CN115099415A - Prediction method, apparatus, electronic device and readable medium based on network model - Google Patents

Abstract

The application provides a prediction method, a prediction device, electronic equipment and a readable medium based on a network model. The method comprises the following steps: acquiring original model parameters of a target network model, wherein the input of the target network model is K variables; generating T sub-network models according to the original model parameters, wherein each sub-network model in the T sub-network models is used for inputting a variable group, the variable groups input by different sub-network models are different from each other, the T variable groups comprise K variables, and each variable group comprises at least one variable; acquiring data to be predicted, wherein the data to be predicted comprises data characteristics corresponding to each variable in K variables; dividing the data to be predicted into data feature groups corresponding to T variable groups according to the variable groups corresponding to each sub-network model; and inputting a corresponding data characteristic group into each sub-network model respectively to generate a model prediction result of the target network model. The method increases the parallelism of calculation in the model and improves the running speed of the model.

Description

Prediction method, apparatus, electronic device and readable medium based on network model

technical field

The present application relates to the field of computer technology, and in particular, to a prediction method, apparatus, electronic device and readable medium based on a network model.

Background technique

With the rapid development of the field of artificial intelligence, the application of machine learning models has become more and more extensive, and the strategic conclusions and label information obtained by various businesses through model training have become more and more accurate. A good model can bring risks. Great improvement in all fields.

In related technologies, the application of machine learning models is usually to use a large amount of data for offline training and calculation.

However, such a model is usually a relatively complex calculation process, so it takes a long time to output the results each time the calculation is performed, and the real-time performance of the calculation is poor, making it difficult to perform real-time calculation and application.

SUMMARY OF THE INVENTION

Based on the above technical problems, the present application provides a prediction method, device, electronic device and readable medium based on a network model, so as to increase the parallelism of calculation in the model, improve the running speed of the model, and thereby improve the real-time performance of calculation, so as to for applications in real-time computing applications.

Other features and advantages of the present application will become apparent from the following detailed description, or be learned in part by practice of the present application.

According to an aspect of the embodiments of the present application, a method for predicting a method based on a network model is provided, including:

Obtain the original model parameters of the target network model, wherein the input of the target network model is K variables, and the K is an integer greater than 1;

T sub-network models are generated according to the original model parameters, wherein each sub-network model in the T sub-network models is used to input a variable group, and the variable groups input by different sub-network models are different from each other, and the T variable groups Including the K variables, each variable group includes at least one variable;

Obtaining data to be predicted, wherein the data to be predicted includes data features corresponding to each of the K variables;

According to the variable group corresponding to each sub-network model, the to-be-predicted data is divided into data feature groups corresponding to the T variable groups;

Corresponding data feature groups are respectively input to each of the sub-network models to generate a model prediction result of the target network model.

In some embodiments of the present application, based on the above technical solutions, the generating T sub-network models according to the original model parameters includes:

According to the implementation language of the original model parameters, the original model parameters are analyzed to obtain a plurality of calculation units, and the calculation units are used to calculate the output result based on the corresponding variable group;

According to the dependencies between the input parameters and the output results of the plurality of computing units, the computing units with mutual dependencies are combined into a sub-network model to obtain T sub-network models.

In some embodiments of the present application, based on the above technical solutions, the generating T sub-network models according to the original model parameters includes:

According to the implementation language of the original model parameters, the original model parameters are parsed to obtain a first calculation unit, a second calculation unit and a third calculation unit, wherein the input parameter of the third calculation unit is the first calculation unit. an output result of a computing unit and the second computing unit;

Based on the first computing unit, the second computing unit, and the third computing unit, a sub-network model is generated, and the T sub-network models are obtained.

In some embodiments of the present application, based on the above technical solutions, the obtaining data to be predicted includes:

According to the K variables, determine the data query statement corresponding to each variable and the corresponding data source;

The business data characteristics corresponding to the K variable pairs are obtained from the data source according to the data query statement, and the data to be predicted is obtained.

