteristic values and the stability evaluation data of the enterprises as target values.
An enterprise stability evaluation device comprising: a memory and a processor;
wherein the processor is configured to execute a program stored in the memory;
the memory is used for storing programs, and the programs are at least used for realizing the enterprise stability evaluation method.
A storage medium stores computer-executable instructions, and the computer-executable instructions are loaded and executed by a processor to realize the enterprise stability evaluation method.
According to the technical scheme, the enterprise stability evaluation method, the enterprise stability evaluation device, the enterprise stability evaluation equipment and the storage medium provided by the embodiment of the application have the advantages that the power utilization data of the target enterprise to be evaluated are input into the plurality of basic models and are simultaneously processed, the evaluation result output by each basic model and aiming at the target enterprise is obtained, the evaluation results are subjected to weighted calculation, and the target evaluation result of the target enterprise is obtained. The basic model is obtained based on power consumption data of a large number of enterprises and enterprise stability data training, the target enterprise stability can be accurately predicted, and the evaluation results of all the target enterprises obtained by the basic model are subjected to weighting processing, so that the evaluation results are optimized, and the most accurate target evaluation result corresponding to the target enterprise is obtained.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the embodiments or the prior art descriptions will be briefly described below, it is obvious that the drawings in the following description are only the embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a schematic flowchart of an enterprise stability evaluation method provided in an embodiment of the present application;
fig. 2 is a schematic structural diagram of an enterprise stability evaluation device according to an embodiment of the present disclosure;
fig. 3 is a schematic structural diagram of an enterprise stability evaluation device according to an embodiment of the present application.
Detailed Description
The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
Referring to fig. 1, a schematic flow chart of a method for implementing enterprise stability evaluation provided by an embodiment of the present application is shown, where the method may include the following steps:
and S110, collecting power utilization data of the target enterprise to be evaluated.
The power utilization data of the target enterprise can be acquired through various ways, for example, the power utilization data of the target enterprise can be called from a power grid system, and the power utilization data actively uploaded by the enterprise can be directly acquired for evaluating the stability of the enterprise. The electricity utilization data can include data information such as enterprise electricity fee level, enterprise payment behavior, enterprise electricity utilization specification and the like, and the information can reflect enterprise stability, fund state and the like of the enterprise to a certain extent. Generally, the directly acquired electricity consumption data can be classified based on the index grade, and the electricity consumption data effective to the scheme can be extracted from the grade-based index during collection.
And step S120, inputting the electricity utilization data into at least two basic models to obtain an evaluation result output by each basic model.
Specifically, the electricity utilization data is input into at least two basic models, and an evaluation result output by each basic model and specific to the target enterprise is obtained. The training targets of the basic models are the same, and the model structures of the basic models are different.
It can be understood that each of the trained basic models is used for evaluating the enterprise stability of the target enterprise based on the electricity utilization data of the target enterprise. However, differences exist among different model structures, and results obtained by training different enterprise models have advantages and disadvantages, and if only a single basic model is adopted to evaluate a target enterprise, the accuracy of the finally obtained evaluation result is low. A plurality of basic models can be trained simultaneously, the power utilization data of the same target enterprise is evaluated, and the evaluation result of each basic model on the target enterprise is obtained. And further calculating the evaluation result to obtain a final target evaluation result.
The different basic models have advantages and disadvantages, and can be trained by selecting a proper basic model according to requirements on a data processing process or by randomly selecting a plurality of groups of basic models, model evaluation is performed on the trained basic models, and a group of basic models with the optimal model evaluation result is determined as a basic model group for evaluating enterprises by the scheme. Based on this, the basic model in the scheme is not a specific model and can be a multi-choice and variable model.
And S130, carrying out weighting processing on the evaluation result to obtain a target evaluation result of the target enterprise.
And weighting each evaluation result according to the data advantages, and performing weighted summation processing on all the evaluation results to obtain the target evaluation result corresponding to the target enterprise.
