A model construction and operation service method and system under an artificial intelligence framework

Abstract

The present invention provides a model construction and operation service method and system under an artificial intelligence framework, wherein the method includes: obtaining a trained model, and registering it in a model operation service system according to relevant parameters of the model; uploading the training code of the model to the GIT library in the model operation service system; the model deployment module in the model operation service system obtains the model code, and deploys the model through the model code; grayscale verification is performed on the deployed model, and the successfully verified model is officially launched; real-time acquisition of model operation data, comparison of the operation data with preset data, and judgment of whether the model is experiencing a decline phenomenon; if so, iterative optimization of the model; if not, continuing to run the model. The present invention greatly simplifies the model deployment and online process, improves model deployment efficiency, reduces manual operations, and reduces model operation costs.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a model building and operation service method and system under an artificial intelligence framework. Background Art

[0002] With the advent of the global artificial intelligence era, artificial intelligence technologies such as big data, machine learning, and deep learning have been widely used in various industries. More and more models are being applied to business scenarios such as risk prevention and control and anti-fraud, and a set of model construction and operation systems have gradually been formed in this process. Modelers in various industries carry out model training and development work on the modeling platform according to actual business scenarios. After completing model training, the model is deployed online and put into production environment application through batch estimation or real-time estimation. At the same time, after the model is deployed online, the modeler can continuously count the operation status and effect indicators of the model through the corresponding tools, iteratively optimize the model after the model performance declines, retrain the model, and redeploy the trained model online after the model performance indicators meet the requirements to replace the original model.

[0003] The existing model construction and operation system has shortcomings, which are mainly reflected in the following points: 1. The model deployment process often requires a lot of manual intervention and has a low degree of automation. For models that have been iteratively optimized for multiple times, multiple deployment operations are required, which consumes a lot of human resources; 2. After the model is deployed online, the model information, model input and output lineage, model performance indicators, etc. have not formed a unified visual monitoring and management, and the offline operation cost is relatively high. Summary of the invention

[0004] To solve the above technical problems, the present invention provides a model construction and operation service method and system under an artificial intelligence framework, which reduces the model operation cost by performing unified visual monitoring and management of model information, model input and output lineage, model performance indicators, etc. after the model is deployed online. At the same time, a separate model deployment API service is used to realize the one-click deployment function of the model, realize automatic model deployment, and improve model deployment efficiency.

[0005] The first aspect of the present invention provides a model construction and operation service method under an artificial intelligence framework, comprising:

[0006] Obtain the trained model and register it in the model operation service system according to the relevant parameters of the model;

[0007] Upload the training code of the model to the GIT library in the model operation service system;

[0008] The model deployment module in the model operation service system obtains the model code and deploys the model through the model code;

[0009] The deployed model is subjected to grayscale verification, and the successfully verified model is officially launched.

[0010] Furthermore, after the successfully verified model is officially launched, it also includes:

[0011] The model operation data is acquired in real time, and the operation data is compared with the preset data to determine whether the model is experiencing a decline; if so, the model is iteratively optimized; if not, the model continues to run.

[0012] Furthermore, after the successfully verified model is officially launched, it also includes:

[0013] The model operation data is acquired in real time, and the operation data is compared with the preset data to determine whether the model is experiencing a decline; if so, the model is iteratively optimized; if not, the model continues to run.

[0014] Further, the iterative optimization of the model includes:

[0015] Re-establish the training model according to the actual scenario to obtain a new model;

[0016] After the decline model is offline, the new model is uploaded to the model operation service system.

[0017] Furthermore, the grayscale verification of the deployed model includes:

[0018] Verify model information, the relationship between model input and output, and model performance indicators.

[0019] The second aspect of the present invention provides a model construction and operation service system under an artificial intelligence framework, including:

[0020] A registration module is used to obtain the trained model and register it in the model operation service system according to the relevant parameters of the model;

[0021] A code uploading module, used to upload the training code of the model to the GIT library in the model operation service system;

[0022] A model deployment module, used for the model deployment module in the model operation service system to obtain the model code and deploy the model through the model code;

[0023] The model verification module is used to perform grayscale verification on the deployed model. The model that passes the verification is officially launched.

[0024] Furthermore, the model construction and operation service system under the artificial intelligence framework also includes:

[0025] The model iteration module is used to obtain model operation data in real time, compare the operation data with preset data, and determine whether the model is experiencing a decline; if so, iteratively optimize the model; if not, continue to run the model.

[0026] Furthermore, the model iteration module is also used for:

[0027] Re-establish the training model according to the actual scenario to obtain a new model;

[0028] After the decline model is offline, the new model is uploaded to the model operation service system.

[0029] Furthermore, the model deployment module is also used to:

[0030] Verify model information, the relationship between model input and output, and model performance indicators.

