Enterprise operation continuity prediction method and device, medium and equipment
By obtaining and processing the characteristics of corporate business behavior, forming feature vectors, and using multiple machine learning models to make predictions, the problem of insufficient accuracy of corporate business continuity prediction in the existing technology is solved, the accuracy and efficiency of prediction are improved, and the accuracy and efficiency of prediction are achieved. More efficient supervision and assistance.
Patent Information
- Application Number
- CN202411296739.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-18
- Publication Date
- 2024-12-31
AI Technical Summary
The existing corporate business continuity prediction methods rely on statistical analysis, and the accuracy needs to be improved, and the efficiency and accuracy are insufficient, so it is necessary to improve the accuracy and efficiency of the prediction results.
Use the enterprise to obtain the business behavior characteristics in the latest time period, perform pre-processing, derivative processing and dimensionality reduction processing, form a business behavior characteristic vector, and input a pre-trained enterprise business continuity prediction model, and use a variety of machine learning models (such as Classified regression trees, random forests, and gradient boosting trees) are used for prediction.
It improves the accuracy and efficiency of enterprise continuity predictions in the future time period, reduces the degree of manual participation, reduces labor costs, and achieves more accurate assistance and supervision.
Smart Images
Figure CN119228425A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of risk prediction, and in particular to a method, device, medium and equipment for predicting the continuity of business operations of an enterprise. Background Art
[0002] Business continuity means that the company will continue to operate in the future and will not face the risk of bankruptcy and liquidation in the short term. By predicting the possibility of deregistration and non-normalization of enterprises, and focusing on supervising enterprises that are no longer continuing, precise assistance can be provided and tax losses can be avoided. At present, the prediction of business continuity of enterprises mostly adopts statistical analysis methods, such as trend prediction analysis, causal prediction analysis and regression analysis, to reveal the regular relationship between relevant variables and draw conclusions on the calculation results. The accuracy of current statistical analysis methods needs to be improved. Summary of the invention
[0003] In response to at least one of the above technical problems, an embodiment of the present invention provides a method, device, medium, and equipment for predicting business continuity of an enterprise.
[0004] According to the first aspect, the enterprise business continuity prediction method provided by the embodiment of the present invention includes:
[0005] Obtain the original operating behavior characteristics of the target enterprise in the most recent predetermined time period;
[0006] Preprocessing the original business behavior characteristics;
[0007] Performing derivative processing on the pre-processed original business behavior characteristics to obtain derived business behavior characteristics;
[0008] Performing dimensionality reduction processing on the preprocessed original business behavior features and the derived business behavior features to obtain a business behavior feature vector including a first preset number of business behavior features;
[0009] The operating behavior feature vector corresponding to the target enterprise is input into a pre-trained enterprise operation continuity prediction model to obtain the continuity prediction result of the target enterprise in a preset time period in the future.
[0010] In one embodiment, the obtaining of the original operating behavior characteristics of the target enterprise within the most recent predetermined time period includes: extracting a second preset number of original operating behavior characteristics from the declared collection information, financial information, invoice information and registration change information of the target enterprise within the most recent predetermined time period.
[0011] In one embodiment, the preprocessing of the original business behavior features includes at least one of the following:
[0012] The qualitative features in the original business behavior features are converted into numerical features by using a qualitative feature dummy coding method;
[0013] The continuous features in the original business behavior features are converted into discrete features by using a continuous feature discretization processing method.
[0014] In one embodiment, the deriving processing of the preprocessed original business behavior characteristics to obtain derived business behavior characteristics includes: based on the preprocessed original business behavior characteristics, using at least one of feature combination and feature operation to generate a third preset number of the derived business behavior characteristics.
[0015] In one embodiment, the dimensionality reduction processing is performed on the pre-processed original business behavior features and the derived business behavior features to obtain a business behavior feature vector including a first preset number of business behavior features, including:
[0016] Performing data exploratory analysis on the pre-processed original business behavior characteristics and the derived business behavior characteristics to obtain data exploratory analysis results;
[0017] According to the data exploratory analysis result, selecting the first preset number of business behavior features from the preprocessed original business behavior features and the derived business behavior features;
[0018] The first preset number of business behavior features form the business behavior feature vector.