In order to ensure that the evaluation result output by each basic model is subjected to weighted summation processing, and the obtained target evaluation result is an optimal result, an optimal weight distribution mode can be determined in an exhaustive mode, and an optimal weighting mode can also be determined in a weight optimization model mode and the like, so that the target evaluation result obtained after weighted summation processing is most accurate.
According to the embodiment of the application, the power utilization data of the target enterprise are input into at least two basic models to be processed, the evaluation result of each basic model for the enterprise stability of the target enterprise is obtained, and the target evaluation result of the enterprise stability of the target enterprise is obtained by performing weighted summation on all the evaluation results. Through the processing of a plurality of basic models, the electricity utilization data of the target enterprise can be evaluated from multiple aspects, the diversity of data evaluation is increased, and the problems that the data processing is not comprehensive and the obtained evaluation result is not accurate due to the fact that a single evaluation model is used for evaluating are solved. Meanwhile, weighting and summing are performed on the obtained multiple evaluation results in a weighted mode, the optimal target evaluation result is obtained, and the obtained target evaluation result is more accurate.
Next, the enterprise stability evaluation method is further introduced in the embodiments of the present application.
In the embodiment of the application, the power utilization data of the target enterprise needs to be processed by at least two basic models respectively to obtain the evaluation result of each basic model for the target enterprise, wherein the basic models can be selected in various ways.
Selecting a model structure of the basic model according to the functional characteristics corresponding to the basic model to obtain at least two blank basic models; and modeling based on the at least two blank basic models to obtain at least two basic models, wherein the basic models are models obtained by training based on the electricity utilization data of the enterprises as characteristic values and the stability evaluation data of the enterprises as target values.
Different basic models have different emphasis points on data processing contained in sample data and training sets, and prediction capabilities of prediction models obtained by training different basic models are different to a certain extent. In the embodiment of the application, three basic models are selected, namely a nonlinear autoregressive Network (NARX) with external source input, a long-short term memory (LSTM) and a threshold cycle unit (GRU), and the three basic models are utilized to construct an N-L-G combined type prediction model for predicting the enterprise stability of a target enterprise. The three basic models have the capability of predicting the operation stability of the enterprise by using the power utilization data of the enterprise to a certain extent.
The three basic models have advantages and disadvantages respectively, and the memory capacity of historical data is enhanced by using a prediction model obtained by NARX training; the LSTM can effectively transmit and express information in a long-time sequence, can not cause useful information to be ignored before a long time, and can also solve the problem of gradient disappearance or explosion in the RNN; the GRU is also used for solving the problems of long-term memory, gradient in back propagation and the like, and compared with the LSTM, the GRU is easier to train and can improve the training efficiency to the maximum extent. Based on the advantages and disadvantages of the basic model as described above as an example, it can be understood that how to select an appropriate basic model is determined based on the data processing emphasis of the final prediction model and consideration of the model processing result.
After obtaining the blank basic model, modeling and training are performed on the basis of the blank basic model to obtain a basic model capable of evaluating enterprise stability, which specifically includes:
acquiring power consumption data and operation data of a plurality of enterprises, and performing data preprocessing on the power consumption data and the operation data to obtain a data set corresponding to each enterprise, wherein the data set comprises power rate fluctuation data and enterprise operation stability evaluation data corresponding to the enterprises; randomly dividing the data set into a training set and a testing set; and respectively training and analyzing each blank basic model based on the training set and the testing set to obtain a basic model for outputting enterprise stability evaluation data based on enterprise electric charge fluctuation data.