[0031] A third aspect of the present invention provides a model operation service system, comprising:

[0032] A model registration unit, through which model information that needs to be deployed online after model training is completed is registered;

[0033] A model deployment unit, through which the registered model is deployed;

[0034] Model Git repository unit, used to store model training code;

[0035] The model performance detection unit is used to check the performance indicator operation data of the online model, and take corresponding measures when the model is degraded or the model operation status is abnormal, so as to ensure the stable operation and optimal effect of the model.

[0036] Furthermore, the model operation service system further includes:

[0037] A model information query unit, through which parameter information of the online model is queried;

[0038] The model deployment API service unit is used to obtain the model code from the model Git library unit after receiving the model deployment request;

[0039] The model input and output unit is used to view the input and output data of the online model.

[0040] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0041] The present invention provides a model construction and operation service method and system under an artificial intelligence framework, wherein the method includes: obtaining a trained model, registering it in a model operation service system according to relevant parameters of the model; uploading the training code of the model to the GIT library in the model operation service system; the model deployment module in the model operation service system obtains the model code, and deploys the model through the model code; grayscale verification of the deployed model is performed, and the successfully verified model is officially launched; real-time acquisition of model operation data, comparison of the operation data with preset data, and judgment of whether the model has a recession phenomenon; if so, iterative optimization of the model; if not, the model continues to run. The present invention realizes the one-click deployment function of the model after the model training is completed, greatly simplifies the model deployment and online process, and improves the model deployment efficiency in the process of multiple iterative optimization of the model; unified visual monitoring and management, reduces manual operations, and reduces model operation costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] In order to more clearly illustrate the technical solution of the present invention, the drawings required for use in the implementation mode will be briefly introduced below. Obviously, the drawings described below are only some implementation modes of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0043] Figure 1 It is a flow chart of a model construction and operation service method under an artificial intelligence framework provided by a certain embodiment of the present invention;

[0044] Figure 2 is a flow chart of a model construction and operation service method under an artificial intelligence framework provided by another embodiment of the present invention;

[0045] Figure 3 It is a structural diagram of a model construction and operation service system under an artificial intelligence framework provided by a certain embodiment of the present invention;

[0046] Figure 4 It is a flow chart of a model construction and operation service method under an artificial intelligence framework provided by another embodiment of the present invention;

[0047] Figure 5 It is a device diagram of a model construction and operation service system under an artificial intelligence framework provided by a certain embodiment of the present invention;

[0048] Figure 6 is a device diagram of a model construction and operation service system under an artificial intelligence framework provided by another embodiment of the present invention;

[0049] Figure 7It is a structural diagram of an electronic device provided by a certain embodiment of the present invention. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0052] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0053] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0054] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0055] The first aspect.

[0056] See also Figure 1-2 An embodiment of the present invention provides a model construction and operation service method under an artificial intelligence framework, including:

[0057] S10: Obtain the trained model and register it in the model operation service system according to relevant parameters of the model.

[0058] S20. Upload the training code of the model to the GIT library in the model operation service system.

[0059] S30: The model deployment module in the model operation service system obtains the model code and deploys the model through the model code.

[0060] S40: Perform grayscale verification on the deployed model, and the successfully verified model will be officially launched.

[0061] In a specific implementation, the grayscale verification of the deployed model includes:

[0062] Verify model information, the relationship between model input and output, and model performance indicators.

[0063] In a specific implementation manner, the step S40 further includes:

[0064] S50, acquiring model operation data in real time, comparing the operation data with preset data, and determining whether the model is experiencing a decline; if so, iteratively optimizing the model; if not, continuing to operate the model.

[0065] In a specific implementation, the iterative optimization of the model includes:

[0066] Re-establish the training model according to the actual scenario to obtain a new model;

[0067] After the decline model is offline, the new model is uploaded to the model operation service system.

[0068] In a specific embodiment, see Figure 3 The present invention provides a model construction and operation service system under an artificial intelligence framework, including:

[0069] 1) Model registration: This module registers the model information that modelers need to deploy online after completing model training on the modeling platform. Only after registration is completed can the model information be displayed on the model construction and operation service system;

[0070] 2) Model deployment: This module enables one-click deployment of the model on the modeling platform by allowing modelers to click on the corresponding menu;

[0071] 3) Model input and output: This module is used to view the input and output data of the online model. Input refers to the table data used for model training, and output refers to the result data after the model completes the estimation. This module can be used to manage the relationship between model input and output.

