CN113515867A - Model training method, business processing method, device and equipment - Google Patents
Model training method, business processing method, device and equipment Download PDFInfo
- Publication number
- CN113515867A CN113515867A CN202110854994.3A CN202110854994A CN113515867A CN 113515867 A CN113515867 A CN 113515867A CN 202110854994 A CN202110854994 A CN 202110854994A CN 113515867 A CN113515867 A CN 113515867A
- Authority
- CN
- China
- Prior art keywords
- service
- view
- data
- sample data
- processing
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The embodiment of the specification provides a model training method, a business processing device and equipment, which can be applied to the technical field of artificial intelligence. The method comprises the following steps: acquiring service sample data; the service sample data corresponds to a sample label; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; determining multi-view integration loss based on the service sample data and the sample label; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service. The method comprehensively considers the relevance between different visual angles corresponding to the business data, improves the training effect of the model, improves the accuracy of business evaluation, and ensures the effective execution of the business.
Description
Technical Field
The embodiment of the specification relates to the technical field of artificial intelligence, in particular to a model training method, a business processing device and equipment.
Background
The service evaluation is generally a necessary step before the service of other clients or enterprises is processed, for example, for the service evaluation, the execution capacity, performance capability and trust level of the client or the enterprise can be comprehensively evaluated, and then whether to process the service or select a corresponding service processing mode is determined according to the evaluation result, so that the service processing effect is ensured, and the risk brought by the service is reduced.
At present, when a service is evaluated, the service is generally classified from a plurality of visual angles, and then a corresponding service evaluation result is determined by combining the characteristics of data in the visual angles. However, although the above evaluation method includes information of a plurality of views, since different views are independent of each other, information included in a single view is insufficient to provide classification information of a sample, and the classification effect of a model is rather reduced. In addition, the way of respectively evaluating the services by using multiple viewing angles ignores the correlation between different viewing angles, also affects the accuracy of the evaluation result, and is not favorable for the effective processing of the services. Therefore, a method for evaluating a service accurately and effectively to achieve better service processing is needed.
Disclosure of Invention
An object of the embodiments of the present specification is to provide a model training method, a service distribution method, an apparatus, and a device, so as to solve a problem how to effectively evaluate a service to improve an effect of service execution.
In order to solve the above technical problem, an embodiment of the present specification provides a model training method based on business evaluation, including: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
An embodiment of the present specification further provides a model training apparatus based on business evaluation, including: a service sample data obtaining module, configured to obtain service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; the single-view data dividing module is used for dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; the visual angle constraint coefficient calculation module is used for calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; the multi-view integration loss determining module is used for determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; the service evaluation model training module is used for integrating the visual angle constraint coefficient and the multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
The embodiment of the present specification further provides a model training device based on business evaluation, including a memory and a processor; the memory to store computer program instructions; the processor to execute the computer program instructions to implement the steps of: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
In order to solve the above technical problem, an embodiment of the present specification further provides a service processing method, including: acquiring target service data of a target service; inputting the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service; and determining a service processing mode corresponding to the target service based on the service evaluation result.
An embodiment of this specification further provides a service processing apparatus, including: the target service data acquisition module is used for acquiring target service data of a target service; the service evaluation result acquisition module is used for inputting the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service; and the service processing mode determining module is used for determining a service processing mode corresponding to the target service based on the service evaluation result.
The embodiment of the present specification further provides a service processing device, which includes a memory and a processor; the memory to store computer program instructions; the processor to execute the computer program instructions to implement the steps of: acquiring target service data of a target service; inputting the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service; and determining a service processing mode corresponding to the target service based on the service evaluation result.