The training model needs a large amount of sample data for training, and the power utilization data and the enterprise operation data of a plurality of enterprises can be obtained through various ways. The power consumption data can call power consumption data information of an enterprise from a power grid system, wherein the operation stability of the enterprise is mainly predicted through the power consumption data of the enterprise in the embodiment of the application, and the power consumption data of the enterprise can be extracted from the power consumption data. The electricity charge data can be collected according to grade indexes, including: the enterprise electric charge level, the enterprise behavior of collecting fee and enterprise's power consumption action etc. first grade index, second grade index such as first grade receivable/real charge electric charge level, receivable/real charge fluctuation condition and electric charge recovery level, and the data condition is shown as following table 1:
TABLE 1 Enterprise Electricity Charge data Condition
The enterprise operation data information can be obtained from a government affair system, and the operation data of the enterprise mainly comprises an enterprise industry status, an enterprise operation stability, an enterprise development trend, an enterprise capital state, an enterprise integrity level and the like, wherein different enterprise operation data can be applied to different prediction scenes, and specifically refer to table 2 as follows:
TABLE 2 prediction scenarios corresponding to different business data
| Business data | Predicting scenes |
| Status of industry | In financial wind-control-loan |
| Stability of operation | In financial wind-control-loan, after joint modeling and financial wind-control-loan |
| Trend of development | Joint modeling in financial wind control-loan |
| Status of funds | Precision marketing (white list) |
| Internal management | Financial wind-loan accurate marketing |
The required operation data can be selected as a target value for training according to different prediction purposes, and the electricity fee data and the operation stability training basic model are selected and used for predicting the operation stability of the enterprise.
Specifically, the obtained electricity charge data and operation data need to be preprocessed to obtain a data set capable of training a basic model, and the method comprises the following steps:
extracting the electricity charge fluctuation condition of the enterprise from the electricity consumption data; extracting enterprise operation stability data of the enterprise from the operation data; and combining the electric charge fluctuation condition corresponding to each enterprise and the enterprise operation stability evaluation data to obtain a data set corresponding to each enterprise.
The receivable or real electric charge fluctuation condition data of the enterprise can be extracted from the electric charge data of the electric power consumption data corresponding to each enterprise, and classification acquisition can be performed according to different indexes, wherein the three-level indexes can be included: the fluctuation condition of the electric charge which is charged in the last 3, 6 and 12 months and the fluctuation condition of the electric charge which is charged in the last 3, 6 and 12 months. And extracting the annual operation stability evaluation data of the enterprises from the operation data of the enterprises, and taking the data as the operation stability evaluation data of the enterprises.
The electric charge fluctuation data and the operation stability evaluation data are combined and processed into data sets corresponding to each enterprise by taking the enterprise as a unit, wherein each data set comprises the electric charge data characteristics of the enterprise, and the operation stability evaluation data of the enterprise in the last year are marked, so that the enterprise is directly used as the unit for training when a basic model is conveniently trained subsequently, and the data confusion among the enterprises is avoided. The structure of a particular data set can be referred to in table 3, as follows:
and training the basic model based on the data set, wherein the data set is required to be randomly divided into a training set and a testing set, wherein the training set is used for training the basic model, so that the basic model obtained by training can obtain the corresponding operation stability evaluation data of the enterprise through the power utilization data of the enterprise. And the test set is used for testing the accuracy of the prediction of the basic model obtained by training the training set and carrying out inspection.
And training based on the fluctuation condition of the electric charge of the enterprises in the training set as a characteristic value and the operation stability evaluation data as a target value to obtain at least two basic models, wherein each basic model can predict the operation stability evaluation of the enterprises according to the input electric power consumption data.
Each basic model can output an evaluation result corresponding to the target enterprise according to the input power utilization data of the target enterprise, but the evaluation results are multiple, and the evaluation results need to be further processed to obtain a comprehensive evaluation result, so that the operation stability of the target enterprise can be more visually represented. According to the embodiment of the application, the evaluation result output by each basic model and aiming at the target enterprise is subjected to weighted calculation to obtain the target evaluation result corresponding to the target enterprise. In order to obtain a more accurate target evaluation result, before this step, a weighted calculation method that optimizes the evaluation result needs to be found so that an optimal target evaluation result can be obtained.
Specifically, electricity consumption data of a plurality of enterprises can be input into a trained basic model to obtain an evaluation result output by the basic model and corresponding to each enterprise; dividing the evaluation result output by each basic model into combined evaluation values corresponding to each enterprise by taking the enterprise as a unit; modeling based on a Markov strategy process to obtain a weight optimization model, and inputting the combined evaluation value into the weight optimization model for weight optimization to obtain a weight matrix corresponding to each enterprise; determining a target weight matrix from the weight matrix corresponding to each enterprise; and modeling based on the target weight matrix to obtain a comprehensive calculation model for carrying out weighted calculation on the evaluation result which is output by each basic model and aims at the target enterprise.
