A model training method, business processing method, device and equipment

By calculating multi-view integration loss and perspective constraint coefficients for business sample data, the business evaluation model is trained, and the business evaluation accuracy problem caused by perspective independence in the existing technology is solved, and the effectiveness of business processing is improved.

CN113515867BActive Publication Date: 2024-11-15INDUSTRIAL AND COMMERCIAL BANK OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202110854994.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-07-28
Publication Date
2024-11-15
Estimated Expiration
2041-07-28

AI Technical Summary

Technical Problem

The prior art causes insufficient information due to the independence between different perspectives during business evaluation, which reduces the classification effect of the model and the accuracy of evaluation results, and affects the effectiveness of business processing.

Method used

By obtaining business sample data, dividing it into single-view data from multiple business processing perspectives, the view angle constraint coefficient is calculated to represent the correlation of different perspectives, and the multi-view integration loss is determined based on the sample label, and the business evaluation model is comprehensively trained to improve the evaluation accuracy .

Benefits of technology

The training effect of the model is improved, the accuracy of business evaluation and the effectiveness of business processing are improved, and the effectiveness of business execution is ensured.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN113515867B_ABST
    Figure CN113515867B_ABST
Patent Text Reader

Abstract

The embodiments of this specification provide a model training method, a business processing method, an apparatus and a device, which can be applied to the field of artificial intelligence technology. The method includes: obtaining business sample data; the business sample data corresponds to a sample label; dividing the business sample data into single-perspective data corresponding to at least two business processing perspectives; the business processing perspectives correspond to pre-set sub-classification models; calculating the perspective constraint coefficient by integrating the business sample data and sub-classification models of each business processing perspective; determining the multi-perspective integration loss based on the business sample data and sample labels; training a business evaluation model by integrating the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business. The above method comprehensively considers the correlation between different perspectives corresponding to the business data, improves the training effect of the model, improves the accuracy of the business evaluation, and ensures the effective execution of the business.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The embodiments of this specification relate to the field of artificial intelligence technology, and in particular to a model training method, business processing method, device and equipment. Background Art

[0002] Business evaluation is generally a necessary step taken before handling the business of other customers or companies. For example, business evaluation can conduct a comprehensive evaluation of the customer or company's execution ability, performance ability, and trustworthiness, and then determine whether to handle the business or select the corresponding business processing method based on the evaluation results, to ensure the effectiveness of business processing and reduce the risks brought by the business.

[0003] At present, when evaluating a business, the business is generally classified from multiple perspectives, and then the corresponding business evaluation results are determined based on the characteristics of the data in these perspectives. However, although the above evaluation method has information from multiple perspectives, since different perspectives are independent of each other, the information contained in a single perspective is not enough to provide sample category discrimination information, which will reduce the classification effect of the model. In addition, the method of using multiple perspectives to evaluate the business separately ignores the correlation between different perspectives, which also affects the accuracy of the evaluation results, and is not conducive to the effective processing of the business. Therefore, there is an urgent need for a method that can accurately and effectively evaluate the business in order to better implement business processing. Summary of the invention

[0004] The purpose of the embodiments of this specification is to provide a model training method, a business allocation method, an apparatus and a device to solve the problem of how to effectively evaluate the business to improve the effect of business execution.

[0005] To solve the above technical problems, an embodiment of the present specification provides a model training method based on business evaluation, including: obtaining business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; the business sample data is divided into single-perspective data corresponding to at least two business processing perspectives; the business processing perspective is used to indicate different categories of data; the business processing perspectives correspond to pre-set sub-classification models; the perspective constraint coefficient is calculated by combining the business sample data and sub-classification models of each business processing perspective; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; based on the business sample data and sample labels, a multi-perspective integration loss is determined; the multi-perspective integration loss is used to represent the deviation value predicted by each business processing perspective; the business evaluation model is trained by combining the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business.

[0006] The embodiment of this specification also proposes a model training device based on business evaluation, including: a business sample data acquisition module, used to acquire business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; a single-perspective data division module, used to divide the business sample data into single-perspective data corresponding to at least two business processing perspectives respectively; the business processing perspective is used to indicate different categories of data; the business processing perspectives respectively correspond to pre-set sub-classification models; a perspective constraint coefficient calculation module, used to comprehensively calculate the perspective constraint coefficient of the business sample data and the sub-classification model of each business processing perspective; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; a multi-perspective integrated loss determination module, used to determine the multi-perspective integrated loss based on the business sample data and sample labels; the multi-perspective integrated loss is used to represent the deviation value predicted by each business processing perspective; a business evaluation model training module, used to comprehensively train the business evaluation model with the perspective constraint coefficient and the multi-perspective integrated loss; the business evaluation model is used to evaluate the execution effect of the business.

[0007] The embodiment of this specification also proposes a model training device based on business evaluation, including a memory and a processor; the memory is used to store computer program instructions; the processor is used to execute the computer program instructions to implement the following steps: obtaining business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; dividing the business sample data into single-perspective data corresponding to at least two business processing perspectives respectively; the business processing perspective is used to indicate different categories of data; the business processing perspectives correspond to pre-set sub-classification models respectively; the business sample data and sub-classification models of each business processing perspective are combined to calculate the perspective constraint coefficient; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; based on the business sample data and sample labels, the multi-perspective integration loss is determined; the multi-perspective integration loss is used to represent the deviation value predicted by each business processing perspective; the business evaluation model is trained by combining the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business.