[0019] In one embodiment, the enterprise business continuity prediction model includes at least two binary classification prediction models and a logistic regression model connected to each binary classification prediction model, and the at least two binary classification prediction models include at least two of a classification regression tree model, a random forest model, and a gradient boosting tree model; wherein: each binary classification prediction model is used to perform a binary classification prediction based on the business behavior feature vector to obtain a respective continuity prediction result; and the logistic regression model is used to output a final continuity prediction result based on at least two continuity prediction results.
[0020] In one embodiment, the training process of the enterprise business continuity prediction model includes:
[0021] Extracting registration information of multiple enterprises from the tax system, and determining enterprises that are currently in an abnormal or cancelled state based on the registration information, and treating the enterprises that are currently in an abnormal or cancelled state as black enterprises; and determining enterprises that are currently in a normal state based on the registration information, and treating the enterprises that are currently in a normal state as white enterprises;
[0022] Extracting the original operating behavior characteristics of each black enterprise in a predetermined time period before the status becomes abnormal or cancelled, and extracting the original operating behavior characteristics of each white enterprise in a recent predetermined time period;
[0023] Pre-process the original operating behavior characteristics of each enterprise;
[0024] Perform derivative processing on the original operating behavior characteristics of each enterprise after preprocessing to obtain the derived operating behavior characteristics of the enterprise;
[0025] Performing dimensionality reduction processing on the original business behavior features and the derived business behavior features of each enterprise after preprocessing to obtain a first preset number of business behavior features of the enterprise, forming a business behavior feature vector with the first preset number of business behavior features, and labeling the business behavior feature vector according to the state of the enterprise to obtain a training sample corresponding to the enterprise;
[0026] Divide the obtained training samples into training set, validation set and test set;
[0027] The training set is used to train the model; after the training converges, the validation set is used to evaluate the model performance, and the model is tuned according to the evaluation results; after the tuning is completed, the test set is used to evaluate and output the final performance of the enterprise business continuity prediction model.
[0028] According to the second aspect, the enterprise business continuity prediction device provided by the embodiment of the present invention includes:
[0029] A feature acquisition module is used to acquire the original operating behavior features of the target enterprise in the most recent predetermined time period;
[0030] A feature processing module, used for pre-processing the original business behavior features;
[0031] A feature derivation module is used to derive the original business behavior features after preprocessing to obtain derived business behavior features;
[0032] A dimension reduction processing module, used for performing dimension reduction processing on the pre-processed original business behavior features and the derived business behavior features to obtain a business behavior feature vector including a first preset number of business behavior features;
[0033] The model prediction module is used to input the business behavior feature vector corresponding to the target enterprise into the pre-trained enterprise business continuity prediction model to obtain the continuity prediction result of the target enterprise in a preset time period in the future.
[0034] According to a third aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed in a computer, enables the computer to execute the method provided in the first aspect.
[0035] According to a fourth aspect, an embodiment of the present invention provides a computing device, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, the method provided in the first aspect is implemented.
[0036] The enterprise business continuity prediction method, device, medium, and equipment provided by the embodiment of the present invention first obtains the original business behavior characteristics of the target enterprise in the most recent predetermined time period, then pre-processes the original business behavior characteristics, and then derives the pre-processed original business behavior characteristics to obtain derived business behavior characteristics. The pre-processed original business behavior characteristics and the derived business behavior characteristics are subjected to dimensionality reduction processing to obtain a business behavior characteristic vector, and finally the business behavior characteristic vector corresponding to the target enterprise is input into the business continuity prediction model, so as to obtain the continuity prediction result of the target enterprise in the future preset time period. It can be seen that compared with the statistical analysis method in the prior art, the present invention is based on a large model to predict whether the enterprise will have continuity in the future period of time, which is more accurate and efficient. The entire process requires very little human participation, which can reduce labor costs. BRIEF DESCRIPTION OF THE DRAWINGS
[0037] Figure 1 A schematic diagram of a process flow of a method for predicting business continuity of an enterprise in one embodiment of the present invention;
[0038] Figure 2 A schematic diagram of a flow chart of a model training process in one embodiment of the present invention;
[0039] Figure 3 It is a structural block diagram of a device for predicting business continuity of an enterprise in one embodiment of the present invention. DETAILED DESCRIPTION
[0040] In a first aspect, an embodiment of the present invention provides a method for predicting business continuity of an enterprise, see Figure 1 The method comprises the following steps S110 to S150:
[0041] S110, obtaining the original operating behavior characteristics of the target enterprise in the most recent predetermined time period;
[0042] Among them, the target enterprise refers to the enterprise that needs to conduct enterprise continuity forecast.