According to the embodiment of the application, weighted calculation needs to be carried out on the evaluation results output by each basic model aiming at the target enterprise, and therefore a large number of evaluation results output by the basic models are needed to be used as samples when the weight optimization model is trained. The data corresponding to the enterprises used for training the basic model can be reused, the electricity consumption data of each enterprise is input into the trained basic model, the evaluation result of the trained basic model for each enterprise is obtained, and all the evaluation results are divided into combined evaluation values corresponding to each enterprise by taking the enterprise as a unit and are used as a data set of the training weight optimization model.
The method and the device for optimizing the weights are based on a Markov decision process, and model building is conducted on a state space S, an action space A and a reward function R to obtain a weight optimization model. Wherein, the evaluation results output by the three basic models of the nonlinear autoregressive Network (NARX) with external source input, the long-short term memory (LSTM) and the threshold cycle unit (GRU) are respectively represented as S 1 、S 2 、S 3 。
The state space S contains the predictors of the respective underlying modelMeasuring the weight of the result, e.g. S = [ w = [ ] 1 ,w 2 ,w 3 ]. Wherein, w 1 、w 2 、w 3 The prediction weights of the evaluation results S1, S2, S3 of the three basic models, and the initial state S 0 Is set as [1/3,1/3 ]]。
The motion space matrix a includes the motion of increasing and decreasing the prediction weight of each base model. A motion matrix of 3 rows and 2 columns is established as follows:
where Δ w represents the magnitude of the weight that increases or decreases each time an action is performed.
The immediate reward function R obtained after each execution of an action is set as follows:
wherein T represents the Tth execution of the action,
the prediction weights of the evaluation results of the basic models NARX and LSTM and GRU model after the execution of Tth operation are respectively shown. />
Respectively representing the predicted values of the three models, y t Representing the actual value and N representing the number of samples of the training set. When the action is executed, if the average absolute error (MAE) of the combined evaluation result is less than the last result, a certain reward is obtained, namely when MAE (T + 1)<The additional reward of custom k is given to MAE (T) in order to avoid the problem of reward sparseness when MAE promotion is small.
And inputting the combined evaluation value corresponding to each enterprise as a training sample into the weight optimization model for training to obtain an optimal weight matrix. The weight optimization model adopts an exploration strategy of epsilon-greedy strategy, is used for training to obtain an optimal weight matrix and an optimal strategy of the weight optimization model, and specifically may include:
initializing current model parameters may include: initializing a current value network parameter theta, a target value network parameter theta' = theta, an updating step alpha, a batch updating size BS, an experience pool B, a discount factor gamma, a search probability epsilon, a training time step =0 and an action space A. For each weight optimization task, the initialization environment state s = s 0 And changing actions of the combined evaluation value weight optimization result according to each optimization model.
Randomly generating a probability p ∈ [0,1 ]]If p is less than epsilon, then an action a is randomly selected t Otherwise, the current state s t Inputting the current value network to obtain a with the maximum Q value t . Execution of a t Changing s, calculating y to obtain a sample (s, a) t ,r,s t+1 ) Storing in an experience pool B, if the number of samples in the experience pool B is more than BS, randomly removing BS samples from B, and obtaining (s, a) each sample t ,r,s t+1 ) Predicting a next state optimum action
Calculating TD-error, i.e. </R>
According to the loss L (theta) = delta 2 2, training in batch to update a current value network parameter theta, and adding 1 to the network update times step; if mod (step, a) =0, the target value network parameter θ' = θ is updated.