[0008] In order to solve the above technical problems, the embodiment of this specification also proposes a business processing method, including: obtaining target business data of a target business; inputting the target business data into a business evaluation model to obtain a business evaluation result; the business evaluation model is obtained in the following manner: obtaining business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; the business sample data is divided into single-perspective data corresponding to at least two business processing perspectives respectively; the business processing perspective is used to indicate different categories of data; the business processing perspectives respectively correspond to pre-set sub-classification models; the business processing perspectives are combined to calculate the perspective constraint coefficient of each business processing perspective and the sub-classification model; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; based on the business sample data and sample labels, a multi-perspective integration loss is determined; the multi-perspective integration loss is used to represent the deviation value predicted by each business processing perspective; the business evaluation model is trained by combining the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business; based on the business evaluation result, the business processing method corresponding to the target business is determined.

[0009] The embodiment of the present specification also proposes a business processing device, including: a target business data acquisition module, used to acquire target business data of a target business; a business evaluation result acquisition module, used to input the target business data into a business evaluation model to obtain a business evaluation result; the business evaluation model is obtained in the following manner: acquiring business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; dividing the business sample data into single-view data corresponding to at least two business processing perspectives respectively; the business processing perspective is used to indicate different categories of data; the business processing perspectives respectively correspond to pre-set sub-classification models; calculating a perspective constraint coefficient by integrating the business sample data and sub-classification models of each business processing perspective; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; determining a multi-perspective integration loss based on the business sample data and the sample label; the multi-perspective integration loss is used to represent the deviation value predicted by each business processing perspective; training a business evaluation model by integrating the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business; a business processing method determination module is used to determine the business processing method corresponding to the target business based on the business evaluation result.

[0010] The embodiment of the present specification also proposes a business processing device, including a memory and a processor; the memory is used to store computer program instructions; the processor is used to execute the computer program instructions to implement the following steps: obtaining target business data of a target business; inputting the target business data into a business evaluation model to obtain a business evaluation result; the business evaluation model is obtained in the following manner: obtaining business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; dividing the business sample data into single-view data corresponding to at least two business processing perspectives respectively; the business processing perspective is used to indicate different categories of data; the business processing perspectives respectively correspond to pre-set sub-classification models; comprehensively calculating the perspective constraint coefficient by combining the business sample data and the sub-classification models of each business processing perspective; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; determining a multi-perspective integration loss based on the business sample data and the sample label; the multi-perspective integration loss is used to represent the deviation value predicted by each business processing perspective; comprehensively training the business evaluation model by combining the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business; and determining the business processing method corresponding to the target business based on the business evaluation result.

[0011] It can be seen from the technical solutions provided by the above embodiments of this specification that after obtaining the business sample data, the embodiments of this specification divide the business sample data into single-perspective data corresponding to multiple business processing perspectives, and then combine the business sample data and sub-classification models corresponding to different business processing perspectives to calculate the perspective constraint coefficient to determine the correlation between the data from different business processing perspectives, and then determine the multi-perspective integration loss based on the business sample data and sample labels, determine the predicted deviation value, and train the business evaluation model by combining the above-mentioned perspective constraint coefficient and multi-perspective integration loss to evaluate the business. After the business data is input into the trained model, the business evaluation results based on the output of the model can be used to process the business. Through the above method, the correlation between different perspectives corresponding to the business data is comprehensively considered, the training effect of the model is improved, the accuracy of the business evaluation is improved, and the effective execution of the business is guaranteed. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0013] Figure 1This is a flow chart of a model training method based on business evaluation according to an embodiment of this specification;

[0014] Figure 2 A schematic diagram of a multi-view model training process according to an embodiment of this specification;

[0015] Figure 3 This is a flow chart of a business evaluation process according to an embodiment of the present specification;

[0016] Figure 4 This is a flow chart of a model training method according to an embodiment of this specification;

[0017] Figure 5 This is a module diagram of a model training device based on business evaluation according to an embodiment of this specification;

[0018] Figure 6 This is a module diagram of a model training device according to an embodiment of this specification;

[0019] Figure 7 This is a structural diagram of a model training device based on business evaluation according to an embodiment of this specification;

[0020] Figure 8 This is a structural diagram of a model training device according to an embodiment of this specification. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of this specification to clearly and completely describe the technical solutions in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0022] In order to solve the above technical problems, a model training method based on business evaluation is first introduced in the embodiment of this specification. The execution subject of the model training method based on business evaluation is a model training device based on business evaluation, and the model training device based on business evaluation includes but is not limited to a server, an industrial computer, a PC, etc. Figure 1 As shown, the business evaluation-based model training method may include the following specific implementation steps.

[0023] S110: Acquire business sample data; the business sample data corresponds to a sample tag; the sample tag is used to describe the business execution result.

[0024] The business sample data may be sample data used to train the model, and may include processing records corresponding to the business, business information itself, and corresponding information of users corresponding to the business.