In some embodiments of the present application, based on the above technical solutions, inputting a corresponding data feature group to each sub-network model to generate a model prediction result of the target network model includes:

Input corresponding data feature groups to each of the sub-network models to obtain T output results;

According to the T output results, model fitting is performed through the loss function of the target network model to obtain a model prediction result of the target network model.

In some embodiments of the present application, based on the above technical solutions, after the corresponding data feature groups are respectively input to the each sub-network model to generate the model prediction result of the target network model, the method further includes:

Obtain independent data features from the data to be predicted, and the independent data features do not correspond to variables in the K variables;

According to the independent data feature and the model prediction result, a decision rule corresponding to the to-be-predicted data is determined, and the decision-making rule is used to determine the risk level of the business transaction corresponding to the to-be-predicted data.

According to an aspect of the embodiments of the present application, a prediction apparatus based on a network model is provided, including:

A parameter obtaining module, used for obtaining the original model parameters of the target network model, wherein the input of the target network model is K variables, and the K is an integer greater than 1;

A sub-network model generation module, configured to generate T sub-network models according to the original model parameters, wherein each sub-network model in the T sub-network models is used to input a variable group, and the variable groups input by different sub-network models Different from each other, T variable groups include the K variables, and each variable group includes at least one variable;

A data acquisition module, configured to acquire data to be predicted, wherein the data to be predicted includes data features corresponding to each of the K variables;

a data division module, configured to divide the to-be-predicted data into data feature groups corresponding to the T variable groups according to the variable group corresponding to each sub-network model;

The result generating module is used for inputting corresponding data feature groups to each sub-network model respectively, so as to generate a model prediction result of the target network model.

In some embodiments of the present application, based on the above technical solutions, the sub-network model generation module includes:

a parameter parsing unit, configured to parse the original model parameters according to the implementation language of the original model parameters, to obtain a plurality of calculation units, the calculation units are used to calculate the output result based on the corresponding variable group;

The combining unit is configured to combine the computing units with mutual dependencies into a sub-network model according to the dependencies between the input parameters and output results of the plurality of computing units to obtain T sub-network models.

According to an aspect of the embodiments of the present application, there is provided an electronic device, the electronic device includes: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to execute by executing the executable instructions Such as the prediction method based on the network model in the above technical solutions.

According to an aspect of the embodiments of the present application, there is provided a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, implements the network model-based prediction method in the above technical solution.

In the embodiment of the present application, by changing the execution order of the model itself, the original execution logic of the model is split and reorganized during the compilation process, so that the parallelism of the calculation in the model is increased, and the running speed of the model is improved, thereby Improve the real-time performance of computing, so that it can be applied in the application field of real-time computing.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not limiting of the present application.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and together with the description serve to explain the principles of the application. Obviously, the drawings in the following description are only some embodiments of the present application, and for those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort. In the attached image:

FIG. 1 schematically shows a schematic diagram of an exemplary system architecture of the technical solution of the present application in an application scenario;

2 is a schematic flowchart of a prediction process in an embodiment of the present application;

3 is a flowchart of a prediction method based on a network model provided by an embodiment of the present application;

FIG. 4 schematically shows a block diagram of the composition of a prediction apparatus based on a network model in an embodiment of the present application;

FIG. 5 shows a schematic structural diagram of a computer system suitable for implementing the electronic device according to the embodiment of the present application.

Detailed ways

Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments, however, can be embodied in various forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided so that this application will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art.

Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided in order to give a thorough understanding of the embodiments of the present application. However, those skilled in the art will appreciate that the technical solutions of the present application may be practiced without one or more of the specific details, or other methods, components, devices, steps, etc. may be employed. In other instances, well-known methods, devices, implementations, or operations have not been shown or described in detail to avoid obscuring aspects of the present application.