And the weight optimization model is trained and optimized through the steps based on the training samples, and when the weight optimization model meets the requirement of task training times, an optimal strategy and an optimal weight matrix obtained after optimization are output. However, the optimal weight matrix is corresponding to the sample data used in the training, that is, each training sample has a corresponding optimal weight matrix. The range of an optimal weight matrix can be determined by counting all the optimal weight matrices, and a determined weight matrix or a determined weight matrix range is obtained as a target weight matrix. And modeling based on the target weight matrix to obtain a comprehensive calculation model. And the comprehensive calculation model is used for performing weighted calculation on the evaluation result aiming at the target enterprise and output by each basic model to obtain a target evaluation result corresponding to the target enterprise.
Specifically, the evaluation results output by each basic model and specific to the target enterprise are input into a comprehensive calculation model, and the comprehensive calculation model performs corresponding weighted summation processing on the evaluation results according to the parameters of the optimal weight matrix to obtain the target evaluation results corresponding to the target enterprise.
The following is an example of practical application provided in the real-time example of the present application, where a bank usually encounters a situation that an enterprise borrows money from the bank, which involves a large amount of money and a long loan term. The bank thus needs to evaluate the enterprise applying for the loan in order to make the most accurate decision. Among them, the business stability of the enterprise is one of the most important reference certificates.
The electric power data can reflect the production conditions, the operation conditions and the like of enterprises, a bank can obtain the electricity utilization data of a target loan enterprise by applying to a power grid system, the electricity utilization data can be further processed to extract useful electricity charge fluctuation data, or the electricity utilization data can be directly input into at least two basic models for data processing to obtain the evaluation result of each basic model aiming at the target loan enterprise, and the basic models can select a proper basic model structure for modeling according to the data bias of the bank for enterprise evaluation in the initial modeling period, so that the situation that prediction and evaluation are performed based on the same data are too single, and the obtained evaluation result is not accurate is prevented.
And calculating each evaluation result of each basic model aiming at the target bandwidth enterprise according to the optimal weight matrix, calculating each evaluation result according to the corresponding weight distribution, and summing the weighted structures of each evaluation result to obtain the final target evaluation result corresponding to the target bandwidth enterprise, wherein the target evaluation structure can reflect the operation stability of the enterprise, and can be referred by a bank based on the target evaluation result so as to make the most correct decision aiming at the target bandwidth enterprise.
The enterprise stability evaluation device provided by the embodiment of the present application is described below, and the enterprise stability evaluation device described below and the enterprise stability evaluation method described above are referred to in correspondence with each other.
First, referring to fig. 2, an enterprise stability evaluation device is described, and as shown in fig. 2, the enterprise stability evaluation device may include:
the data acquisition unit 100 is used for acquiring power consumption data of a target enterprise to be evaluated;
the data model processing unit 200 is configured to input the power consumption data into at least two basic models respectively for processing, so as to obtain an evaluation result output by each basic model for the target enterprise, where training target values of the basic models are the same, and model structures of the basic models are different;
the evaluation result obtaining unit 300 is configured to perform weighted calculation on the evaluation result for the target enterprise output by each base model to obtain a target evaluation result corresponding to the target enterprise.
Preferably, before executing the data model processing unit 200, the method further includes:
the basic model selection unit is used for selecting the model structure of the basic model according to the functional characteristics corresponding to the basic model to obtain at least two blank basic models;
and the basic model establishing unit is used for establishing a model based on the at least two blank basic models to obtain at least two basic models, wherein the basic models are models obtained by training based on the electricity utilization data of the enterprise as characteristic values and the stability evaluation data of the enterprise as target values.
Preferably, the basic model building unit comprises
The system comprises a sample data acquisition subunit, a data processing subunit and a data processing subunit, wherein the sample data acquisition subunit is used for acquiring power utilization data and operation data of a plurality of enterprises;
the data preprocessing subunit is used for performing data preprocessing on the electricity utilization data and the operation data to obtain a data set corresponding to each enterprise, wherein the data set comprises electricity fee fluctuation data and enterprise operation stability evaluation data corresponding to the enterprises;
the data dividing subunit is used for randomly dividing the data set into a training set and a testing set;
and the model training subunit is used for respectively training and analyzing each blank basic model based on the training set and the test set to obtain a basic model for outputting enterprise stability evaluation data based on the electric charge fluctuation data of the enterprise, and the model structures corresponding to the blank basic model are at least two of a nonlinear autoregressive network with external input, a long-term and short-term memory and a threshold cycle unit.