[0025] The business sample data may correspond to a sample tag. The sample tag is used to describe the execution result of the business. Based on the execution result corresponding to the sample tag, information such as the execution difficulty of the business can be reflected, thereby realizing the evaluation of the business. For example, a successfully executed business may have a better evaluation result.

[0026] In some implementations, the business sample data may include labeled data and unlabeled data. In the field of machine learning, modeling and learning based on partially labeled data and partially unlabeled data is fast and accurate, and has good application value.

[0027] The labeled data is data labeled with corresponding sample labels, and correspondingly, the unlabeled data does not have sample labels. Based on the labeled data and the unlabeled data, a semi-supervised learning method can be used for model training, thereby reducing the time spent on labeling sample labels while ensuring the accuracy of training.

[0028] Preferably, in order to reduce the time consumed in labeling data, the labeled data may correspond to only a small proportion of the overall business sample data.

[0029] In some embodiments, after the business sample data is obtained, the business sample data may be preprocessed. The preprocessing includes: completing the business sample data based on preset feature fields. In order to ensure that the business sample data can be effectively utilized in subsequent processes, missing value columns in the business sample data may be completed. The preset feature fields may be fields corresponding to different completion rules. For example, for missing values ​​of numerical features in the original features, the preset feature fields may be completed with a "0" value; for missing values ​​of non-numerical features, the preset feature fields may be completed with an "unknown" value. In actual applications, other preset feature fields may be used to complete the original features as needed, which is not limited to the above examples and will not be repeated here.

[0030] S120: Divide the business sample data into single-perspective data corresponding to at least two business processing perspectives respectively; the business processing perspectives are used to indicate different categories of data; the business processing perspectives respectively correspond to pre-set sub-classification models.

[0031] The business processing perspective may be used to represent different information categories, so that the business sample data may be distinguished by using the business processing perspective, that is, used to indicate different categories of data.

[0032] In some implementations, the business processing perspective may include at least one of a basic information perspective, an annual report information perspective, a tax information perspective, a change information perspective, a fund transaction information perspective, and a news and public opinion perspective.

[0033] Specifically, when the business sample data is data for an enterprise, different perspectives of the sample can be constructed based on the categories to which these features belong. Specifically, the basic information of the enterprise constitutes the basic information perspective, the annual report information of the enterprise constitutes the annual report information perspective, the tax information of the enterprise constitutes the tax information perspective, the change information of the enterprise constitutes the change information perspective, the capital flow information of the enterprise constitutes the capital flow information perspective, and the news and public opinion information constitutes the news and public opinion perspective.

[0034] Based on the above-mentioned various business processing perspectives and the corresponding characteristics of the data under different business processing perspectives, these business sample data can be divided to obtain the corresponding data under different business processing perspectives.

[0035] The single-view data is the data corresponding to a certain business processing perspective after the division is completed. It should be noted that different single-view data can be completely different, or there can be some identical data, and there is no limitation on this.

[0036] In some embodiments, when dividing the single-view data, the statistical information and deviation features corresponding to the business sample data may be determined first. The statistical information is used to describe the overall characteristics of the business sample data, and may specifically include at least one of the maximum value, minimum value, mean value, and variance. The deviation feature may be used to describe the comparison result between the derived features and the original features after the business sample data is evolved, and may specifically include at least one of the minimum difference, maximum difference, and mean difference between the derived features and the original features.

[0037] After obtaining the statistical information and deviation features, since the above information can describe the nature of the business sample features, the business sample data can be divided into single-view data in combination with the statistical information and deviation features. The specific division rules can be set based on the needs of the actual application and will not be repeated here.

[0038] In some implementations, in order to ensure the reasonable use of the business sample data, the data distribution density of the business sample data may also be obtained. The data distribution density may be determined according to the spatial distribution of the data, so that the distribution of the main body of the business sample data can be obtained.

[0039] After obtaining the data distribution density, the weight of each business sample data can be determined according to the data distribution density, and based on the weight, steps such as calculating the perspective constraint coefficient, determining the multi-perspective integration loss, and training the business evaluation model are performed.

[0040] By constructing the data distribution density and assigning larger weights to samples with high local density and smaller weights to samples with low local density, we can prevent abnormal samples from affecting the learning process of the model and ensure the accuracy of model training.

[0041] S130: Calculate a perspective constraint coefficient by integrating the business sample data of each business processing perspective and the sub-classification model; the perspective constraint coefficient is used to represent the correlation of data of different business processing perspectives.

[0042] Each business processing perspective can be pre-set with a sub-classification model. The sub-classification model can determine the corresponding business evaluation result based on the business sample data under the current perspective. Since different business processing perspectives correspond to corresponding types of data, the accuracy of the sub-classification model for business evaluation is guaranteed.

[0043] In some embodiments, the sub-classification model can classify the business in combination with the execution effect of the business. The output result of the sub-classification model can be combined with the sample label corresponding to the business sample data. Specifically, the sub-classification model can determine a first business category and a second business category, and the first business category and the second business category can be used to represent the business that can be effectively executed and the business that cannot be effectively executed, respectively, and then the business evaluation result can be determined based on the output result of the sub-classification model.