[0043] In one embodiment, obtaining the original operating behavior characteristics of the target enterprise in the most recent predetermined time period in S120 may include:
[0044] A second preset number of original business behavior features are extracted from the target enterprise's declared collection information, financial information, invoice information and registration change information within a recent predetermined time period.
[0045] For example, 90 original business behavior features are extracted from the target company's declared collection information, financial information, invoice information and registration change information in the last three months.
[0046] S120, pre-processing the original business behavior characteristics;
[0047] In one embodiment, the preprocessing of the original business behavior feature in S120 includes at least one of the following:
[0048] 1. Using the qualitative feature dummy coding method, the qualitative features in the original business behavior features are converted into numerical features;
[0049] 2. The continuous features in the original business behavior features are converted into discrete features by using a continuous feature discretization processing method.
[0050] Among them, the dummy coding method of qualitative features can convert qualitative features into numerical features for subsequent use. Converting continuous features into discrete features is also conducive to the use of subsequent features.
[0051] In S120, it is necessary to perform preprocessing according to the characteristics of the original business behavior characteristics to achieve the transition from data characterization to generalization, which is convenient for data analysis and model training. Of course, in addition to the dummy coding of qualitative features and the discretization of continuous features, there can be other preprocessing methods, which are not limited here.
[0052] S130, performing derivative processing on the pre-processed original business behavior characteristics to obtain derived business behavior characteristics;
[0053] In one embodiment, the deriving processing of the preprocessed original business behavior characteristics in S130 to obtain derived business behavior characteristics may include: generating a third preset number of the derived business behavior characteristics by at least one of feature combination and feature operation based on the preprocessed original business behavior characteristics.
[0054] For example, based on the 90 original business behavior features extracted previously, 24 new features, namely, 24 derived business behavior features, are derived by combination, calculation or other methods after preprocessing.
[0055] It is understandable that the original business behavior features are derived and processed to obtain the derived business behavior features. The original business behavior features and the derived business behavior features are used together in the subsequent steps to improve the accuracy of model prediction. In the model training scenario, adding a certain number of derived business behavior features can improve the generalization ability and robustness of the model.
[0056] S140, performing dimensionality reduction processing on the preprocessed original business behavior features and the derived business behavior features to obtain a business behavior feature vector including a first preset number of business behavior features;
[0057] For example, there are 114 original business behavior features and derived business behavior features, which is a relatively large number. Therefore, some dimensionality reduction processing can be performed to obtain 106 business behavior features. These 106 business behavior features form a business behavior feature vector.
[0058] In the process of dimensionality reduction, in order to prevent dimensionality reduction from affecting the prediction of subsequent models, it is necessary to retain more important features during dimensionality reduction.
[0059] In one embodiment, the dimensionality reduction processing of the pre-processed original business behavior features and the derived business behavior features in S140 to obtain a business behavior feature vector including a first preset number of business behavior features may include:
[0060] 1. Perform data exploratory analysis on the pre-processed original business behavior characteristics and the derived business behavior characteristics to obtain data exploratory analysis results;
[0061] 2. According to the data exploratory analysis result, selecting the first preset number of business behavior features from the pre-processed original business behavior features and the derived business behavior features;
[0062] 3. Form the business behavior feature vector from the first preset number of business behavior features.
[0063] Among them, data feature analysis, for example, can explore the correlation between business behavior characteristics and labels, and the similarity between business behavior characteristics. When reducing the dimension, try to retain the business behavior characteristics with a strong correlation with the labels, and remove redundant business behavior characteristics with a high similarity.
[0064] S150, inputting the operating behavior feature vector corresponding to the target enterprise into a pre-trained enterprise operation continuity prediction model to obtain a continuity prediction result of the target enterprise in a preset time period in the future.
[0065] Among them, the target enterprise's business behavior feature vector is input into the enterprise business continuity prediction model, and the model can output the continuity prediction results of the target enterprise in the future preset time period.
[0066] Among them, the enterprise business continuity prediction model includes at least two binary classification prediction models and a logistic regression model connected to each binary classification prediction model, and the at least two binary classification prediction models include at least two of a classification regression tree model, a random forest model and a gradient boosting tree model; wherein: each binary classification prediction model is used to perform a binary classification prediction based on the business behavior feature vector to obtain a respective continuity prediction result; the logistic regression model is used to output a final continuity prediction result based on at least two continuity prediction results.