The block diagrams shown in the figures are merely functional entities and do not necessarily necessarily correspond to physically separate entities. That is, these functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor devices and/or microcontroller devices entity.

The flowcharts shown in the figures are only exemplary illustrations and do not necessarily include all contents and operations/steps, nor do they have to be performed in the order described. For example, some operations/steps can be decomposed, and some operations/steps can be combined or partially combined, so the actual execution order may be changed according to the actual situation.

It should be understood that the solution of the present application can be applied to the field of machine learning, and is specifically used in a scenario where a machine model is used for real-time computing. The solution of the present application can dynamically parse the output of various model packages, normalize it into a unified model execution object, and then dynamically generate a DAG execution graph according to the input parameters that each subtree depends on, and then base the DAG execution graph on the dependencies. Incorporated into the overall DAG flow chart of the application. In practical applications, relying on the topological sorting algorithm, the calculation of the model is integrated into the application execution process, and distributed computing is performed at the same time, thereby reducing the time-consuming of model introduction to the system.

FIG. 1 schematically shows a schematic diagram of an exemplary system architecture of the technical solution of the present application in an application scenario. As shown in Figure 1, the application scenario includes a server and multiple terminal devices. A real-time service is deployed in the server, and the real-time service performs related operations through the machine model configured according to the solution of the present application. The terminal can access the real-time service on the server through the network to request the operation. The service calls the machine model to perform real-time operation on the data sent by the terminal and returns the operation result.

The server shown in FIG. 1 may specifically be a server, a server cluster including multiple servers, or a cloud server. In one embodiment, the system architecture may not use any terminal device, and the operations performed by the terminal device are run by a background service on a server, or run by a dedicated server.

It can be understood that the scenario shown in FIG. 1 is only an example of the scenario applied by the solution of the present application, and the actual application scenario may adopt other suitable network structures, such as adding a proxy server and a multi-level network, etc. No restrictions apply.

FIG. 2 is a schematic flowchart of a prediction process in an embodiment of the present application. As shown in Figure 2, models written in PMML or SPARK are dynamically compiled into model objects and normalized into a unified model execution object, that is, a subtree. Then, the DAG execution graph is constructed according to the variable dependencies of each subtree, so as to determine the execution process of the subtree, and then the model result is obtained by fitting the model. In practical applications, the variables required by the model are proposed from external information according to the input data of the model and calculated, so as to obtain the results of the model and make decisions based on the results.

The technical solutions provided by the present application will be described in detail below with reference to specific embodiments. For the convenience of introduction, please refer to FIG. 3 , which is a flowchart of a prediction method based on a network model provided by an embodiment of the present application. This method can be applied to the execution of the above-mentioned server, and the server can be regarded as a kind of computer equipment. In the embodiment of the present application, the computer equipment is used as the execution subject to introduce the prediction based on the network model. The prediction method based on the network model can be It includes the following steps S310 to S350:

Step S310, obtaining the original model parameters of the target network model, wherein the input of the target network model is K variables, and the K is an integer greater than 1;

The original model parameters include the structure of the target network model itself, the calculation process, and the model variables. The original model parameters are usually in the form of files or file packages, wherein the original model parameters are written in various programming languages and exported into corresponding files, such as PMML, PYTHON and other languages. The target network model itself is a trained machine learning model. The target network model can be a network model used to achieve any target, and the structure adopted by the model can also adopt various common network models. For example, the targetable network model can be a model for risk assessment of bank transactions, and the input data are various types of data of each credit card transaction, such as the account, amount, remarks, transaction history of both parties, etc. , and the output result is the prediction result of whether the transaction is illegal or not.