As can be seen from the technical solutions provided by the embodiments of the present specification, after acquiring service sample data, the embodiments of the present specification divide the service sample data into single view data corresponding to multiple service processing views, calculate a view constraint coefficient by integrating the service sample data and the sub-classification models corresponding to different service processing views to determine correlation between the data at different service processing views, determine a multi-view integration loss according to the service sample data and the sample labels, determine a predicted deviation value, and train a service evaluation model by integrating the view constraint coefficient and the multi-view integration loss to realize service evaluation. After the business data is input into the trained model, the business can be processed based on the business evaluation result output by the model. By the method, the relevance between different visual angles corresponding to the business data is comprehensively considered, the training effect of the model is improved, the accuracy of business evaluation is improved, and the effective execution of the business is guaranteed.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments described in the specification, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a flowchart illustrating a method for model training based on business evaluation according to an embodiment of the present disclosure;
FIG. 2 is a flowchart illustrating a multi-view-based model training process according to an embodiment of the present disclosure;
FIG. 3 is a flow chart illustrating a business evaluation-based process according to an embodiment of the present disclosure;
FIG. 4 is a flow chart of a model training method according to an embodiment of the present disclosure;
FIG. 5 is a block diagram of a model training apparatus based on business evaluation according to an embodiment of the present disclosure;
FIG. 6 is a block diagram of a model training apparatus according to an embodiment of the present disclosure;
FIG. 7 is a block diagram of a model training device based on business evaluation according to an embodiment of the present disclosure;
fig. 8 is a block diagram of a model training apparatus according to an embodiment of the present disclosure.
Detailed Description
The technical solutions in the embodiments of the present disclosure will be clearly and completely described below with reference to the drawings in the embodiments of the present disclosure, and it is obvious that the described embodiments are only a part of the embodiments of the present disclosure, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments in the present specification without any creative effort shall fall within the protection scope of the present specification.
In order to solve the above technical problem, a model training method based on business evaluation in the embodiment of the present specification is first introduced. The execution subject of the model training method based on the business evaluation is model training equipment based on the business evaluation, and the model training equipment based on the business evaluation comprises but is not limited to a server, an industrial personal computer, a PC and the like. As shown in fig. 1, the model training method based on business evaluation may include the following implementation steps.
S110: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result.
The traffic sample data may be sample data used for training the model. The service sample data may be a processing record corresponding to the service, information of the service itself, and corresponding information of a user corresponding to the service.
The service sample data may correspond to a sample tag. The sample label is used for describing the execution result of the service. Based on the execution result corresponding to the sample label, the information such as the execution difficulty of the service can be reflected, and further the service evaluation is realized. For example, performing a successful transaction may have a better assessment result.
In some embodiments, the traffic sample data may include tagged data and untagged data. In the field of machine learning, the modeling learning based on part of labeled data and part of unlabeled data has the characteristics of rapidness and accuracy, and has a good application value.
The labeled data is data labeled with a corresponding sample label, and correspondingly, the unlabeled data does not have the sample label. Model training can be performed in a semi-supervised learning mode based on labeled data and unlabeled data, so that the time spent on labeling sample labels is reduced while the training accuracy is ensured.
Preferably, in order to reduce the time consumed by tagging data, the tagged data may have only a small proportion of the whole corresponding to the service sample data.
In some embodiments, after the service sample data is obtained, the service sample data may be further preprocessed. The pretreatment comprises the following steps: and completing the service sample data based on a preset characteristic field. In order to ensure that the service sample data can be effectively utilized in the subsequent process, the missing value column in the service sample data can be complemented. The preset feature field may be a field corresponding to a different completion rule. For example, for a missing value of a numerical feature in the original feature, the preset feature field may be complemented by a value of "0"; for missing values where the non-numeric feature is missing, the default feature field may be completed for an "unknown" value. In practical application, the completion of the original feature can be realized by using other preset feature fields according to requirements, which is not limited to the above examples and is not described herein again.
S120: dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; and the service processing visual angles correspond to preset sub-classification models respectively.
The service processing view can be used for representing different information types, so that the service processing view is used for distinguishing service sample data, namely different types of the data are indicated.
In some embodiments, the business process perspective may include at least one of a basic information perspective, an annual newspaper information perspective, a tax payment information perspective, a change information perspective, a financial transaction information perspective, and a news public opinion perspective.
Specifically, when the business sample data is data for an enterprise, different view angles of the sample can be constructed based on categories to which the features belong, specifically, the enterprise basic information forms a basic information view angle, the annual report information of the enterprise forms an annual report information view angle, the tax payment information of the enterprise forms a tax payment information view angle, the change information of the enterprise forms a change information view angle, the fund to information of the enterprise forms a fund to information view angle, and the public opinion information forms a news public opinion view angle.
Based on the various service processing perspectives and the corresponding characteristics of the data at different service processing perspectives, the service sample data can be divided to obtain corresponding data at different service processing perspectives.