[0067] Among them, the classification and regression tree model is CART tree, the full name is classification and regression tree. CART tree can be used for both classification and regression.
[0068] Among them, the gradient boosting tree model is XGBoost, whose full name is eXtreme Gradient Boosting. It is a distributed tree model based on gradient boosting. It prevents overfitting and improves efficiency by adding regularization terms and second-order derivatives.
[0069] Among them, the random forest model refers to a classifier that uses multiple trees to train and predict samples.
[0070] For example, the enterprise operation continuity prediction model includes three binary classification prediction models, namely, the classification regression tree model, the random forest model and the gradient boosting tree model. The classification regression tree model, the random forest model and the gradient boosting tree model can all perform binary classification predictions, and each model will output a prediction result: continuity or non-continuity. Then the three prediction results will be input into the logistic regression model, and the logistic regression model will output the final continuity prediction result according to the three prediction results.
[0071] In one embodiment, see Figure 2 The training process of the enterprise business continuity prediction model may include 7 steps:
[0072] 1. Extracting registration information of multiple enterprises from the tax system, and determining enterprises that are currently in an abnormal or cancelled state as black enterprises and enterprises that are currently in a normal state as white enterprises based on the registration information;
[0073] 2. Extract the original operating behavior characteristics of each black enterprise in a predetermined time period before its status becomes abnormal or cancelled, and extract the original operating behavior characteristics of each white enterprise in a recent predetermined time period;
[0074] Specifically, the business characteristics of each enterprise are extracted from its declared collection information, financial information, invoice information, and registration change information as the original business behavior characteristics. Then, the availability of the original business behavior characteristics is evaluated from the three dimensions of acquisition difficulty, coverage, and accuracy, and the original business behavior characteristics with higher availability are retained. Among them, the feature acquisition time of black enterprises is within the predetermined time period before the status becomes abnormal or is cancelled. The feature acquisition time of white enterprises is within the most recent predetermined time period. Finally, 90 original business behavior characteristics are obtained for each enterprise.
[0075] 3. Pre-process the original operating behavior characteristics of each enterprise;
[0076] For example, the original operating behavior characteristics of each enterprise can be preprocessed by using qualitative feature dummy coding, continuous feature discretization or other special feature processing methods.
[0077] 4. Perform derivative processing on the original operating behavior characteristics of each enterprise after preprocessing to obtain the derived operating behavior characteristics of the enterprise;
[0078] For example, by combining, calculating or otherwise deriving the original business behavior characteristics of each enterprise, 24 derived business behavior characteristics are obtained, and 114 business behavior characteristics can be obtained for one enterprise. Through derivation processing, it is convenient to fully explore the characteristics related to the sustainable operation of the enterprise, which are used for the subsequent feature selection and model training, and improve the accuracy of model prediction.
[0079] 5. Perform dimensionality reduction processing on the original business behavior features and the derived business behavior features of each enterprise after preprocessing to obtain a first preset number of business behavior features of the enterprise, form a business behavior feature vector with the first preset number of business behavior features, and label the business behavior feature vector according to the state of the enterprise to obtain a training sample corresponding to the enterprise;
[0080] For example, for an enterprise, there are 114 original business behavior characteristics and the derived business behavior characteristics. Data exploratory analysis is performed on these 114 business behavior characteristics to understand the relationship between the business behavior characteristics and the relationship between the business behavior characteristics and the labels, so as to better select features and build models, including but not limited to data overview, missing value and abnormal value viewing, label value distribution, category characteristics and numerical characteristics analysis. Combined with the results of data exploratory analysis, the 114 business behavior characteristics are reduced in dimension to reduce the number of business behavior characteristics and find truly relevant business behavior characteristics, which can improve the accuracy of the prediction model and reduce the running time. Finally, the business behavior characteristics of an enterprise are reduced to 106.
[0081] It is understandable that the label of the operating behavior feature vector of the black enterprise is no continuity, and the label of the operating behavior feature vector of the white enterprise is continuity.
[0082] 6. Divide the obtained training samples into training set, validation set and test set;
[0083] For example, each training sample is divided into a training set, a validation set, and a test set in a ratio of 6:2:2.
[0084] 7. Use the training set to train the model; after the training converges, use the validation set to evaluate the model performance, and perform model tuning according to the evaluation results; after the tuning is completed, use the test set to evaluate and output the final performance of the enterprise business continuity prediction model.