Step S320, generating T sub-network models according to the original model parameters, wherein each sub-network model in the T sub-network models is used to input a variable group, and the variable groups input by different sub-network models are different from each other, and T variable groups include the K variables, and each variable group includes at least one variable;

Specifically, the sub-network model usually adopts a tree structure, and the T sub-network models can be formed in the form of a directed acyclic graph. The overall effect of the T sub-network models is the same as that of the target network model. The original model parameters are transformed into one or more tree-structured networks, and each network corresponds to one or more computation processes in the original model parameters. Each of the T sub-network models is used to input a variable group, and the variable groups input by different sub-network models are different from each other, and the T variable groups include the K variables. The intersection of the variable groups of all sub-network models is the original input variables of the target network model. The variable groups of each sub-network model are different, but there are some variable conflicts among the variable components of different sub-network models. The T sub-network models can be executed in parallel.

In an embodiment of the present application, the process of generating T sub-network models according to the original model parameters includes the following steps: parsing the original model parameters according to the implementation language of the original model parameters to obtain multiple a calculation unit, which is used to calculate the output result based on the corresponding variable group; according to the dependencies between the input parameters and the output results of the plurality of calculation units, the calculation units with mutual dependencies are combined into a sub-network model to obtain T sub-network models.

In an embodiment of the present application, the process of generating T sub-network models according to the original model parameters includes the following steps: parsing the original model parameters according to the implementation language of the original model parameters to obtain the first A calculation unit, a second calculation unit and a third calculation unit, wherein the input parameter of the third calculation unit is the output result of the first calculation unit and the second calculation unit; based on the first calculation unit, The second computing unit and the third computing unit generate sub-network models to obtain the T sub-network models.

Step S330, obtaining data to be predicted, wherein the data to be predicted includes data features corresponding to each of the K variables;

The data to be predicted is the actual data generated in the actual system. During the acquisition process, relevant data can be collected from the actual system according to the attributes and classification of the K variables. According to the actual data collected, feature extraction can be performed on the data to obtain the data features corresponding to each variable. It can be understood that the data to be predicted is usually data within a predetermined time period, for example, business data within three months or business data within three months.

In an embodiment of the present application, the process of acquiring the data to be predicted includes the following steps: according to the K variables, determine a data query statement corresponding to each variable and a corresponding data source; according to the data query statement The business data features corresponding to the K variable pairs are obtained from the data source to obtain the data to be predicted.

Step S340, according to the variable group corresponding to each sub-network model, divide the to-be-predicted data into data feature groups corresponding to the T variable groups;

The input variable group of each sub-network model includes a part of the K variables. Therefore, according to the corresponding relationship, a data feature group corresponding to the variable group can be generated according to the variables to be predicted according to the variables included in the variable group.

Step S350: Input corresponding data feature groups to each of the sub-network models to generate a model prediction result of the target network model.

The sub-network model operates on the input data feature groups respectively. Since the input parameters between the sub-network models all come from K variables, there is no interdependence between the operation processes of the sub-network models, so they can be executed in parallel. After all the sub-network models are calculated, the output results of each sub-network model can be summarized, and the regression operation of the loss function can be further performed to obtain the model prediction result of the target network model. In practical applications, based on the model prediction results, the business operations corresponding to the data to be predicted can be further processed. For example, if it is determined that the transaction is suspected of being illegal, the transaction can be frozen for further confirmation.

In an embodiment of the present application, the process of inputting corresponding data feature groups to each sub-network model to generate a model prediction result of the target network model includes the following steps: The model inputs corresponding data feature groups to obtain T output results; according to the T output results, model fitting is performed through the loss function of the target network model to obtain model prediction results of the target network model.

In an embodiment of the present application, after the corresponding data feature groups are respectively input to the each sub-network model to generate the model prediction result of the target network model, the method further includes the following steps: Obtain independent data features from the data to be predicted, and the independent data features do not correspond to variables in the K variables; according to the independent data features and the model prediction result, determine the decision rule corresponding to the data to be predicted , the decision rule is used to determine the risk level of the business transaction corresponding to the data to be predicted.

In the embodiment of the present application, by changing the execution order of the model itself, the original execution logic of the model is split and reorganized during the compilation process, so that the parallelism of the calculation in the model is increased, and the running speed of the model is improved, thereby Improve the real-time performance of computing, so that it can be applied in the application field of real-time computing.