[0044] Specifically, when calculating the viewing angle constraint coefficient, the above-mentioned service sample data and sub-classification model may be substituted into the formula Calculate the viewing angle constraint coefficient, where R mcor is the perspective constraint coefficient, V is the number of business processing perspectives, p and q are used to refer to different business processing perspectives, and f p 、f q are the sub-classification models corresponding to the business processing perspectives p and q respectively, Respectively represent the classifier f p 、f q The standard deviation of the prediction results, X p , X q Respectively represent the characteristic matrices of the business sample data corresponding to the business processing perspectives p and q. Among them, cov(f p ,f q ) represents the classifier f p and f qThe covariance of the prediction results on the training set, where E(·) is the mathematical expectation. It should be noted that p < q aims to avoid duplicate calculations. For example, if p < q is not restricted, cov(f1,f2) and cov(f2,f1) will both be calculated once. The role of the negative sign is to minimize That is, to maximize

[0045] S140: Determine the multi-perspective integrated loss based on the service sample data and sample labels; the multi-perspective integrated loss is used to represent the deviation values predicted by each service processing perspective.

[0046] The multi-perspective integrated loss can be used to evaluate the difference between the service evaluation result obtained by using the sub-classification model and the actual service evaluation result. Since there may be certain errors when using the sub-classification model after it is set up, the sub-classification model can be further optimized in the subsequent process by analyzing the errors.

[0047] Specifically, the multi-perspective integrated loss can be calculated using the formula where R eemp is the multi-perspective integrated loss, V is the number of service processing perspectives, X v is the feature matrix of the service sample data in the v-th service processing perspective, f v is the sub-classification model corresponding to the service processing perspective v, and Y is the sample label.

[0048] S150: Train the service evaluation model by integrating the perspective constraint coefficient and the multi-perspective integrated loss; the service evaluation model is used to evaluate the execution effect of the service.

[0049] After obtaining the perspective constraint coefficient and the multi-perspective integrated loss, the above coefficients can be effectively used for the evaluation effect of the service to complete the training of the service evaluation model.

[0050] After presetting the corresponding service evaluation model, the service evaluation model may not achieve a completely accurate evaluation effect, and the service evaluation model needs to be further optimized. Therefore, the model can be further optimized according to the training results of the model in each training batch to improve the accuracy of the model for user classification.

[0051] Specifically, for example, the service evaluation model can be optimized using the gradient descent method until the number of optimization times reaches the preset iteration threshold or the loss values of the two objective functions are less than the preset loss threshold. The actual optimization process can be set and adjusted based on the details in the specific application and will not be elaborated here.

[0052] Since the perspective constraint coefficient and the multi-perspective integration loss can effectively evaluate the data, the corresponding objective function can be constructed in combination with the perspective constraint coefficient and the multi-perspective integration loss, and then the calculation result of the objective function can be used to evaluate the training effect of the model to achieve the result of optimizing the model.

[0053] Specifically, the objective function can be constructed as L = R emp +αR mcor +βR eemp , where L is the objective function, R emp is the experience loss, where V is the number of business processing perspectives, X v is the feature matrix of the business sample data in the vth business processing perspective, f v is the sub-classification model corresponding to the business processing perspective v, Y is the sample label, α and β are hyperparameters, and R mcor is the viewing angle constraint coefficient, R eemp Ensemble loss for multiple views.

[0054] After the model is optimized, it can be tested. Specifically, based on the test sample x, the formula can be used Test, where F(x) is the business evaluation model, f v is the sub-classification model, x is the test sample, ω1 is the first service category classified by the sub-classifier, and ω2 is the second service category classified by the sub-classifier. The above example is only a formula set when the sub-classification model generates two evaluation results for the service. In actual applications, the corresponding formula for generating the service evaluation model can be determined according to the specific information category, which is not limited to the above example and will not be repeated here.

[0055] The following uses a specific scenario example to introduce the above user classification-based model training method. Figure 2 The figure shows a flow chart of a model training method. After obtaining the training data set, the training sample data in the training data set are divided into different perspectives to obtain different single-perspective data. The experience loss, multi-perspective integrated experience loss and the maximum correlation constraint between perspectives are further calculated for these single-perspective data. The objective function is constructed according to these parameters, and the constructed objective function is used to train the business evaluation model, so as to obtain a model that can accurately and effectively evaluate the business.

[0056] Based on the introduction of the above embodiments and scenario examples, it can be seen that after obtaining the business sample data, the method divides the business sample data into single-perspective data corresponding to multiple business processing perspectives, and then combines the business sample data corresponding to different business processing perspectives and the sub-classification model to calculate the perspective constraint coefficient to determine the correlation between the data from different business processing perspectives, and then determines the multi-perspective integration loss based on the business sample data and the sample label, determines the predicted deviation value, and trains the business evaluation model by combining the above perspective constraint coefficient and multi-perspective integration loss to evaluate the business. After the business data is input into the trained model, the business evaluation results based on the output of the model can be used to process the business. Through the above method, the correlation between different perspectives corresponding to the business data is comprehensively considered, the training effect of the model is improved, the accuracy of the business evaluation is improved, and the effective execution of the business is guaranteed.