The single view data is the data corresponding to a certain service processing view after the division is finished. It should be noted that different single-view data may be completely different from each other, or some of the same data may exist, which is not limited to this.
In some embodiments, when dividing the single-view data, statistical information and deviation value characteristics corresponding to the traffic sample data may also be determined first. The statistical information is used to describe the overall characteristic condition of the service sample data, and may specifically include at least one of a maximum value, a minimum value, a mean value, and a variance. The deviation value feature may be used to describe a comparison result between the evolved derived feature and the original feature after the service sample data is evolved, and specifically may include at least one of a minimum difference value, a maximum difference value, and a mean difference value between the derived feature and the original feature.
After the statistical information and the deviation value characteristics are obtained, the information can describe the properties of the service sample characteristics, so that the service sample data can be divided into single-view data by combining the statistical information and the deviation value characteristics. The specific division rule may be set based on the requirement of the actual application, and is not described herein again.
In some embodiments, in order to ensure reasonable utilization of the service sample data, the data distribution density of the service sample data may also be obtained. The data distribution density can be determined according to the spatial distribution condition of the data, so that the distribution condition of the main body of the service sample data can be acquired.
After the data distribution density is obtained, the weight of each service sample data can be determined according to the data distribution density, and the steps of calculating a visual angle constraint coefficient, determining multi-visual angle integration loss, training a service evaluation model and the like are executed based on the weight.
By constructing the data distribution density, the larger weight of the sample with high local density is distributed, and the smaller weight of the sample with low local density can prevent the abnormal sample from influencing the learning process of the model, thereby ensuring the accuracy of model training.
S130: calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficients are used to represent the correlation of data from different traffic processing views.
Each business processing perspective can be preset with a sub-classification model. The sub-classification model may determine a corresponding service evaluation result based on the service sample data at the current view angle. Due to the fact that the corresponding types of data correspond to different service processing visual angles, the accuracy of the sub-classification model for service evaluation is guaranteed.
In some embodiments, the sub-classification model may classify the business in combination with the execution effect of the business. The output result of the sub-classification model may be combined with the sample label corresponding to the service sample data. Specifically, the sub-classification model may determine a first service class and a second service class, where the first service class and the second service class may be used to respectively represent a service that can be effectively executed and a service that cannot be effectively executed, and further may determine a service evaluation result based on an output result of the sub-classification model.
Specifically, when the view angle constraint coefficient is calculated, the service sample data and the sub-classification model may be substituted into a formula
Calculating a view angle constraint coefficient, wherein RmcorFor view constraint coefficients, V is the number of traffic processing views, p and q are used to refer to different traffic processing views, fp、fqSub-classification models corresponding to traffic processing views p, q,
respectively representing classifiers fp、fqStandard deviation of predicted outcome, Xp、XqAnd respectively representing the feature matrixes of the service sample data corresponding to the service processing visual angles p and q. Wherein, cov (f)p,fq) Representation classifier fpAnd fqThe covariance of the results is predicted on the training set, E (-) as the mathematical expectation. In addition, p is<q aims to avoid duplicate calculations, for example: if not limiting p<q，cov(f1,f2) And cov (f)2,f1) Will count once. The effect of the negative sign is to minimize in the process of minimizing the objective function
I.e. to maximise
S140: determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the deviation value of each service processing view prediction.
The multi-view integration 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. After the sub-classification model is set, certain errors may exist when the sub-classification model is utilized, so that the sub-classification model can be further optimized in a subsequent process by analyzing the errors.
In particular, a formula may be utilized
Calculating the multi-view integration loss, wherein ReempFor multi-view integration loss, V is the number of service processing views, XvFor the feature matrix of the service sample data in the v-th service processing view, fvY is a sample label for the sub-classification model corresponding to the business process perspective v.
S150: integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
After the visual angle constraint coefficient and the multi-visual angle integration loss are obtained, the evaluation effect of the coefficient on the service can be effectively utilized, and the training of a service evaluation model is completed.
After the corresponding service evaluation model is preset, the service evaluation model may not obtain a completely accurate evaluation effect, and further optimization of the service evaluation model is required. Therefore, the model can be further optimized according to the training results of the model in each training batch, so that the accuracy of the model for classifying the user is improved.
Specifically, for example, the service evaluation model may be optimized by using a gradient descent method until the number of times of optimization reaches a preset iteration threshold or a loss value of two objective functions is smaller than a preset loss threshold. The actual optimization process may be set and adjusted based on the details of the specific application, which are not described herein.