[0072] 4) Model performance indicator monitoring: This module is used to view the performance indicator data of the online model. When the model effect declines or the model operation status is abnormal, it can be discovered in time and corresponding measures can be taken to ensure the stable operation and optimal effect of the model. The model performance indicators mainly include operation indicators, technical indicators and operation indicators;

[0073] 5) Model information query: This module is used to view and manage the basic information of the online model, including model name, modeler, modeling platform, model GIT library address, model scenario and other information;

[0074] 6) Model deployment API service: After receiving a model deployment request, this module will pull the corresponding code from the model GIT library to create a project on the modeling platform to complete the model deployment operation;

[0075] 7) Model GIT library: This module is used to save the model code uploaded to the model construction and operation service system to achieve unified management of model versions.

[0076] See also Figure 4 ,The basic process of this system includes:

[0077] 1) After the modeler has trained the model on the modeling platform, the system is used to register the trained model. The registration content includes basic model information, model performance indicators, model input and output data, etc. The registration process is completed by importing an Excel spreadsheet on the system interface;

[0078] 2) After the model registration is completed, upload the trained model code on the system interface, and the model code will be saved to the model GIT library;

[0079] 3) After uploading the model code to the GIT repository, find the model deployment menu button corresponding to this model on the system interface. After clicking it, the system model deployment API service will pull the corresponding model code from the model GIT repository to create a project on the modeling platform;

[0080] 4) After the model grayscale verification is passed, the model is officially launched. At this time, the system can perform unified visual monitoring and management of model information, model input and output relationships, model performance indicators, etc.;

[0081] 5) Compare the monitored model performance indicator data with the pre-set normal target value to determine whether the model is experiencing degradation. If the model is experiencing degradation, it is iteratively optimized. If the model is not experiencing degradation, the model is allowed to continue running. During the model optimization process, the model needs to be retrained on the modeling platform according to the actual scenario. After the model training is completed, the old model is taken offline first, and then the newly trained model code is uploaded to the system model GIT library, and the model is deployed online according to the previous process, and this cycle is repeated.

[0082] The beneficial effects of the present invention are:

[0083] 1. This model construction and operation service system can be used to implement the one-click deployment function of the model after the model training is completed, which greatly simplifies the model deployment and launch process and improves the model deployment efficiency during the process of multiple iterative optimization of the model;

[0084] 2. This model can be used to build an operation service system to achieve unified visual monitoring and management of model information, model input and output relationships, and model performance indicators, thereby reducing manual operations and lowering model operation costs.

[0085] This invention is mainly aimed at the field of artificial intelligence, and mainly includes the module functions of one-click model deployment, model information management, model input and output relationship management, model performance evaluation index monitoring, etc. of the model construction and operation service system under the artificial intelligence framework:

[0086] The main points of protection are:

[0087] 1. Use the one-click deployment function module of this "model construction and operation service system" to deploy the model online after model training is completed;

[0088] 2. Use the model information management, model input and output relationship management, model performance evaluation index monitoring and other module functions of this "model construction and operation service system" to apply to model operation management after the model is deployed and launched.

[0089] The second aspect.

[0090] See also Figure 5-6 The present invention provides a model construction and operation service system under an artificial intelligence framework, including:

[0091] The registration module 10 is used to obtain the trained model and register it in the model operation service system according to the relevant parameters of the model.

[0092] The code uploading module 20 is used to upload the training code of the model to the GIT library in the model operation service system.

[0093] The model deployment module 30 is used in the model operation service system to obtain the model code and deploy the model through the model code.

[0094] In a specific implementation, the model deployment module 30 is further used to:

[0095] Verify model information, the relationship between model input and output, and model performance indicators.

[0096] The model verification module 40 is used to perform grayscale verification on the deployed model, and the model that succeeds the verification is officially launched.

[0097] In a specific embodiment, it also includes:

[0098] The model iteration module 50 is used to obtain model operation data in real time, compare the operation data with preset data, and determine whether the model is experiencing a decline phenomenon; if so, iteratively optimize the model; if not, continue to run the model.

[0099] In a specific implementation, the model iteration module 50 is further configured to:

[0100] Re-establish the training model according to the actual scenario to obtain a new model;

[0101] After the decline model is offline, the new model is uploaded to the model operation service system.

[0102] The third aspect.

[0103] An embodiment of the present invention provides a model operation service system, including:

[0104] A model registration unit, through which model information that needs to be deployed online after model training is completed is registered;

[0105] A model deployment unit, through which the registered model is deployed;

[0106] Model Git repository unit, used to store model training code;

[0107] The model performance detection unit is used to check the performance indicator operation data of the online model, and take corresponding measures when the model is degraded or the model operation status is abnormal, so as to ensure the stable operation and optimal effect of the model.

[0108] In a specific embodiment, it also includes:

[0109] A model information query unit, through which parameter information of the online model is queried;

[0110] The model deployment API service unit is used to obtain the model code from the model Git library unit after receiving the model deployment request;

[0111] The model input and output unit is used to view the input and output data of the online model.