[0057] based on Figure 1 Corresponding to the model training method based on business evaluation, the embodiment of this specification also proposes a business processing method. The execution subject of the business processing method can be a business processing device, and the business processing device includes but is not limited to a server, an industrial computer, a PC, etc. Figure 3 As shown, the business processing method includes the following specific implementation steps.

[0058] S310: Acquire target business data of the target business.

[0059] The target business may be a business that needs to be evaluated to determine whether it needs to be processed or how to process it. The target business data is data corresponding to the target business. For a specific introduction to the target business and the target business data, please refer to the description in step S110, which will not be repeated here.

[0060] S320: Input the target business data into the business evaluation model to obtain the business evaluation result; the business evaluation model is obtained in the following manner: obtain business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; divide the business sample data into single-perspective data corresponding to at least two business processing perspectives; the business processing perspective is used to indicate different categories of data; the business processing perspectives correspond to pre-set sub-classification models; the perspective constraint coefficient is calculated by combining the business sample data and sub-classification models of each business processing perspective; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; determine the multi-perspective integration loss based on the business sample data and sample labels; the multi-perspective integration loss is used to represent the deviation value predicted by each business processing perspective; train the business evaluation model by combining the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business.

[0061] After obtaining the target business data, the target business data can be directly input into the business evaluation model to obtain the corresponding business evaluation result. The business evaluation result can be a result obtained based on two classifications, for example, a low risk evaluation result and a high risk evaluation result; the business evaluation result can also have multiple classification results, so that in subsequent steps, each classification result is processed separately according to the processing method corresponding to the classification result.

[0062] For a detailed introduction to the business evaluation model, please refer to Figure 1 The introduction of the corresponding model training method based on user classification will not be repeated here.

[0063] S330: Determine a service processing method corresponding to the target service based on the service evaluation result.

[0064] Based on different business assessment results, corresponding business processing methods can be pre-set. For example, when the business assessment results include high-risk assessment results and low-risk assessment results, the high-risk assessment results can be chosen not to be processed, and the low-risk assessment results can be processed normally. In the case where the business assessment results include multiple categories, different processing methods can be set for different categories, such as different audit strengths, to ensure normal and effective processing of the business.

[0065] The specific correspondence between the business evaluation results and the business processing methods can be directly specified by the management personnel, or can be obtained through training based on historical data. The specific acquisition method can be set based on the actual application situation, which will not be repeated here.

[0066] The above business processing method is introduced using a scenario example. Figure 4 The figure shows a flowchart of a business processing process, wherein after obtaining the corresponding business data from the data warehouse, the business data is preprocessed, and after extracting the corresponding data features, the data is divided into training samples and test samples, the training samples are used to train the model, and the final business evaluation model is obtained in combination with the test samples, thereby obtaining the final prediction result to realize the evaluation of the corresponding business, and then the processing method for the business can be determined according to the evaluation result.

[0067] based on Figure 1 The corresponding model training method based on business evaluation introduces a model training device based on business evaluation in an embodiment of this specification. The model training device based on business evaluation can be set in a model training device based on business evaluation. Figure 5 As shown, the business evaluation-based model training device includes the following modules.

[0068] The business sample data acquisition module 510 is used to acquire business sample data; the business sample data corresponds to a sample tag; the sample tag is used to describe the business execution result.

[0069] The single-perspective data division module 520 is used to divide the business sample data into single-perspective data corresponding to at least two business processing perspectives; the business processing perspectives are used to indicate different categories of data; the business processing perspectives correspond to pre-set sub-classification models.

[0070] The perspective constraint coefficient calculation module 530 is used to calculate the perspective constraint coefficient by integrating the business sample data of each business processing perspective and the sub-classification model; the perspective constraint coefficient is used to represent the correlation of data of different business processing perspectives.

[0071] The multi-view integrated loss determination module 540 is used to determine the multi-view integrated loss based on the business sample data and the sample label; the multi-view integrated loss is used to represent the deviation value predicted by each business processing perspective.

[0072] The business evaluation model training module 550 is used to train a business evaluation model by comprehensively considering the view constraint coefficient and the multi-view integrated loss; the business evaluation model is used to evaluate the execution effect of the business.

[0073] based on Figure 4 The corresponding service processing method introduces a service processing device in the embodiment of this specification. The service processing device is set in the service processing device. Figure 6 As shown, the business processing device includes the following modules.

[0074] The target business data acquisition module 610 is used to acquire target business data of a target business.

[0075] The business evaluation result acquisition module 620 is used to input the target business data into the business evaluation model to obtain the business evaluation result; the business evaluation model is obtained in the following manner: obtaining business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; dividing the business sample data into single-perspective data corresponding to at least two business processing perspectives; the business processing perspective is used to indicate different categories of data; the business processing perspectives correspond to pre-set sub-classification models; the perspective constraint coefficient is calculated by combining the business sample data and sub-classification models of each business processing perspective; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; determining the multi-perspective integration loss based on the business sample data and sample labels; the multi-perspective integration loss is used to represent the deviation value predicted by each business processing perspective; training the business evaluation model by combining the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business.

[0076] The service processing method determination module 630 is used to determine the service processing method corresponding to the target service based on the service evaluation result.

[0077] based on Figure 1 Corresponding to the model training method based on business evaluation, this specification embodiment provides a model training device based on business evaluation. Figure 7 As shown, the business evaluation-based model training device may include a memory and a processor.