It should be noted that although the various steps of the methods of the present application are depicted in the figures in a particular order, this does not require or imply that the steps must be performed in that particular order, or that all illustrated steps must be performed to achieve the desired the result of. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step for execution, and/or one step may be decomposed into multiple steps for execution, and the like.

The following describes the implementation of the apparatus of the present application, which can be used to execute the prediction method based on the network model in the foregoing embodiments of the present application. FIG. 4 schematically shows a block diagram of the composition of the prediction apparatus based on the network model in the embodiment of the present application. As shown in FIG. 4 , the prediction apparatus 400 based on the network model may mainly include:

A parameter obtaining module 410, configured to obtain the original model parameters of the target network model, wherein the input of the target network model is K variables, and the K is an integer greater than 1;

The sub-network model generation module 420 is configured to generate T sub-network models according to the original model parameters, wherein each sub-network model in the T sub-network models is used to input a variable group, and the variables input by different sub-network models The groups are different from each other, the T variable groups include the K variables, and each variable group includes at least one variable;

A data acquisition module 430, configured to acquire data to be predicted, wherein the data to be predicted includes data features corresponding to each of the K variables;

A data division module 440, configured to divide the to-be-predicted data into data feature groups corresponding to the T variable groups according to the variable group corresponding to each sub-network model;

The result generating module 450 is configured to respectively input corresponding data feature groups to each of the sub-network models, so as to generate a model prediction result of the target network model.

In some embodiments of the present application, based on the above technical solutions, the sub-network model generation module includes:

a parameter parsing unit, configured to parse the original model parameters according to the implementation language of the original model parameters, to obtain a plurality of calculation units, the calculation units are used to calculate the output result based on the corresponding variable group;

The combining unit is configured to combine the computing units with mutual dependencies into a sub-network model according to the dependencies between the input parameters and output results of the plurality of computing units to obtain T sub-network models.

In some embodiments of the present application, based on the above technical solutions, the sub-network model generation module includes:

According to the implementation language of the original model parameters, the original model parameters are parsed to obtain a first calculation unit, a second calculation unit and a third calculation unit, wherein the input parameter of the third calculation unit is the first calculation unit. an output result of a computing unit and the second computing unit;

Based on the first computing unit, the second computing unit, and the third computing unit, a sub-network model is generated, and the T sub-network models are obtained.

In some embodiments of the present application, based on the above technical solutions, the data acquisition module includes:

a data source determination unit, configured to determine the data query statement corresponding to each variable and the corresponding data source according to the K variables;

A feature acquisition unit, configured to acquire business data features corresponding to the K variable pairs from the data source according to the data query statement to obtain the data to be predicted.

In some embodiments of the present application, based on the above technical solutions, the result generation module includes:

an eigenvalue input unit for inputting corresponding data feature groups to each of the sub-network models respectively, to obtain T output results;

A model fitting unit, configured to perform model fitting through the loss function of the target network model according to the T output results, to obtain a model prediction result of the target network model.

In some embodiments of the present application, based on the above technical solutions, the network model-based prediction apparatus further includes:

an independent feature acquisition module, used for acquiring independent data features from the data to be predicted, the independent data features not corresponding to the variables in the K variables;

A decision rule determination module, configured to determine a decision rule corresponding to the to-be-predicted data according to the independent data feature and the model prediction result, where the decision-making rule is used to determine the risk level of the business transaction corresponding to the to-be-predicted data .

It should be noted that the apparatuses provided in the above embodiments and the methods provided by the above embodiments belong to the same concept, and the specific manners in which each module performs operations have been described in detail in the method embodiments, which will not be repeated here.

FIG. 5 shows a schematic structural diagram of a computer system suitable for implementing the electronic device according to the embodiment of the present application.