Preferably, the data preprocessing subunit includes:
the electricity charge data extraction subunit is used for extracting the electricity charge fluctuation condition of the enterprise from the electricity consumption data;
the business data extraction subunit is used for extracting enterprise stability evaluation data of the enterprise from the business data;
and the data set acquisition subunit is used for combining the electric charge fluctuation condition and the enterprise stability evaluation data corresponding to each enterprise to obtain a data set corresponding to each enterprise.
Preferably, the method further comprises the following steps:
the evaluation result acquisition subunit is used for inputting the electricity utilization data of a plurality of enterprises into the trained basic model to obtain the evaluation result output by the basic model and corresponding to each enterprise;
the evaluation result processing subunit is used for combining the evaluation results output by each basic model of each enterprise to obtain a combined evaluation value corresponding to each enterprise;
the modeling subunit is used for modeling based on a Markov decision process to obtain a weight optimization model;
a weight matrix obtaining subunit, configured to input the combined evaluation value into the weight optimization model to perform weight optimization, so as to obtain a weight matrix corresponding to each enterprise;
a target weight matrix determining subunit, configured to determine a target weight matrix from the weight matrix corresponding to each enterprise;
and the comprehensive calculation model obtaining subunit is used for modeling based on the target weight matrix to obtain a comprehensive calculation model for performing weighted calculation on the evaluation result output by each basic model and aiming at the target enterprise.
Preferably, the evaluation result acquiring unit 300 includes:
the evaluation result output subunit is used for inputting the evaluation result output by each basic model and aiming at the target enterprise into the comprehensive calculation model;
and the evaluation result processing subunit is used for carrying out weighted summation processing on the evaluation results based on the calculation parameters of the comprehensive calculation model to obtain target evaluation results corresponding to the target enterprises.
According to the method and the device, the power utilization data of the target enterprise are input into the at least two basic models to be processed, the evaluation result of each basic model for the enterprise stability of the target enterprise is obtained, and the target evaluation result of the enterprise stability of the target enterprise is obtained by performing weighted summation processing on all the evaluation results. Through the processing of a plurality of basic models, the power utilization data of the target enterprise can be evaluated in multiple ways, the diversity of data evaluation is increased, and the problems that the data processing is not comprehensive and the obtained evaluation result is not accurate due to the fact that a single evaluation model is used for evaluation are solved. Meanwhile, weighting and summing are carried out on the obtained multiple evaluation results with emphasis, so that the optimal target evaluation result is obtained, and the obtained target evaluation result is more accurate.
The enterprise stability evaluation device provided by the embodiment of the application can be applied to enterprise stability evaluation equipment, and the enterprise stability evaluation equipment can be a network server side and also can be terminal equipment. Fig. 3 shows a schematic structural diagram of an enterprise stability evaluation device, and referring to fig. 3, the enterprise stability evaluation device may include: at least one processor 10, at least one memory 20, at least one communication bus 30, and at least one communication interface.
In the embodiment of the present application, the number of the processor 10, the memory 20, the communication bus 30 and the communication interface 40 is at least one, and the processor 10, the memory 20 and the communication interface 40 all communicate with each other through the communication bus 30.
The processor 10 may be a central processing unit CPU, or may be an Application Specific Integrated Circuit ASIC (Application Specific Integrated Circuit), or may be one or more Integrated circuits configured to implement the embodiments of the present Application, or the like.
The memory 20 may comprise a high-speed RAM memory, and may further comprise a non-volatile memory or the like, such as at least one disk memory.
The processor can call the program stored in the memory, and the program is used for realizing the enterprise stability evaluation method.
Embodiments of the present application also provide a storage medium that may store computer-executable instructions that are loaded and executed by a processor. The computer-executable instructions may be in the form of a computer program for implementing the various steps of the enterprise stability evaluation method described above.