Because the visual angle constraint coefficient and the multi-visual angle integration loss can effectively evaluate the data, a corresponding objective function can be constructed by combining the visual angle constraint coefficient and the multi-visual angle integration loss, and then the training effect of the model is evaluated by utilizing the calculation result of the objective function, so that the result of optimizing the model is realized.
Specifically, the objective function may be constructed as L ═ Remp+αRmcor+βReempWherein L is an objective function, RempIs a loss of experience in which, among other things,
v is the number of service processing views, XvFor the feature matrix of the service sample data in the v-th service processing view, fvIs a sub-classification model corresponding to a business processing visual angle v, Y is a sample label, alpha and beta are hyper-parameters, RmcorAs a view angle constraint coefficient, ReempIs a multi-view integration penalty.
After the model is optimized, the model can be tested, specifically, based on the test sample x, a formula can be utilized
Performing a test, wherein F (x) is a service evaluation model, fvIs a sub-classification model, x is a test sample, ω1A first traffic class, ω, divided for the sub-classifiers2A second traffic class divided for the sub-classifiers. The above example is only a formula set in the case that the sub-classification model generates two evaluation results for the service, and in practical application, the corresponding formula for generating the service evaluation model may be determined according to a specific information category, and is not limited to the above example, and is not described herein again.
The following describes the model training method based on user classification by using a specific scenario example, and as shown in fig. 2, the method is a schematic flow chart of the model training method. After the training data set is obtained, the training sample data in the training data set are respectively divided into different visual angles to obtain different single-visual-angle data. The empirical losses, multi-view integration empirical losses and inter-view maximum correlation constraints are further calculated for these single-view data. And constructing an objective function according to the parameters, and training a service evaluation model by using the constructed objective function, thereby obtaining a model capable of accurately and effectively evaluating the service.
Based on the above embodiments and the introduction of the scenario example, it can be seen that, after the service sample data is acquired, the service sample data is divided into single view data corresponding to multiple service processing views, then the view constraint coefficients are calculated by integrating the service sample data corresponding to different service processing views and the sub-classification model to determine the correlation between the data at different service processing views, then the multi-view integration loss is determined according to the service sample data and the sample label, the predicted deviation value is determined, and the service evaluation model is trained by integrating the view constraint coefficients and the multi-view integration loss to realize the service evaluation. After the business data is input into the trained model, the business can be processed based on the business evaluation result output by the model. By the method, the relevance between different visual angles corresponding to the business data is comprehensively considered, the training effect of the model is improved, the accuracy of business evaluation is improved, and the effective execution of the business is guaranteed.
Based on the model training method based on business evaluation corresponding to fig. 1, an embodiment of the present specification further provides a business processing method. The execution main body of the service processing method can be service processing equipment, and the service processing equipment comprises but is not limited to a server, an industrial personal computer, a pc machine and the like. As shown in fig. 3, the service processing method includes the following specific implementation steps.
S310: and acquiring target service data of the target service.
The target service may be a service that needs to be evaluated to determine whether or how processing is needed. The target service data is data corresponding to the target service. For specific introduction of the target service and the target service data, reference may be made to the description in step S110, and details are not described herein again.
S320: inputting the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
After the target service data is obtained, the target service data can be directly input into a service evaluation model to obtain a corresponding service evaluation result. The business evaluation result can be a result obtained based on the second classification, for example, a low risk evaluation result and a high risk evaluation result; the service evaluation result may also have a plurality of classification results, so that in the subsequent steps, the processing is performed according to the processing mode corresponding to each classification result.
For the specific description of the service evaluation model, reference may be made to the description in the user classification-based model training method corresponding to fig. 1, and details are not repeated here.
S330: and determining a service processing mode corresponding to the target service based on the service evaluation result.
Based on different service evaluation results, corresponding service processing modes can be preset. For example, when the service evaluation result includes a high risk evaluation result and a low risk evaluation result, the high risk evaluation result may be selected not to be processed, and the low risk evaluation result may be processed normally. Under the condition that the service evaluation result comprises a plurality of categories, different processing modes can be respectively set for different categories, for example, different auditing strengths can be provided, so as to ensure normal and effective processing of the service.