[0078] In this embodiment, the memory may be implemented in any appropriate manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid-state hard disk, or a USB flash drive, etc. The memory may be used to store computer program instructions.

[0079] In this embodiment, the processor can be implemented in any appropriate manner. For example, the processor can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, etc. The processor can execute the computer program instructions to implement the following steps: obtain business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; the business sample data is divided into single-view data corresponding to at least two business processing perspectives; the business processing perspective is used to indicate different categories of data; the business processing perspective corresponds to a pre-set sub-classification model; the business sample data and the sub-classification model of each business processing perspective are combined to calculate the perspective constraint coefficient; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; based on the business sample data and the sample label, the multi-view integration loss is determined; the multi-view integration loss is used to represent the deviation value predicted by each business processing perspective; the business evaluation model is trained by combining the perspective constraint coefficient and the multi-view integration loss; the business evaluation model is used to evaluate the execution effect of the business.

[0080] based on Figure 4 Corresponding to the service processing method, this specification embodiment provides a service processing device. Figure 8 As shown, the business processing device may include a memory and a processor.

[0081] In this embodiment, the memory may be implemented in any appropriate manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid-state hard disk, or a USB flash drive, etc. The memory may be used to store computer program instructions.

[0082] In this embodiment, the processor may be implemented in any suitable manner. For example, the processor may take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro)processor, a logic gate, a switch, an application specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller, etc. The processor can execute the computer program instructions to implement the following steps: obtain target business data of the target business; input the target business data into a business evaluation model to obtain a business evaluation result; the business evaluation model is obtained in the following manner: obtain business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; divide the business sample data into single-perspective data corresponding to at least two business processing perspectives respectively; the business processing perspective is used to indicate different categories of data; the business processing perspectives correspond to pre-set sub-classification models respectively; the business sample data and sub-classification models of each business processing perspective are combined to calculate the perspective constraint coefficient; the perspective constraint coefficient is used to represent the correlation of data from different business processing perspectives; determine the multi-perspective integration loss based on the business sample data and sample labels; the multi-perspective integration loss is used to represent the deviation value predicted by each business processing perspective; train the business evaluation model by combining the perspective constraint coefficient and the multi-perspective integration loss; the business evaluation model is used to evaluate the execution effect of the business; determine the business processing method corresponding to the target business based on the business evaluation result.

[0083] It should be noted that the above-mentioned model training methods, business processing methods, devices and equipment can be applied to the field of artificial intelligence technology, and can also be applied to other technical fields except the field of artificial intelligence technology, without any limitation.

[0084] In the 1990s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the method flow). However, with the development of technology, many improvements to the method flow today can be regarded as direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask a chip manufacturer to design and produce a dedicated integrated circuit chip. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0085] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0086] It can be known from the above description of the implementation mode that the technicians in this field can clearly understand that the present specification can be implemented by means of software plus the necessary first hardware platform. Based on such an understanding, the technical solution of the present specification can be essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present specification or some parts of the embodiments.

[0087] 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 system 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.

[0088] This specification can be used in many first or special computer system environments or configurations, such as personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc.

[0089] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0090] Although the present specification is described through embodiments, those skilled in the art will appreciate that there are many modifications and changes to the present specification without departing from the spirit of the present specification, and it is intended that the appended claims include these modifications and changes without departing from the spirit of the present specification.

Claims

1. A model training method based on business evaluation, characterized in that: include: Acquire business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; The business sample data is divided into single-view data corresponding to at least two business processing perspectives respectively; the business processing perspectives are used to indicate different categories of business sample data; the single-view data includes business sample data under the same business processing perspective, and the business processing perspectives respectively correspond to pre-set sub-classification models, and the sub-classification models are used to determine business evaluation results of the business sample data under the corresponding business processing perspectives; Calculate the perspective constraint coefficient by integrating the business sample data and sub-classification model of each business processing perspective In the formula, R mcor is the perspective constraint coefficient, V is the number of business processing perspectives, p and q are used to refer to different business processing perspectives, and f p 、f q are the sub-classification models corresponding to the business processing perspectives p and q respectively, Respectively represent the classifier f p 、f q The standard deviation of the prediction results, X p , X q Respectively represent the characteristic matrices of the business sample data corresponding to the business processing perspectives p and q; the perspective constraint coefficient is used to represent the correlation of the business sample data under different business processing perspectives; Determine a multi-view integration loss based on the business sample data and the sample label; the multi-view integration loss is used to evaluate the difference between the business evaluation result obtained by using the sub-classification model and the actual business evaluation result; The objective function is constructed by combining the perspective constraint coefficient and the multi-perspective integrated loss, and the business evaluation model is optimized according to the objective function; the business evaluation model is used to evaluate the execution effect of the business.

2. The method according to claim 1, characterized in that Before the comprehensive viewing angle constraint coefficient and multi-view integrated loss training service evaluation model, the method further includes: Determining the data distribution density of the business sample data; Determine the weight of each business sample data based on the data distribution density; Accordingly, the comprehensive viewing angle constraint coefficient and multi-view integrated loss training service evaluation model includes: Combined with the weight of the business sample data, the perspective constraint coefficient and the multi-perspective integrated loss are used to train the business evaluation model.