It should be noted that the computer system 500 of the electronic device shown in FIG. 5 is only an example, and should not impose any limitations on the functions and scope of use of the embodiments of the present application.

As shown in FIG. 5 , the computer system 500 includes a central processing unit (Central Processing Unit, CPU) 501, which can be loaded into a random device according to a program stored in a read-only memory (Read-Only Memory, ROM) 502 or from a storage part 508 Various appropriate operations and processes are executed by accessing a program in a memory (Random Access Memory, RAM) 503 . In the RAM 503, various programs and data necessary for system operation are also stored. The CPU 501 , the ROM 502 , and the RAM 503 are connected to each other through a bus 504 . An Input/Output (I/O) interface 505 is also connected to the bus 504 .

The following components are connected to the I/O interface 505: an input section 506 including a keyboard, a mouse, etc.; an output section 507 including a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc. ; a storage part 508 including a hard disk and the like; and a communication part 509 including a network interface card such as a LAN (Local Area Network) card, a modem, and the like. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I/O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is mounted on the drive 510 as needed so that a computer program read therefrom is installed into the storage section 508 as needed.

In particular, according to the embodiments of the present application, the processes described in the flowcharts of the respective methods may be implemented as computer software programs. For example, embodiments of the present application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the method illustrated in the flowchart. In such an embodiment, the computer program may be downloaded and installed from the network via the communication portion 509 and/or installed from the removable medium 511 . When the computer program is executed by the central processing unit (CPU) 501, various functions defined in the system of the present application are executed.

It should be noted that the computer-readable medium shown in the embodiments of the present application may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the above two. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or a combination of any of the above. More specific examples of computer readable storage media may include, but are not limited to, electrical connections with one or more wires, portable computer disks, hard disks, random access memory (RAM), read only memory (ROM), erasable Erasable Programmable Read Only Memory (EPROM), flash memory, optical fiber, portable Compact Disc Read-Only Memory (CD-ROM), optical storage device, magnetic storage device, or any suitable of the above The combination. In this application, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, however, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code therein. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. A computer-readable signal medium can also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device . Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination of the foregoing.

The flowchart and block diagrams in the Figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code that contains one or more logical functions for implementing the specified functions executable instructions. It should also be noted that, in some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It is also noted that each block of the block diagrams or flowchart illustrations, and combinations of blocks in the block diagrams or flowchart illustrations, can be implemented in special purpose hardware-based systems that perform the specified functions or operations, or can be implemented using A combination of dedicated hardware and computer instructions is implemented.

It should be noted that although several modules or units of the apparatus for action performance are mentioned in the above detailed description, this division is not mandatory. Indeed, according to embodiments of the present application, the features and functions of two or more modules or units described above may be embodied in one module or unit. Conversely, the features and functions of one module or unit described above may be further divided into multiple modules or units to be embodied.

From the description of the above embodiments, those skilled in the art can easily understand that the exemplary embodiments described herein may be implemented by software, or may be implemented by software combined with necessary hardware. Therefore, the technical solutions according to the embodiments of the present application may be embodied in the form of software products, and the software products may be stored in a non-volatile storage medium (which may be CD-ROM, U disk, mobile hard disk, etc.) or on the network , which includes several instructions to cause a computing device (which may be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the embodiments of the present application.

Other embodiments of the present application will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses or adaptations of this application that follow the general principles of this application and include common knowledge or conventional techniques in the technical field not disclosed in this application .

It is to be understood that the present application is not limited to the precise structures described above and illustrated in the accompanying drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.