[0085] For example, a training set is used to train a prediction model, which includes CART tree, random forest, XGBoost, and logistic regression models. After training, the three models of CART tree, random forest, and XGBoost can be used to output three continuity prediction results. These three continuity prediction results are input into a logistic regression model, which will make a comprehensive judgment on the three continuity prediction results and output the final continuity prediction result. The validation set evaluates the effect of the prediction model so that the prediction model can be tuned. The test set evaluates the final performance of the prediction model.
[0086] It is understandable that the prediction of business continuity is a binary classification problem. The performance of the prediction model can be evaluated through common evaluation indicators such as confusion matrix, accuracy, precision, recall, AUC, etc. Through repeated debugging, training, and testing, the prediction model finally reaches a level that can be used in practice.
[0087] It can be seen that in the training scenario, the original business behavior characteristics of black enterprises before they were cancelled or converted to non-business and the original business behavior characteristics of white enterprises in the recent period were extracted, and the original business behavior characteristics were preprocessed and derived to obtain the derived business behavior characteristics. After feature dimensionality reduction, machine learning technology was used to train the model to obtain a prediction model.
[0088] After executing step 7 above, the prediction model can be deployed and applied. After the model is deployed and applied, the target enterprise's operating behavior characteristics are processed on a monthly basis and the prediction model is used to predict whether the target enterprise has continuity. The enterprises whose taxpayer status has been changed to abnormal or deregistered will be added to the black sample library, and the number of black samples will continue to increase.
[0089] It can be seen that the embodiments of the present invention can realize the intelligent prediction of the possibility of an enterprise being deregistered or becoming abnormal in the near future, that is, to realize the prediction of the continuity of enterprise operations, and assist the grassroots to carry out precise and intelligent supervision.
[0090] In a second aspect, an embodiment of the present invention provides a device for predicting business continuity of an enterprise, see Figure 3 , the device 100 comprises:
[0091] The feature acquisition module 110 is used to acquire the original business behavior features of the target enterprise in the most recent predetermined time period;
[0092] A feature processing module 120, used for pre-processing the original business behavior features;
[0093] The feature derivation module 130 is used to derive the pre-processed original business behavior features to obtain derived business behavior features;
[0094] A dimension reduction processing module 140 is used to perform dimension reduction processing on the pre-processed original business behavior features and the derived business behavior features to obtain a business behavior feature vector including a first preset number of business behavior features;
[0095] The model prediction module 150 is used to input the business behavior feature vector corresponding to the target enterprise into the pre-trained enterprise business continuity prediction model to obtain the continuity prediction result of the target enterprise in a preset time period in the future.
[0096] In one embodiment, the feature acquisition module is specifically used to extract a second preset number of original business behavior features from the target enterprise's declaration and collection information, financial information, invoice information and registration change information within a recent predetermined time period.
[0097] In one embodiment, the feature processing module is specifically used to perform at least one of the following: using qualitative feature dummy coding to convert the qualitative features in the original business behavior features into numerical features; using continuous feature discretization processing to convert the continuous features in the original business behavior features into discrete features.
[0098] In one embodiment, the feature derivation module is specifically used to generate a third preset number of the derived business behavior features by adopting at least one of feature combination and feature operation according to the preprocessed original business behavior features.
[0099] In one embodiment, the dimensionality reduction processing module is specifically used to: perform data exploratory analysis on the preprocessed original business behavior characteristics and the derived business behavior characteristics to obtain data exploratory analysis results; select the first preset number of business behavior characteristics from the preprocessed original business behavior characteristics and the derived business behavior characteristics based on the data exploratory analysis results; and form the business behavior feature vector with the first preset number of business behavior characteristics.
[0100] In one embodiment, the enterprise business continuity prediction model includes at least two binary classification prediction models and a logistic regression model connected to each binary classification prediction model, and the at least two binary classification prediction models include at least two of a classification regression tree model, a random forest model, and a gradient boosting tree model; wherein: each binary classification prediction model is used to perform a binary classification prediction based on the business behavior feature vector to obtain a respective continuity prediction result; and the logistic regression model is used to output a final continuity prediction result based on at least two continuity prediction results.