Finally, it should also be noted that, in this document, relational terms such as first and second, and the like are used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrases "comprising a," "8230," "8230," or "comprising" does not exclude the presence of additional like elements in a process, method, article, or apparatus that comprises the element.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the application. Thus, the present application is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (10)
1. An enterprise stability evaluation method is characterized by comprising the following steps:
collecting power utilization data of a target enterprise to be evaluated;
respectively inputting the electricity utilization data into at least two basic models for processing to obtain an evaluation result output by each basic model and aiming at the target enterprise, wherein the training target values of the basic models are the same, and the model structures of the basic models are different;
and performing weighted calculation on the evaluation result output by each basic model and aiming at the target enterprise to obtain a target evaluation result corresponding to the target enterprise.
2. The method according to claim 1, before the inputting the electricity consumption data into at least two basic models respectively for processing, and obtaining the evaluation result for the target enterprise output by each basic model, further comprising:
selecting a model structure of the basic model according to the functional characteristics corresponding to the basic model to obtain at least two blank basic models;
and modeling based on the at least two blank basic models to obtain at least two basic models, wherein the basic models are models obtained by training based on the electricity utilization data of enterprises as characteristic values and the stability evaluation data of the enterprises as target values.
3. The method of claim 2, wherein modeling based on the at least two blank basis models results in at least two basis models, including
Acquiring power utilization data and operation data of a plurality of enterprises;
performing data preprocessing on the electricity utilization data and the operation data to obtain a data set corresponding to each enterprise, wherein the data set comprises electricity charge fluctuation data and enterprise operation stability evaluation data corresponding to the enterprises;
randomly dividing the data set into a training set and a testing set;
and respectively training and analyzing each blank basic model based on the training set and the testing set to obtain a basic model for outputting enterprise stability evaluation data based on the electric charge fluctuation data of the enterprise, wherein model structures corresponding to the blank basic model are at least two of a nonlinear autoregressive network with external input, long-short term memory and a threshold cycle unit.
4. The method of claim 3, wherein the pre-processing the electricity consumption data and the business data to obtain a data set corresponding to each enterprise comprises:
extracting the electricity charge fluctuation condition of the enterprise from the electricity consumption data;
extracting enterprise stability evaluation data of the enterprise from the business data;
and combining the electric charge fluctuation condition and the enterprise stability evaluation data corresponding to each enterprise to obtain a data set corresponding to each enterprise.
5. The method of claim 3, further comprising:
inputting the electricity utilization data of a plurality of enterprises into the trained basic model to obtain the evaluation result output by the basic model and corresponding to each enterprise;
combining the evaluation results output by each basic model of each enterprise to obtain a combined evaluation value corresponding to each enterprise;
modeling based on a Markov decision process to obtain a weight optimization model;
inputting the combined evaluation value into the weight optimization model to perform weight optimization to obtain a weight matrix corresponding to each enterprise;
determining a target weight matrix from the weight matrix corresponding to each enterprise;
and modeling based on the target weight matrix to obtain a comprehensive calculation model for performing weighted calculation on the evaluation result output by each basic model and aiming at the target enterprise.
6. The method of claim 5, wherein the performing a weighted calculation on the evaluation result for the target enterprise output by each base model comprises:
inputting the evaluation result output by each basic model and aiming at the target enterprise into a comprehensive calculation model;
and carrying out weighted summation processing on the evaluation results based on the calculation parameters of the comprehensive calculation model to obtain target evaluation results corresponding to the target enterprises.
7. An enterprise stability evaluation device, comprising:
the data acquisition unit is used for acquiring the electricity utilization data of a target enterprise to be evaluated;
the data model processing unit is used for respectively inputting the power utilization data into at least two basic models to be processed to obtain an evaluation result output by each basic model and aiming at the target enterprise, the training target values of the basic models are the same, and the model structures of the basic models are different;
and the evaluation result acquisition unit is used for performing weighted calculation on the evaluation result output by each basic model and aiming at the target enterprise to obtain a target evaluation result corresponding to the target enterprise.