[0112] The fourth aspect.

[0113] The present invention provides an electronic device, the electronic device comprising:

[0114] processor, memory, and bus;

[0115] The bus is used to connect the processor and the memory;

[0116] The memory is used to store operation instructions;

[0117] The processor is used to call the operation instruction, and the executable instruction enables the processor to perform operations corresponding to the model construction and operation service method under an artificial intelligence framework as shown in the first aspect of the present application.

[0118] In an alternative embodiment, an electronic device is provided, such as Figure 7 As shown, Figure 7 The electronic device 5000 shown includes: a processor 5001 and a memory 5003. The processor 5001 and the memory 5003 are connected, such as through a bus 5002. Optionally, the electronic device 5000 may also include a transceiver 5004. It should be noted that in actual applications, the transceiver 5004 is not limited to one, and the structure of the electronic device 5000 does not constitute a limitation on the embodiments of the present application.

[0119] Processor 5001 may be a CPU, a general-purpose processor, a DSP, an ASIC, an FPGA or other programmable logic device, a transistor logic device, a hardware component or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 5001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0120] The bus 5002 may include a path to transmit information between the above components. The bus 5002 may be a PCI bus or an EISA bus, etc. The bus 5002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 7 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0121] The memory 5003 may be a ROM or other type of static storage device that can store static information and instructions, a RAM or other type of dynamic storage device that can store information and instructions, or an EEPROM, a CD-ROM or other optical disk storage, an optical disk storage (including a compressed optical disk, a laser disk, an optical disk, a digital versatile disk, a Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited to these.

[0122] The memory 5003 is used to store application code for executing the solution of the present application, and the execution is controlled by the processor 5001. The processor 5001 is used to execute the application code stored in the memory 5003 to implement the content shown in any of the above method embodiments.

[0123] Among them, electronic devices include but are not limited to: mobile terminals such as mobile phones, laptops, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc.

[0124] The fifth aspect.

[0125] The present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the model construction and operation service method under an artificial intelligence framework shown in the first aspect of the present application is implemented.

[0126] Another embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer-readable storage medium is run on a computer, the computer can execute the corresponding content in the aforementioned method embodiment.

Claims

1. A model construction and operation service method under an artificial intelligence framework, characterized in that: include: Obtain the trained model and register it in the model operation service system according to the relevant parameters of the model; Upload the training code of the model to the GIT library in the model operation service system; The model deployment module in the model operation service system obtains the model code and deploys the model through the model code; The deployed model is subjected to model grayscale verification, and the successfully verified model is officially launched; wherein, after the successfully verified model is officially launched, the following steps are further included: obtaining model operation data in real time, comparing the operation data with preset data, and determining whether the model is experiencing a decline; if so, iteratively optimizing the model, and the iterative optimization includes: re-establishing the training model according to the actual scenario to obtain a new model; after the decline model is offline for processing, uploading the new model to the model operation service system; if not, continuing to run the model; the model grayscale verification of the deployed model includes: verifying model information, the blood relationship between the model input and output, and model performance indicators.

2. A model construction and operation service system under an artificial intelligence framework, characterized in that: include: A registration module is used to obtain the trained model and register it in the model operation service system according to the relevant parameters of the model; A code uploading module, used to upload the training code of the model to the GIT library in the model operation service system; A model deployment module, used for the model deployment module in the model operation service system to obtain the model code and deploy the model through the model code; The model verification module is used to perform grayscale verification on the deployed model, and the successfully verified model is officially launched; wherein, after the successfully verified model is officially launched, it also includes: real-time acquisition of model operation data, comparing the operation data with preset data, and judging whether the model is experiencing a decline; if so, iterative optimization of the model, the iterative optimization includes: re-establishing the training model according to the actual scenario to obtain a new model; after the decline model is offline, uploading the new model to the model operation service system; if not, continuing to run the model; the grayscale verification of the deployed model includes: verifying model information, the blood relationship between the input and output of the model, and model performance indicators.

3. A model operation service system, characterized in that: include: A model registration unit, through which model information that needs to be deployed online after model training is completed is registered; A model deployment unit, through which the registered model is deployed; Model Git repository unit, used to store model training code; The model performance detection unit is used to check the performance indicator operation data of the online model, and take corresponding measures when the model is degraded or the model operation status is abnormal, so as to ensure the stable operation and optimal effect of the model.

4. A model operation service system as claimed in claim 3, characterized in that: Also includes: A model information query unit, through which parameter information of the online model is queried; The model deployment API service unit is used to obtain the model code from the model Git library unit after receiving the model deployment request; The model input and output unit is used to view the input and output data of the online model.

Images