[0101] In one embodiment, the device may further include:
[0102] The model training module is used to extract the registration information of multiple enterprises from the tax system, and determine the enterprises that are currently in an abnormal or cancelled state based on the registration information, and regard the enterprises that are currently in an abnormal or cancelled state as black enterprises; and determine the enterprises that are currently in a normal state based on the registration information, and regard the enterprises that are currently in a normal state as white enterprises; extract the original business behavior characteristics of each black enterprise within a predetermined time period before the state becomes abnormal or cancelled, and extract the original business behavior characteristics of each white enterprise within a recent predetermined time period; pre-process the original business behavior characteristics of each enterprise; derive the original business behavior characteristics of each enterprise after pre-processing to obtain the derived business characteristics of the enterprise business behavior characteristics; perform dimensionality reduction processing on the original business behavior characteristics of each enterprise after preprocessing and the derived business behavior characteristics to obtain a first preset number of business behavior characteristics of the enterprise, form a business behavior feature vector with the first preset number of business behavior characteristics, and label the business behavior feature vector according to the status of the enterprise to obtain a training sample corresponding to the enterprise; divide the obtained training samples into a training set, a validation set and a test set; use the training set to train the model; after the training converges, use the validation set to evaluate the model performance, and perform model tuning processing according to the evaluation results; after the tuning is completed, use the test set to evaluate and output the final performance of the enterprise business continuity prediction model.
[0103] It is understandable that the explanation, specific implementation, beneficial effects, examples, etc. of the relevant contents in the device provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.
[0104] In a third aspect, an embodiment of the present invention provides a computer-readable medium, on which computer instructions are stored. When the computer instructions are executed by a processor, the processor executes the method provided in the first aspect.
[0105] Specifically, a system or device equipped with a storage medium can be provided, on which software program code that implements the functions of any of the above-mentioned embodiments is stored, and a computer (or CPU or MPU) of the system or device can be enabled to read and execute the program code stored in the storage medium.
[0106] In this case, the program code itself read from the storage medium can realize the function of any one of the above-mentioned embodiments, and thus the program code and the storage medium storing the program code constitute a part of the present invention.
[0107] The storage medium embodiments for providing the program code include a floppy disk, a hard disk, a magneto-optical disk, an optical disk (such as CD-ROM, CD-R, CD-RW, DVD-ROM, DVD-RAM, DVD-RW, DVD+RW), a magnetic tape, a non-volatile memory card, and a ROM. Alternatively, the program code can be downloaded from a server computer by a communication network.
[0108] In addition, it should be clear that the functions of any of the above embodiments can be implemented not only by executing the program code read by the computer, but also by enabling an operating system operating on the computer to complete part or all of the actual operations based on instructions from the program code.
[0109] In addition, it can be understood that the program code read from the storage medium is written to a memory provided in an expansion board inserted into the computer or to a memory provided in an expansion module connected to the computer, and then based on the instructions of the program code, a CPU installed on the expansion board or expansion module is enabled to perform part or all of the actual operations, thereby realizing the functions of any of the above-mentioned embodiments.
[0110] It is understandable that the explanation, specific implementation methods, beneficial effects, examples, etc. of the relevant contents in the computer-readable medium provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.
[0111] In a fourth aspect, an embodiment of the present specification provides a computing device, including a memory and a processor, wherein the memory stores executable code, and when the processor executes the executable code, it implements the method in any embodiment of the specification.
[0112] It is understandable that the explanation, specific implementation, beneficial effects, examples, etc. of the relevant contents in the computing device provided in the embodiment of the present invention can be found in the corresponding parts of the method provided in the first aspect, and will not be repeated here.
[0113] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.
[0114] Those skilled in the art should be aware that in one or more of the above examples, the functions described in the present invention can be implemented by hardware, software, widgets, or any combination thereof. When implemented by software, these functions can be stored in a computer-readable medium or transmitted as one or more instructions or codes on a computer-readable medium.
[0115] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made on the basis of the technical solution of the present invention should be included in the scope of protection of the present invention.
Claims
1. A method for predicting business continuity of an enterprise, characterized in that: include: Obtain the original operating behavior characteristics of the target enterprise in the most recent predetermined time period; Preprocessing the original business behavior characteristics; Performing derivative processing on the pre-processed original business behavior characteristics to obtain derived business behavior characteristics; Performing dimensionality reduction processing on the preprocessed original business behavior features and the derived business behavior features to obtain a business behavior feature vector including a first preset number of business behavior features; The operating behavior feature vector corresponding to the target enterprise is input into a pre-trained enterprise operation continuity prediction model to obtain the continuity prediction result of the target enterprise in a preset time period in the future.