The specific corresponding relation between the service evaluation result and the service processing mode can be directly specified by a manager or obtained by training based on historical data. The specific obtaining mode may be set based on the actual application situation, and is not described herein again.
A scene example is used to introduce the service processing method, as shown in fig. 4, which is a flow diagram of a service processing process, wherein after corresponding service data is obtained from a data warehouse, the service data is subjected to data preprocessing, corresponding data features are extracted from the data, the data are divided into training samples and testing samples, model training is performed by using the training samples, and a final service evaluation model is obtained by combining the testing samples, so that a final prediction result is obtained, so as to realize evaluation of a corresponding service, and further, a processing mode for the service can be determined according to the evaluation result.
A model training device based on business evaluation according to an embodiment of the present specification is introduced based on the model training method based on business evaluation corresponding to fig. 1. The model training device based on business evaluation can be arranged on model training equipment based on business evaluation. As shown in fig. 5, the model training apparatus based on business evaluation includes the following modules.
A service sample data obtaining module 510, configured to obtain service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result.
A single-view data dividing module 520, configured to divide the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; and the service processing visual angles correspond to preset sub-classification models respectively.
A view constraint coefficient calculation module 530, configured to calculate a view constraint coefficient by integrating the service sample data and the sub-classification model at each service processing view; the view constraint coefficients are used to represent the correlation of data from different traffic processing views.
A multi-view integration loss determining module 540, configured to determine a multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the deviation value of each service processing view prediction.
A service evaluation model training module 550, configured to train a service evaluation model by integrating the view constraint coefficient and the multi-view integration loss; the service evaluation model is used for evaluating the execution effect of the service.
A service processing apparatus according to an embodiment of the present description is introduced based on a service processing method corresponding to fig. 4. The service processing device is arranged in the service processing equipment. As shown in fig. 6, the service processing apparatus includes the following modules.
And a target service data obtaining module 610, configured to obtain target service data of the target service.
A service evaluation result obtaining module 620, configured to input the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
A service processing manner determining module 630, configured to determine a service processing manner corresponding to the target service based on the service evaluation result.
Based on the model training method based on business evaluation corresponding to fig. 1, an embodiment of the present specification provides a model training device based on business evaluation. As shown in FIG. 7, the business valuation based model training apparatus may include a memory and a processor.
In this embodiment, the memory may be implemented in any suitable manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid state disk, a U disk, or the like. The memory may be used to store computer program instructions.
In this embodiment, the processor may be implemented in any suitable manner. For example, the processor may take the form of, for example, a microprocessor or processor and a computer-readable medium that stores computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, an embedded microcontroller, and so forth. The processor may execute the computer program instructions to perform the steps of: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
Based on the service processing method corresponding to fig. 4, an embodiment of the present specification provides a service processing device. As shown in fig. 8, the traffic processing device may include a memory and a processor.
In this embodiment, the memory may be implemented in any suitable manner. For example, the memory may be a read-only memory, a mechanical hard disk, a solid state disk, a U disk, or the like. The memory may be used to store computer program instructions.
In this embodiment, the processor may be implemented in any suitable manner. For example, the processor may take the form of, for example, a microprocessor or processor and a computer-readable medium that stores computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, an embedded microcontroller, and so forth. The processor may execute the computer program instructions to perform the steps of: acquiring target service data of a target service; inputting the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service; and determining a service processing mode corresponding to the target service based on the service evaluation result.
It should be noted that the model training method, the service processing device and the equipment can be applied to the technical field of artificial intelligence, and can also be applied to other technical fields except the technical field of artificial intelligence, which is not limited to this.
In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an Integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Hardware Description Language), traffic, pl (core universal Programming Language), HDCal (jhdware Description Language), lang, Lola, HDL, laspam, hardward Description Language (vhr Description Language), vhal (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.
The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smartphone, 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.
From the above description of the embodiments, it is clear to those skilled in the art that the present specification can be implemented by software plus the necessary first hardware platform. Based on such understanding, the technical solutions of the present specification may be essentially or partially implemented in the form of software products, which may be stored in a storage medium, such as ROM/RAM, magnetic disk, optical disk, etc., and include instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in the embodiments or some parts of the embodiments of the present specification.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the system embodiment, since it is substantially similar to the method embodiment, the description is simple, and for the relevant points, reference may be made to the partial description of the method embodiment.