3. The method according to claim 1, characterized in that Before dividing the business sample data into single-view data corresponding to at least two business processing perspectives, the method further includes: The business sample data is preprocessed; the preprocessing includes: completing the business sample data based on preset feature fields.

4. The method according to claim 1, characterized in that The business processing perspective includes at least one of a basic information perspective, an annual report information perspective, a tax information perspective, a change information perspective, a fund transaction information perspective, and a news and public opinion perspective.

5. The method according to claim 1, characterized in that The dividing the business sample data into single-view data corresponding to at least two business processing perspectives respectively includes: Determine the statistical information and deviation characteristics of the business sample data; the statistical information includes at least one of a maximum value, a minimum value, a mean value, and a variance; the deviation characteristics include at least one of a minimum value difference, a maximum value difference, and a mean value difference between the derived feature and the original feature; The business sample data is divided into single-view data corresponding to at least two business processing perspectives according to the statistical information and the deviation value characteristics.

6. The method according to claim 1, characterized in that The determining of the multi-view integration loss based on the business sample data and the sample labels includes: Using the formula Calculate the multi-view integration loss, where R eemp is the multi-view integration loss, V is the number of business processing perspectives, X v is the feature matrix of the business sample data in the vth business processing perspective, f v is the sub-classification model corresponding to the business processing perspective v, and Y is the sample label.

7. The method according to claim 1, characterized in that The objective function is constructed by integrating the perspective constraint coefficient and the multi-perspective integrated loss, including: Construct the objective function as L = R emp +αR mcor +βR eemp , where L is the objective function, R emp is the experience loss, where V is the number of business processing perspectives, X v is the feature matrix of the business sample data in the vth business processing perspective, f v is the sub-classification model corresponding to the business processing perspective v, Y is the sample label, α and β are hyperparameters, and R mcor is the viewing angle constraint coefficient, R eemp Ensemble loss for multiple views.

8. The method according to claim 1, characterized in that Optimizing the business evaluation model according to the objective function includes: The business evaluation model is optimized using the gradient descent method until the number of optimizations reaches a preset iteration threshold or the loss value of two objective functions is less than a preset loss threshold.

9. A model training device based on business evaluation, characterized in that: include: A business sample data acquisition module is used to acquire business sample data; the business sample data corresponds to a sample label; The sample label is used to describe the business execution result; A single-view data division module, used to divide the business sample data into single-view data corresponding to at least two business processing perspectives; the business processing perspectives are used to indicate different categories of business sample data; the single-view data includes business sample data under the same business processing perspective, and the business processing perspectives respectively correspond to pre-set sub-classification models, and the sub-classification models are used to determine the business evaluation results of the business sample data under the corresponding business processing perspectives; The view constraint coefficient calculation module is used to calculate the view constraint coefficient by integrating the business sample data and sub-classification model of each business processing view In the formula, R mcor is the perspective constraint coefficient, V is the number of business processing perspectives, p and q are used to refer to different business processing perspectives, and f p 、f q are the sub-classification models corresponding to the business processing perspectives p and q respectively, Respectively represent the classifier f p 、f q The standard deviation of the prediction results, X p , X q Respectively represent the characteristic matrices of the business sample data corresponding to the business processing perspectives p and q; the perspective constraint coefficient is used to represent the correlation of the business sample data under different business processing perspectives; A multi-view integrated loss determination module, used to determine the multi-view integrated loss based on the business sample data and the sample label; the multi-view integrated loss is used to evaluate the difference between the business evaluation result obtained by using the sub-classification model and the actual business evaluation result; The business evaluation model training module is used to construct an objective function by comprehensively considering the perspective constraint coefficient and the multi-perspective integrated loss, and optimize the business evaluation model according to the objective function; the business evaluation model is used to evaluate the execution effect of the business.

10. A model training device based on business evaluation, comprising a memory and a processor; The memory is used to store computer program instructions; The processor is used to execute the computer program instructions to implement the following steps: obtaining business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; dividing the business sample data into single-perspective data corresponding to at least two business processing perspectives; the business processing perspective is used to indicate different categories of business sample data; the single-perspective data includes business sample data under the same business processing perspective, and the business processing perspectives respectively correspond to preset sub-classification models, and the sub-classification models are used to determine the business evaluation results of the business sample data under the corresponding business processing perspective; comprehensively calculating the perspective constraint coefficient by combining the business sample data and the sub-classification models of each business processing perspective In the formula, R mcor is the perspective constraint coefficient, V is the number of business processing perspectives, p and q are used to refer to different business processing perspectives, and f p 、f q are the sub-classification models corresponding to the business processing perspectives p and q respectively, Respectively represent the classifier f p 、f q The standard deviation of the prediction results, X p , X q Respectively represent the feature matrices of the business sample data corresponding to the business processing perspectives p and q; the perspective constraint coefficient is used to represent the correlation of the business sample data under different business processing perspectives; the multi-perspective integration loss is determined based on the business sample data and the sample label; the multi-perspective integration loss is used to evaluate the difference between the business evaluation results obtained using the sub-classification model and the actual business evaluation results; the objective function is constructed by combining the perspective constraint coefficient and the multi-perspective integration loss, and the business evaluation model is optimized according to the objective function; the business evaluation model is used to evaluate the execution effect of the business.