Claims (10)

1. a prediction method based on a network model, is characterized in that, comprises: Obtain the original model parameters of the target network model, wherein the input of the target network model is K variables, and the K is an integer greater than 1; T sub-network models are generated according to the original model parameters, wherein each sub-network model in the T sub-network models is used to input a variable group, and the variable groups input by different sub-network models are different from each other, and the T variable groups Including the K variables, each variable group includes at least one variable; Obtaining data to be predicted, wherein the data to be predicted includes data features corresponding to each of the K variables; According to the variable group corresponding to each sub-network model, the to-be-predicted data is divided into data feature groups corresponding to the T variable groups; Corresponding data feature groups are respectively input to each of the sub-network models to generate a model prediction result of the target network model. 2. The method according to claim 1, wherein the generating T sub-network models according to the original model parameters comprises: According to the implementation language of the original model parameters, the original model parameters are analyzed to obtain a plurality of calculation units, and the calculation units are used to calculate the output result based on the corresponding variable group; According to the dependencies between the input parameters and the output results of the plurality of computing units, the computing units with mutual dependencies are combined into a sub-network model to obtain T sub-network models. 3. The method according to claim 1, wherein the generating T sub-network models according to the original model parameters comprises: According to the implementation language of the original model parameters, the original model parameters are parsed to obtain a first calculation unit, a second calculation unit and a third calculation unit, wherein the input parameter of the third calculation unit is the first calculation unit. an output result of a computing unit and the second computing unit; Based on the first computing unit, the second computing unit, and the third computing unit, a sub-network model is generated, and the T sub-network models are obtained. 4. The method according to claim 1, wherein the obtaining the data to be predicted comprises: According to the K variables, determine the data query statement corresponding to each variable and the corresponding data source; The business data characteristics corresponding to the K variable pairs are obtained from the data source according to the data query statement, and the data to be predicted is obtained. 5. The method according to claim 1, wherein the inputting a corresponding data feature group to the each sub-network model respectively, to generate a model prediction result of the target network model, comprising: Input corresponding data feature groups to each of the sub-network models to obtain T output results; According to the T output results, model fitting is performed through the loss function of the target network model to obtain a model prediction result of the target network model. 6. The method according to any one of claims 1 to 5, wherein after inputting corresponding data feature groups to each sub-network model respectively, to generate a model prediction result of the target network model , the method also includes: Obtain independent data features from the data to be predicted, and the independent data features do not correspond to variables in the K variables; According to the independent data feature and the model prediction result, a decision rule corresponding to the to-be-predicted data is determined, and the decision-making rule is used to determine the risk level of the business transaction corresponding to the to-be-predicted data. 7. A prediction device based on a network model, characterized in that, comprising: A parameter obtaining module, used for obtaining the original model parameters of the target network model, wherein the input of the target network model is K variables, and the K is an integer greater than 1; A sub-network model generation module, configured to generate T sub-network models according to the original model parameters, wherein each sub-network model in the T sub-network models is used to input a variable group, and the variable groups input by different sub-network models Different from each other, T variable groups include the K variables, and each variable group includes at least one variable; A data acquisition module, configured to acquire data to be predicted, wherein the data to be predicted includes data features corresponding to each of the K variables; a data division module, configured to divide the to-be-predicted data into data feature groups corresponding to the T variable groups according to the variable group corresponding to each sub-network model; The result generating module is used for inputting corresponding data feature groups to each sub-network model respectively, so as to generate a model prediction result of the target network model. 8. The method according to claim 7, wherein the sub-network model generation module comprises: a parameter parsing unit, configured to parse the original model parameters according to the implementation language of the original model parameters, to obtain a plurality of calculation units, the calculation units are used to calculate the output result based on the corresponding variable group; The combining unit is configured to combine the computing units with mutual dependencies into a sub-network model according to the dependencies between the input parameters and output results of the plurality of computing units to obtain T sub-network models. 9. An electronic device, characterized in that, comprising: processor; a memory for storing executable instructions for the processor; Wherein, the processor is configured to execute the network model-based prediction method of any one of claims 1 to 6 by executing the executable instructions. 10 . A computer-readable medium having a computer program stored thereon, wherein the computer program implements the prediction method based on a network model according to any one of claims 1 to 6 when the computer program is executed by a processor. 11 .

Images