2. The method according to claim 1, characterized in that The acquisition of the original operating behavior characteristics of the target enterprise in the most recent predetermined time period includes: A second preset number of original business behavior features are extracted from the target enterprise's declared collection information, financial information, invoice information and registration change information within a recent predetermined time period.
3. The method according to claim 1, characterized in that The preprocessing of the original business behavior characteristics includes at least one of the following: The qualitative features in the original business behavior features are converted into numerical features by using a qualitative feature dummy coding method; The continuous features in the original business behavior features are converted into discrete features by using a continuous feature discretization processing method.
4. The method according to claim 1, characterized in that: The derivation processing of the pre-processed original business behavior characteristics to obtain the derived business behavior characteristics includes: According to the preprocessed original business behavior features, a third preset number of the derived business behavior features are generated by adopting at least one of feature combination and feature operation.
5. The method according to claim 1, characterized in that The dimensionality reduction process is performed on the pre-processed original business behavior features and the derived business behavior features to obtain a business behavior feature vector including a first preset number of business behavior features, including: Performing data exploratory analysis on the pre-processed original business behavior characteristics and the derived business behavior characteristics to obtain data exploratory analysis results; According to the data exploratory analysis result, selecting the first preset number of business behavior features from the preprocessed original business behavior features and the derived business behavior features; The first preset number of business behavior features form the business behavior feature vector.
6. The method according to claim 1, characterized in that The enterprise business continuity prediction model includes at least two binary classification prediction models and a logistic regression model connected to each binary classification prediction model, wherein the at least two binary classification prediction models include at least two of a classification regression tree model, a random forest model, and a gradient boosting tree model; wherein: Each binary classification prediction model is used to perform binary classification prediction according to the business behavior feature vector to obtain respective continuity prediction results; The logistic regression model is used to output a final continuity prediction result based on at least two continuity prediction results.
7. The method according to claim 1, characterized in that The training process of the enterprise business continuity prediction model includes: Extracting registration information of multiple enterprises from the tax system, and determining enterprises that are currently in an abnormal or cancelled state based on the registration information, and treating the enterprises that are currently in an abnormal or cancelled state as black enterprises; and determining enterprises that are currently in a normal state based on the registration information, and treating the enterprises that are currently in a normal state as white enterprises; Extracting the original operating behavior characteristics of each black enterprise in a predetermined time period before the status becomes abnormal or cancelled, and extracting the original operating behavior characteristics of each white enterprise in a recent predetermined time period; Pre-process the original operating behavior characteristics of each enterprise; Perform derivative processing on the original operating behavior characteristics of each enterprise after preprocessing to obtain the derived operating behavior characteristics of the enterprise; Performing dimensionality reduction processing on the original business behavior features and the derived business behavior features of each enterprise after preprocessing to obtain a first preset number of business behavior features of the enterprise, forming a business behavior feature vector with the first preset number of business behavior features, and labeling the business behavior feature vector according to the state of the enterprise to obtain a training sample corresponding to the enterprise; Divide the obtained training samples into training set, validation set and test set; The training set is used to train the model; after the training converges, the validation set is used to evaluate the model performance, and the model is tuned according to the evaluation results; after the tuning is completed, the test set is used to evaluate and output the final performance of the enterprise business continuity prediction model.
8. A device for predicting business continuity of an enterprise, characterized in that: include: A feature acquisition module is used to acquire the original operating behavior features of the target enterprise in the most recent predetermined time period; A feature processing module, used for pre-processing the original business behavior features; A feature derivation module is used to derive the original business behavior features after preprocessing to obtain derived business behavior features; A dimension reduction processing module, used for performing dimension reduction processing on the pre-processed original business behavior features and the derived business behavior features to obtain a business behavior feature vector including a first preset number of business behavior features; The model prediction module is used to input the business behavior feature vector corresponding to the target enterprise into the pre-trained enterprise business continuity prediction model to obtain the continuity prediction result of the target enterprise in a preset time period in the future.
9. A computer-readable storage medium, characterized in that: A computer program is stored thereon, and when the computer program is executed in a computer, the computer is caused to execute the method according to any one of claims 1 to 7.
10. A computing device, characterized in that The method comprises a memory and a processor, wherein the memory stores executable codes, and when the processor executes the executable codes, the method according to any one of claims 1 to 7 is implemented.