11. A business processing method, characterized in that: include: Acquire target business data of target business; The target business data is input into the business evaluation model to obtain the business evaluation result; the business evaluation model is obtained in the following manner: obtaining business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; the business sample data is divided into single-perspective data corresponding to at least two business processing perspectives; the business processing perspective is used to indicate different categories of business sample data; the single-perspective data includes business sample data under the same business processing perspective, and the business processing perspectives correspond to pre-set sub-classification models, and the sub-classification models are used to determine the business evaluation results of the business sample data under the corresponding business processing perspective; the perspective constraint coefficient is calculated by combining the business sample data and the sub-classification models of each business processing perspective. In the formula, R mcor is the perspective constraint coefficient, V is the number of business processing perspectives, p and q are used to refer to different business processing perspectives, and f p 、f q are the sub-classification models corresponding to the business processing perspectives p and q respectively, Respectively represent the classifier f p 、f q The standard deviation of the prediction results, X p , X q Respectively represent the characteristic matrices of the business sample data corresponding to the business processing perspectives p and q; the perspective constraint coefficient is used to represent the correlation of the business sample data under different business processing perspectives; Determine a multi-view integration loss based on business sample data and sample labels; the multi-view integration loss is used to evaluate the difference between the business evaluation result obtained by using the sub-classification model and the actual business evaluation result; construct an objective function by combining the view constraint coefficient and the multi-view integration loss, and optimize the business evaluation model according to the objective function; the business evaluation model is used to evaluate the execution effect of the business; A business processing method corresponding to the target business is determined based on the business evaluation result.

12. A service processing device, characterized in that: include: A target business data acquisition module is used to acquire target business data of a target business; A business evaluation result acquisition module is used to input the target business data into a business evaluation model to obtain a business evaluation result; the business evaluation model is obtained in the following manner: obtaining business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; the business sample data is divided into single-perspective data corresponding to at least two business processing perspectives; the business processing perspective is used to indicate different categories of business sample data; the single-perspective data includes business sample data under the same business processing perspective, and the business processing perspectives respectively correspond to pre-set sub-classification models, and the sub-classification models are used to determine the business evaluation results of the business sample data under the corresponding business processing perspective; the perspective constraint coefficient is calculated by integrating the business sample data and the sub-classification models of each business processing perspective. In the formula, R mcor is the perspective constraint coefficient, V is the number of business processing perspectives, p and q are used to refer to different business processing perspectives, and f p 、f q are the sub-classification models corresponding to the business processing perspectives p and q respectively, Respectively represent the classifier f p 、f q The standard deviation of the prediction results, X p , X q Respectively represent the characteristic matrices of the business sample data corresponding to the business processing perspectives p and q; the perspective constraint coefficient is used to represent the correlation of the business sample data under different business processing perspectives; Determine a multi-view integration loss based on business sample data and sample labels; the multi-view integration loss is used to evaluate the difference between the business evaluation result obtained by using the sub-classification model and the actual business evaluation result; construct an objective function by combining the view constraint coefficient and the multi-view integration loss, and optimize the business evaluation model according to the objective function; the business evaluation model is used to evaluate the execution effect of the business; The service processing mode determination module is used to determine the service processing mode corresponding to the target service based on the service evaluation result.

13. A business processing device comprising a memory and a processor; The memory is used to store computer program instructions; The processor is used to execute the computer program instructions to implement the following steps: obtain target business data of the target business; input the target business data into a business evaluation model to obtain a business evaluation result; the business evaluation model is obtained in the following manner: obtain business sample data; the business sample data corresponds to a sample label; the sample label is used to describe the business execution result; divide the business sample data into single-perspective data corresponding to at least two business processing perspectives; the business processing perspective is used to indicate different categories of business sample data; the single-perspective data includes business sample data under the same business processing perspective, and the business processing perspectives respectively correspond to pre-set sub-classification models, and the sub-classification models are used to determine the business evaluation results of the business sample data under the corresponding business processing perspective; comprehensively calculate the perspective constraint coefficient by combining the business sample data and the sub-classification models of each business processing perspective In the formula, R mcor is the perspective constraint coefficient, V is the number of business processing perspectives, p and q are used to refer to different business processing perspectives, and f p 、f q are the sub-classification models corresponding to the business processing perspectives p and q respectively, Respectively represent the classifier f p 、f q The standard deviation of the prediction results, X p , X q Respectively represent the characteristic matrices of the business sample data corresponding to the business processing perspectives p and q; the perspective constraint coefficient is used to represent the correlation of the business sample data under different business processing perspectives; the multi-perspective integration loss is determined based on the business sample data and the sample label; the multi-perspective integration loss is used to evaluate the difference between the business evaluation results obtained by using the sub-classification model and the actual business evaluation results; the objective function is constructed by combining the perspective constraint coefficient and the multi-perspective integration loss, and the business evaluation model is optimized according to the objective function; the business evaluation model is used to evaluate the execution effect of the business; based on the business evaluation result, the business processing method corresponding to the target business is determined.