8. The method of claim 7, further comprising, prior to executing the data model processing unit:
the basic model selection unit is used for selecting the model structure of the basic model according to the functional characteristics corresponding to the basic model to obtain at least two blank basic models;
and the basic model establishing unit is used for establishing a model based on the at least two blank basic models to obtain at least two basic models, wherein the basic models are models obtained by training based on the electricity utilization data of the enterprise as characteristic values and the stability evaluation data of the enterprise as target values.
9. An enterprise stability evaluation device, comprising: a memory and a processor;
the processor is used for executing the program stored in the memory;
the memory is configured to store a program for implementing at least the enterprise stability evaluation method of any of claims 1-6.
10. A storage medium having stored therein computer-executable instructions, which are loaded and executed by a processor, to implement the enterprise stability evaluation method of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211676264.XA CN115936243A (en) | 2022-12-26 | 2022-12-26 | Enterprise stability evaluation method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211676264.XA CN115936243A (en) | 2022-12-26 | 2022-12-26 | Enterprise stability evaluation method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115936243A true CN115936243A (en) | 2023-04-07 |
Family
ID=86697624
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211676264.XA Pending CN115936243A (en) | 2022-12-26 | 2022-12-26 | Enterprise stability evaluation method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115936243A (en) |
-
2022
- 2022-12-26 CN CN202211676264.XA patent/CN115936243A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107103171B (en) | Modeling method and device of machine learning model | |
| US11650968B2 (en) | Systems and methods for predictive early stopping in neural network training | |
| CN110738564A (en) | Post-loan risk assessment method and device and storage medium | |
| CN110415036B (en) | User grade determining method, device, computer equipment and storage medium | |
| CN107766929A (en) | model analysis method and device | |
| CN110634060A (en) | User credit risk assessment method, system, device and storage medium | |
| CN114048436A (en) | Construction method and construction device for forecasting enterprise financial data model | |
| CN113807469A (en) | Multi-energy user value prediction method, device, storage medium and equipment | |
| CN110796485A (en) | Method and device for improving prediction precision of prediction model | |
| CN113409150A (en) | Operation risk and credit risk assessment method, device and computer storage medium | |
| CN112884569A (en) | Credit assessment model training method, device and equipment | |
| CN111598329A (en) | Time sequence data prediction method based on automatic parameter adjustment recurrent neural network | |
| CN113449919B (en) | Power consumption prediction method and system based on feature and trend perception | |
| CN113032367A (en) | Dynamic load scene-oriented cross-layer configuration parameter collaborative tuning method and system for big data system | |
| CN115936243A (en) | Enterprise stability evaluation method, device, equipment and storage medium | |
| CN114004530B (en) | Enterprise electric power credit modeling method and system based on ordering support vector machine | |
| CN115829123A (en) | Natural gas demand prediction method and device based on grey model and neural network | |
| CN115292381A (en) | Virtual currency mining behavior identification method based on extreme gradient lifting algorithm | |
| CN113762591A (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM counterstudy | |
| CN114549233A (en) | Floating population prediction method based on combination of LGB algorithm and ARIMA algorithm | |
| CN109308565B (en) | Crowd performance grade identification method and device, storage medium and computer equipment | |
| CN111026661A (en) | Method and system for comprehensively testing usability of software | |
| CN111179070A (en) | Loan risk timeliness prediction system and method based on LSTM | |
| CN110275773B (en) | Paas resource recycling index system based on real data model fitting | |
| CN117474585A (en) | Electricity charge prediction method, model training method and related equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20230705 Address after: 071000 201, Building C, Corporate Office Area, Xiong'an Civil Service Center, Rongcheng County, Xiong'an District, China (Hebei) Pilot Free Trade Zone, Hebei Province Applicant after: State Grid commercial big data Co.,Ltd. Applicant after: Guowang Xiongan Finance Technology Group Co.,Ltd. Address before: 071000 201, Building C, Corporate Office Area, Xiong'an Civil Service Center, Rongcheng County, Xiong'an District, China (Hebei) Pilot Free Trade Zone, Baoding City, Hebei Province Applicant before: State Grid commercial big data Co.,Ltd. |