The description is operational with numerous first or special purpose computing system environments or configurations. For example: personal computers, server computers, hand-held or portable devices, tablet-type devices, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like.
This description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
While the specification has been described with examples, those skilled in the art will appreciate that there are numerous variations and permutations of the specification that do not depart from the spirit of the specification, and it is intended that the appended claims include such variations and modifications that do not depart from the spirit of the specification.
Claims (15)
1. A model training method based on business evaluation is characterized by comprising the following steps:
acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result;
dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models;
calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views;
determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view;
integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
2. The method of claim 1, wherein the integrating the view constraint coefficients and the multi-view integration loss training traffic assessment model further comprises, prior to:
determining the data distribution density of the service sample data;
determining the weight of each service sample data based on the data distribution density;
correspondingly, the integrating the view constraint coefficients and the multi-view integrated loss training service evaluation model includes:
and combining the weight of the service sample data, and integrating the visual angle constraint coefficient and the multi-visual angle integrated loss training service evaluation model.
3. The method of claim 1, wherein before said dividing said traffic sample data into single view data respectively corresponding to at least two traffic processing views, further comprising:
preprocessing the service sample data; the pretreatment comprises the following steps: and completing the service sample data based on a preset characteristic field.
4. The method of claim 1, wherein the business process perspective comprises at least one of a basic information perspective, an annual newspaper information perspective, a tax payment information perspective, a change information perspective, a financial transaction information perspective, and a news public opinion perspective.
5. The method of claim 1, wherein said dividing the traffic sample data into single view data respectively corresponding to at least two traffic processing views comprises:
determining statistical information and deviation value characteristics of the service sample data; the statistical information comprises at least one of a maximum value, a minimum value, a mean value and a variance; the deviation value feature comprises at least one of a minimum difference value, a maximum difference value and a mean difference value between the derivative feature and the original feature;
and dividing the service sample data into single-view data respectively corresponding to at least two service processing views according to the statistical information and the deviation value characteristics.
6. The method of claim 1, wherein said integrating the traffic sample data and the sub-classification models for each traffic processing view to compute the view constraint coefficients comprises:
substituting the service sample data and the sub-classification model of each service processing visual angle into a formula
Calculating a view angle constraint coefficient, wherein RmcorFor view constraint coefficients, V is the number of traffic processing views, p and q are used to refer to different traffic processing views, fp、fqSub-classification models corresponding to traffic processing views p, q,
respectively representing classifiers fp、fqStandard deviation of predicted outcome, Xp、XqAnd respectively representing the feature matrixes of the service sample data corresponding to the service processing visual angles p and q.
7. The method of claim 1, wherein said determining a multi-view integration loss based on traffic sample data and sample labels comprises:
using formulas
Calculating the multi-view integration loss, wherein ReempFor multi-view integration loss, V is the number of service processing views, XvFor the feature matrix of the service sample data in the v-th service processing view, fvY is a sample label for the sub-classification model corresponding to the business process perspective v.
8. The method of claim 1, wherein the synthesizing the view constraint coefficients and the multi-view integration loss training traffic assessment model comprises:
constructing an objective function based on the view constraint coefficients and the multi-view integration loss;
and optimizing a service evaluation model according to the objective function.
9. The method of claim 8, wherein constructing an objective function based on the view constraint coefficients and multi-view integration losses comprises:
constructing the objective function as L ═ Remp+αRmcor+βReempWherein L is an objective function, RempIs a loss of experience in which, among other things,
v is the number of service processing views, XvFor the feature matrix of the service sample data in the v-th service processing view, fvIs a sub-classification model corresponding to a business processing visual angle v, Y is a sample label, alpha and beta are hyper-parameters, RmcorAs a view angle constraint coefficient, ReempIs a multi-view integration penalty.
10. The method of claim 8, wherein optimizing a traffic assessment model according to the objective function comprises:
and optimizing the service evaluation model by using a gradient descent method until the optimization times reach a preset iteration threshold or the loss value of the two objective functions is less than a preset loss threshold.
11. A model training device based on business evaluation is characterized by comprising:
a service sample data obtaining module, configured to obtain service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result;
the single-view data dividing module is used for dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models;
the visual angle constraint coefficient calculation module is used for calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views;
the multi-view integration loss determining module is used for determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view;
the service evaluation model training module is used for integrating the visual angle constraint coefficient and the multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
12. A model training device based on business evaluation comprises a memory and a processor;
the memory to store computer program instructions;
the processor to execute the computer program instructions to implement the steps of: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service.
13. A method for processing a service, comprising:
acquiring target service data of a target service;
inputting the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service;
and determining a service processing mode corresponding to the target service based on the service evaluation result.
14. A traffic processing apparatus, comprising:
the target service data acquisition module is used for acquiring target service data of a target service;
the service evaluation result acquisition module is used for inputting the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service;
and the service processing mode determining module is used for determining a service processing mode corresponding to the target service based on the service evaluation result.
15. A traffic processing device comprising a memory and a processor;
the memory to store computer program instructions;
the processor to execute the computer program instructions to implement the steps of: acquiring target service data of a target service; inputting the target service data into a service evaluation model to obtain a service evaluation result; the business evaluation model is obtained by the following method: acquiring service sample data; the service sample data corresponds to a sample label; the sample label is used for describing a service execution result; dividing the service sample data into single-view data respectively corresponding to at least two service processing views; the service processing view is used for indicating different types of data; the service processing view angles respectively correspond to preset sub-classification models; calculating a visual angle constraint coefficient by integrating the service sample data and the sub-classification model of each service processing visual angle; the view constraint coefficient is used for representing the correlation of data of different service processing views; determining multi-view integration loss based on the service sample data and the sample label; the multi-view integration loss is used for representing the predicted deviation value of each business processing view; integrating the visual angle constraint coefficient and a multi-visual angle integrated loss training service evaluation model; the service evaluation model is used for evaluating the execution effect of the service; and determining a service processing mode corresponding to the target service based on the service evaluation result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110854994.3A CN113515867A (en) | 2021-07-28 | 2021-07-28 | Model training method, business processing method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110854994.3A CN113515867A (en) | 2021-07-28 | 2021-07-28 | Model training method, business processing method, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113515867A true CN113515867A (en) | 2021-10-19 |
Family
ID=78068658
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110854994.3A Pending CN113515867A (en) | 2021-07-28 | 2021-07-28 | Model training method, business processing method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113515867A (en) |
-
2021
- 2021-07-28 CN CN202110854994.3A patent/CN113515867A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110909165B (en) | Data processing method, device, medium and electronic equipment | |
| CN106651057A (en) | Mobile terminal user age prediction method based on installation package sequence table | |
| WO2020232874A1 (en) | Modeling method and apparatus based on transfer learning, and computer device and storage medium | |
| CN111401777A (en) | Enterprise risk assessment method and device, terminal equipment and storage medium | |
| US20210374582A1 (en) | Enhanced Techniques For Bias Analysis | |
| US20200057632A1 (en) | Automatically evaluating software project requirements | |
| CN112990281A (en) | Abnormal bid identification model training method, abnormal bid identification method and abnormal bid identification device | |
| CN111611390B (en) | Data processing method and device | |
| CN114144770A (en) | System and method for generating data sets for model retraining | |
| CN112328869A (en) | User loan willingness prediction method and device and computer system | |
| CN112950347B (en) | Resource data processing optimization method and device, storage medium and terminal | |
| CN113569955A (en) | Model training method, user portrait generation method, device and equipment | |
| CN113379528A (en) | Wind control model establishing method and device and risk control method | |
| CN113159213A (en) | Service distribution method, device and equipment | |
| CN116756281A (en) | Knowledge question-answering method, device, equipment and medium | |
| CN108830302B (en) | Image classification method, training method, classification prediction method and related device | |
| CN112446777A (en) | Credit evaluation method, device, equipment and storage medium | |
| CN116229211A (en) | Sample generation method, model training method, object detection method and device | |
| CN113591881B (en) | Intention recognition method and device based on model fusion, electronic equipment and medium | |
| CN113515867A (en) | Model training method, business processing method, device and equipment | |
| CN114897607A (en) | Data processing method and device for product resources, electronic equipment and storage medium | |
| CN110570301B (en) | Risk identification method, device, equipment and medium | |
| CN114493853A (en) | Credit rating evaluation method, credit rating evaluation device, electronic device and storage medium | |
| CN112950225A (en) | Customer category determination method, device and storage medium | |
| CN112632284A (en) | Information extraction method and system for unlabeled text data set |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |