Artificial intelligence modeling method and device

Abstract

The invention discloses an artificial intelligence modeling method and device, and relates to the technical field of artificial intelligence, and the method comprises the steps: constructing a visual modeling tool and a preset model library in the form of a plug-in based on a Jupyterlab framework; the visual modeling tool is used for connecting a plurality of function modules on a canvas according to an operation instruction of a user; according to the visual modeling tool and a preset model library, a development platform for artificial intelligence model modeling is built, and the development platform comprises a front end and a rear end; wherein the front end is used for constructing an artificial intelligence model assembly line graph and sending a network request to the rear end according to the artificial intelligence model assembly line graph and an operation instruction, the rear end is used for scheduling execution of an algorithm task corresponding to the artificial intelligence model assembly line graph and performing artificial intelligence model task management, and the front end and the rear end are both developed based on a Jupterlab plug-in technology. The artificial intelligence modeling efficiency can be improved, and the user experience can be improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence and machine learning technology, and in particular to an artificial intelligence modeling method and device. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the invention recited in the claims. No admission is made that the description herein is prior art by inclusion in this section.

[0003] Nowadays, with the continuous development of artificial intelligence technology, there are more and more application scenarios, and the demand for artificial intelligence in various industries is also increasing. Different actual business scenarios have higher and higher requirements for the maturity and accuracy of model algorithms in artificial intelligence.

[0004] Traditional machine learning modeling process: 1. Perform feature preprocessing and feature engineering operations; 2. Divide the data set into training sets and test sets; 3. Select a machine learning algorithm and its parameter training; 4. Perform model prediction and evaluation; 5. If the evaluation does not meet the requirements, change the parameters or reselect the model to continue training.

[0005] However, the current industry big data analysis and artificial intelligence applications face major bottlenecks and constraints. The industry lacks AI (Artificial Intelligence) technical talents, AI modeling technology has high barriers to entry, relies heavily on expert experience, has a long modeling cycle, is costly, and has low efficiency. In order to enable more ordinary industries and enterprises to complete big data intelligent analysis and accelerate the implementation of big data and AI applications, there is an urgent need for an efficient and easy-to-use automated AI modeling method or tool to reduce dependence on AI experts and improve AI modeling efficiency.

[0006] Automated AI modeling methods or tools. Specifically, during the AI ​​visual modeling process, users can more intuitively operate data set configuration, input and output parameter setting, model training, evaluation, prediction, log viewing, model saving, scheduled tasks and other operations and obtain more accurate evaluation and prediction.

[0007] There are some machine learning model visualization modeling methods in the prior art, for example, presetting machine learning model modeling components, setting component parameters, and establishing a directed acyclic graph of a machine learning model. Although the existing methods can achieve visual modeling, all functions need to be developed in advance, the development cost is high, and implementation is difficult. In addition, in the process of artificial intelligence model modeling, many algorithms, data, etc. need to be applied, and even interaction with other platforms is required. The existing methods cannot achieve this, the modeling capability is insufficient, the artificial intelligence modeling efficiency is low, and the user experience is poor. Summary of the invention

[0008] The embodiment of the present invention provides an artificial intelligence modeling method, which is used to reduce the difficulty of visual artificial intelligence modeling, improve the artificial intelligence modeling capability and artificial intelligence modeling efficiency, and enhance user experience. The method includes:

[0009] Based on the Jupyterlab framework, a visual modeling tool and a preset model library are constructed in the form of plug-ins; the visual modeling tool is used to connect multiple functional modules on the canvas according to the user's operation instructions; the functional modules are used to represent the components of the artificial intelligence model, and the preset model library includes multiple preset machine learning algorithm toolkits;

[0010] According to the visual modeling tools and preset model library, a development platform for artificial intelligence model modeling is built, and the development platform includes a front-end and a back-end; wherein, the front-end is used to receive the user's operation instructions, build the artificial intelligence model pipeline graph according to the operation instructions, and send network requests to the back-end according to the artificial intelligence model pipeline graph and the operation instructions; the back-end is used to receive network requests, and according to the network requests, schedule the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph, and perform artificial intelligence model task management. Both the front-end and the back-end are developed based on Jupyterlab plug-in technology.

[0011] In one embodiment, the functional modules include any one or any combination of the following:

[0012] Data source, data processing, feature engineering, classification algorithm, regression analysis, clustering model, evaluation testing.

[0013] In one embodiment, the front end includes any one or any combination of the following:

[0014] Preset model library display area, canvas and canvas function management area, function module parameter setting area, function module execution status real-time feedback area, artificial intelligence model calculation result preview area, artificial intelligence model calculation result display area, user customized code area; the function module execution status includes waiting, executing, error, and completed execution. The canvas is used to organize function modules to form an artificial intelligence model pipeline. The user customized code area is used to receive user customized code.

[0015] In one embodiment, the preset model library includes one or any combination of the following:

[0016] Data sources, data processing algorithms, feature engineering algorithms, classification algorithms, clustering models, regression analysis algorithms, and model evaluation algorithms;

[0017] The data processing algorithms include one or any combination of the following algorithms: removing null values, filling missing values, random shuffling, data deletion, data concatenation, data splitting, data merging, deleting duplicates, type conversion, and one-hot encoding.

[0018] Feature engineering algorithms include one or any combination of the following algorithms: principal component analysis, polynomial features, feature selection selectkbest, sliding window,

[0019] The classification algorithm includes one or any combination of the following algorithms: decision tree, logistic regression, random forest, support vector machine, gradient boosting decision tree,

[0020] The clustering model includes one of the following or any combination of algorithms: K-means, density clustering, fuzzy clustering, isolation forest,

[0021] The regression analysis algorithm includes one of the following or any combination of algorithms: linear regression, Lasso regression, ridge regression, elastic network regression Elastic Net,

[0022] The model evaluation algorithm includes one or any combination of the following algorithms: AUC evaluation, receiver operating characteristic curve, determination coefficient, mean square error.

[0023] In one embodiment, the development platform supports multiple users to share code and data under the Jupyterlab file system.

[0024] In one embodiment, the development platform supports compiling algorithms in a Notebook environment or a Python environment.

[0025] In one embodiment, the visual modeling tool is started in the Lancher component of Jupyterlab, and the visual modeling tool is rendered in the Shell component of Jupyterlab.

[0026] The embodiment of the present invention further provides an artificial intelligence modeling device, which is used to reduce the difficulty of visual artificial intelligence modeling, improve the artificial intelligence modeling capability and artificial intelligence modeling efficiency, and enhance user experience. The device includes:

[0027] A visual modeling tool and a preset model library construction module, which is used to build a visual modeling tool and a preset model library in the form of a plug-in based on the Jupyterlab framework; the visual modeling tool is used to connect multiple functional modules on the canvas according to the user's operation instructions; the functional modules are used to represent the components of the artificial intelligence model, and the preset model library includes multiple preset machine learning algorithm toolkits;

[0028] The development platform building module is used to build a development platform for artificial intelligence model modeling based on visual modeling tools and preset model libraries. The development platform includes a front-end and a back-end; wherein, the front-end is used to receive user operation instructions, build an artificial intelligence model pipeline graph according to the operation instructions, and send network requests to the back-end according to the artificial intelligence model pipeline graph and the operation instructions; the back-end is used to receive network requests, and according to the network requests, schedule the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph, and perform artificial intelligence model task management. Both the front-end and the back-end are developed based on Jupyterlab plug-in technology.

[0029] In one embodiment, the functional modules include any one or any combination of the following:

[0030] Data source, data processing, feature engineering, classification algorithm, regression analysis, clustering model, evaluation testing.

[0031] In one embodiment, the front end includes any one or any combination of the following:

[0032] Preset model library display area, canvas and canvas function management area, function module parameter setting area, function module execution status real-time feedback area, artificial intelligence model calculation result preview area, artificial intelligence model calculation result display area, user customized code area; the function module execution status includes waiting, executing, error, and completed execution. The canvas is used to organize function modules to form an artificial intelligence model pipeline. The user customized code area is used to receive user customized code.

[0033] In one embodiment, the preset model library includes one or any combination of the following:

[0034] Data sources, data processing algorithms, feature engineering algorithms, classification algorithms, clustering models, regression analysis algorithms, and model evaluation algorithms;

[0035] The data processing algorithms include one or any combination of the following algorithms: removing null values, filling missing values, random shuffling, data deletion, data concatenation, data splitting, data merging, deleting duplicates, type conversion, and one-hot encoding.

[0036] Feature engineering algorithms include one or any combination of the following algorithms: principal component analysis, polynomial features, feature selection selectkbest, sliding window,

[0037] The classification algorithm includes one or any combination of the following algorithms: decision tree, logistic regression, random forest, support vector machine, gradient boosting decision tree,

[0038] The clustering model includes one of the following or any combination of algorithms: K-means, density clustering, fuzzy clustering, isolation forest,

[0039] The regression analysis algorithm includes one of the following or any combination of algorithms: linear regression, Lasso regression, ridge regression, elastic network regression Elastic Net,

[0040] The model evaluation algorithm includes one or any combination of the following algorithms: AUC evaluation, receiver operating characteristic curve, determination coefficient, mean square error.

[0041] In one embodiment, the development platform supports multiple users to share code and data under the Jupyterlab file system.

[0042] In one embodiment, the development platform supports compiling algorithms in a Notebook environment or a Python environment.

[0043] In one embodiment, the visual modeling tool is started in the Lancher component of Jupyterlab, and the visual modeling tool is rendered in the Shell component of Jupyterlab.

[0044] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned artificial intelligence modeling method when executing the computer program.

[0045] An embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned artificial intelligence modeling method is implemented.

[0046] An embodiment of the present invention also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the above-mentioned artificial intelligence modeling method is implemented.

[0047] In an embodiment of the present invention, a visual modeling tool and a preset model library are constructed in the form of plug-ins based on the Jupyterlab framework; the visual modeling tool is used to connect multiple functional modules on the canvas according to the user's operating instructions; the functional modules are used to represent the components of the artificial intelligence model, and the preset model library includes multiple preset machine learning algorithm toolkits; based on the visual modeling tool and the preset model library, a development platform for artificial intelligence model modeling is built, and the development platform includes a front end and a back end; wherein, the front end: is used to receive the user's operating instructions, build an artificial intelligence model pipeline graph according to the operating instructions, and send a network request to the back end according to the artificial intelligence model pipeline graph and the operating instructions; the back end: is used to receive network requests, and according to the network requests, schedules the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph, and performs artificial intelligence model task management, and both the front end and the back end are developed based on Jupyterlab plug-in technology. In an embodiment of the present invention, a visual modeling tool and a preset model library are constructed in the form of a plug-in based on the Jupyterlab framework. According to the visual modeling tool and the preset model library, a development platform for artificial intelligence model modeling is built. Both the front end and the back end are developed based on the Jupyterlab plug-in technology to realize rapid modeling of artificial intelligence models. Since the pluggable structure of Jupyterlab is used for development, the development cost is low and the development difficulty is reduced. Moreover, based on the development platform, users can call other components and services of Jupyterlab at will, which greatly reduces the difficulty of user use, reduces the difficulty of visual artificial intelligence modeling, improves the ability and efficiency of artificial intelligence modeling, and improves the user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the prior art descriptions. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. In the drawings:

[0049] Figure 1 A schematic diagram of a flow chart of an artificial intelligence modeling method according to an embodiment of the present invention;

[0050] Figure 2 It is a specific embodiment of the artificial intelligence modeling method in the embodiment of the present invention;

[0051] Figure 3 It is a specific embodiment of the artificial intelligence modeling method in the embodiment of the present invention;

[0052] Figure 4 It is a specific embodiment of the artificial intelligence modeling method in the embodiment of the present invention;

[0053] Figure 5 It is a specific embodiment of the artificial intelligence modeling method in the embodiment of the present invention;

[0054] Figure 6 A schematic diagram of an oilfield AI visualization modeling scenario in an embodiment of the present invention;

[0055] Figure 7 Schematic diagram of an artificial intelligence modeling device in an embodiment of the present invention. DETAILED DESCRIPTION

[0056] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0057] First, technical terms involved in the embodiments of the present invention are explained.

[0058] Jupyterlab: A popular, interactive web framework designed for data scientists, software developers, and researchers to interactively process data and code. Jupyterlab is a comprehensive upgrade of Jupyter Notebook. In fact, Jupyterlab is an all-round IDE (integrated development environment) that integrates Jupyter Notebook, text editor, terminal, and various personalized components (similar to VScode).

[0059] Jupyterlab plug-in technology: Jupyterlab plug-in is a cutting-edge development technology stack that transplants a wide range of front-end applications to the Jupyter data science development environment based on the Jupyterlab ecosystem and development specifications. A basic Jupyterlab plug-in development process includes at least the following steps:

[0060] First, initialize a new project in the project directory using npm (Node Package Manager, JavaScript package manager);

[0061] Then, write the plugin code, which may involve defining new classes, writing logic to handle user interactions, using React (a JavaScript library for building user interfaces) to create views, and so on;

[0062] You also need to use Jupyterlab's APIs to interact with the core parts of Jupyterlab, for example, to read the user's currently opened Notebook, or to open a new view when the user clicks a button;

[0063] After the plugin is completed, you need to use npm to build it and then publish it to the npm public library, or install it directly in the Jupyterlab instance;

[0064] During development, make sure you follow Jupyterlab's design guidelines and best practices to ensure that your plugin integrates seamlessly with the rest of Jupyterlab. You should also write tests to verify your code functionality and use continuous integration tools to automatically run these tests.

[0065] Kubeflow: A machine learning toolkit. Kubeflow is a technology stack running on K8S. This technology stack contains many components. The relationship between the components is relatively loose. We can use them together or use some of them separately.

[0066] React flow: React flow is a library for building node-based applications.

[0067] Argo: Argo is an open source project launched by Applatix, which provides container-native (each step in the workflow is implemented by a container) workflow for Kubernetes. It provides workflows that run in local containers for Kubernetes and implements each step in the workflow as a container. Argo allows users to start multi-step pipelines using a custom DSL (Domain Specific Language) similar to traditional YAML. The framework provides complex loops, conditions, and dependency management with DAG (directed acyclic graph), which helps improve the flexibility of deploying application stacks and the flexibility of configuration and dependencies.

[0068] It should be noted that the acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations.

[0069] The applicant found that there are some machine learning model visualization modeling methods in the prior art, such as presetting machine learning model modeling components, setting component parameters, and establishing a directed acyclic graph of a machine learning model. Although the existing methods can realize visual modeling, all functions need to be developed in advance, the development cost is high, and it is difficult to implement. In addition, in the process of artificial intelligence model modeling, many algorithms, data, etc. need to be applied, and even interaction with other platforms is required. The existing methods cannot achieve this, the modeling ability is insufficient, the artificial intelligence modeling efficiency is also low, and the user experience is poor. Based on this, the applicant proposed an artificial intelligence modeling method.

[0070] Figure 1 FIG. 1 is a flow chart of an artificial intelligence modeling method in an embodiment of the present invention. Figure 1 As shown, the method includes:

[0071] Step 101: Based on the Jupyterlab framework, a visual modeling tool and a preset model library are constructed in the form of plug-ins; the visual modeling tool is used to connect multiple functional modules on the canvas according to the user's operation instructions; the functional modules are used to represent the components of the artificial intelligence model, and the preset model library includes multiple preset machine learning algorithm toolkits;

[0072] Step 102. Build a development platform for artificial intelligence model modeling based on the visual modeling tool and the preset model library. The development platform includes a front-end and a back-end. The front-end is used to receive user operation instructions, build an artificial intelligence model pipeline graph according to the operation instructions, and send a network request to the back-end according to the artificial intelligence model pipeline graph and the operation instructions. The back-end is used to receive network requests, schedule the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph according to the network request, and perform artificial intelligence model task management. Both the front-end and the back-end are developed based on Jupyterlab plug-in technology.

[0073] from Figure 1 As can be seen from the shown process, in the embodiment of the present invention, based on the Jupyterlab framework, a visual modeling tool and a preset model library are constructed in the form of plug-ins. According to the visual modeling tool and the preset model library, a development platform for artificial intelligence model modeling is built. The front end and the back end are developed based on the Jupyterlab plug-in technology to realize rapid modeling of artificial intelligence models. Since the pluggable structure of Jupyterlab is used for development, the development cost is low and the development difficulty is reduced. Moreover, based on the development platform, users can call other components and services of Jupyterlab at will, which greatly reduces the difficulty of user use, reduces the difficulty of visual artificial intelligence modeling, improves the ability of visual artificial intelligence modeling and the efficiency of artificial intelligence modeling, and improves the user experience.

[0074] The artificial intelligence modeling method in the embodiment of the present invention is explained in detail below.

[0075] First, based on the actual workflow of artificial intelligence modeling and analysis and starting from user needs, a visual modeling tool and a preset model library are built in the form of plug-ins based on the Jupyterlab framework; the visual modeling tool is used to connect multiple functional modules on the canvas according to user operation instructions; the functional modules are used to represent the components of the artificial intelligence model, and the preset model library includes multiple preset machine learning algorithm toolkits.

[0076] Table 1 shows the key technologies involved in building visual modeling tools. Visual modeling tools can significantly reduce the technical barriers and costs in operating environment and software installation, analysis process design, code development, etc., and improve the efficiency of artificial intelligence modeling and analysis.

[0077] Table 1 Key technologies involved in building visual modeling tools

[0078] Serial number Key technologies Description 1 React flow Visual workflow drawing board 2 Sklearn Python Machine Learning Development Kit 3 Argo Cloud-native pipeline scheduling engine

[0079] During implementation, the visual modeling tool supports modeling analysts to use drag-and-drop to connect data reading, data processing, modeling, training, prediction and other functional modules to build a visual modeling analysis process. There is no need to write analysis program code in the whole process. Users can organize algorithm functional modules on the canvas by dragging and dropping to form an artificial intelligence model pipeline, which provides a more intuitive and interactive way to create and optimize artificial intelligence model pipelines. Compared with completing machine learning tasks only through coding and debugging, this method is more user-friendly.

[0080] Then, based on the visual modeling tools and preset model library, a development platform for artificial intelligence model modeling is built, and the development platform includes a front-end and a back-end; wherein, the front-end is used to receive user operation instructions, build an artificial intelligence model pipeline graph according to the operation instructions, and send network requests to the back-end according to the artificial intelligence model pipeline graph and operation instructions; the back-end is used to receive network requests, and according to the network requests, schedule the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph, and perform artificial intelligence model task management. Both the front-end and the back-end are developed based on Jupyterlab plug-in technology.

[0081] Figure 2 This is a specific embodiment of the artificial intelligence modeling method in the embodiment of the present invention, such as Figure 2 As shown in the figure, the visual modeling tool is integrated into Jupyterlab in a plug-in manner. The human-computer interaction interface occupies three window areas of Jupyterlab. The left area is the model tree, the middle area is the canvas, and the right side is the parameter setting area for the functional modules in the pipeline graph. The algorithm module in the model tree can be dragged to the canvas on the right. Click the node in the canvas and the node parameters will be displayed on the right.

[0082] Since the visual modeling tool is integrated into Jupyterlab, the development platform supports users to interact with other components and services of Jupyterlab after integration, making it easier for users to perform artificial intelligence modeling, reducing the difficulty of modeling and improving the user experience.

[0083] In one embodiment, the visual modeling tool is started in the Lancher component of Jupyterlab, and the visual modeling tool is rendered in the Shell component of Jupyterlab.

[0084] In one embodiment, the visual modeling tool interacts with Jupyterlab's file management system to enable the import and export of data, models, and algorithm modules.

[0085] The development platform for artificial intelligence model modeling in the embodiment of the present invention is developed based on the JupyterLab open source framework. The front-end and back-end development must comply with the Jupyterlab ecology and development specifications. In addition, an object-oriented design method can be adopted. The front-end is responsible for human-computer interaction and completes the construction of the artificial intelligence model pipeline graph. The back-end is responsible for scheduling the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph and performing artificial intelligence model task management.

[0086] The development platform for artificial intelligence model modeling in the embodiment of the present invention can also adopt a design method that combines object-oriented and service-oriented design methods to split the tasks that the backend is responsible for into tasks that are jointly executed by the backend and the underlying layer. For example, the front end is still responsible for human-computer interaction and completes the construction of the artificial intelligence model pipeline graph. The back end connects the upstream and downstream and is responsible for the algorithm task scheduling and artificial intelligence model task management corresponding to the artificial intelligence model pipeline graph, while the underlying layer is responsible for the actual algorithm execution corresponding to the artificial intelligence model pipeline graph.

[0087] The front end of the development platform uses the lumino framework, which provides a rich set of widgets, layouts, events, and data structure toolkits, enabling developers to build scalable, high-performance desktop Web applications. The back end of the development platform uses the Tornado Web framework encapsulated by Jupyter server, which encapsulates all drag-and-drop functions, as well as submit, run, stop and other functions in the back end framework. When the user performs an action on the front end interface, the front end sends an http network request, the back end service receives it, executes the corresponding code, and completes the back end response.

[0088] Different from the front-end and back-end interaction in traditional web development, the embodiment of the present invention is based on the Jupyterlab open source framework. The front-end and back-end interact in Jupyterlab. The front-end uses the lumino framework, and the back-end uses the secondary development interface provided by Jupyter server to complete the scheduling, running, saving, viewing and other functions of the machine learning pipeline. The execution of the node adopts the python subprocess method, which can execute python scripts and use ipykernal to execute Jupyter notebook scripts.

[0089] In the embodiment of the present invention, the development platform for artificial intelligence model modeling can use Jupyterlab's Kernal management system to schedule the execution engine; the execution Kernal of the visual modeling tool can be managed through Jupyterlab's Kernal management component.

[0090] Table 2 is a functional description of each object / service of the visual modeling function of the development platform for artificial intelligence model modeling in an embodiment of the present invention.

[0091] Table 2 Functional description of each object / service of visual modeling function

[0092]

[0093] In order to improve the artificial intelligence modeling capability, efficiency and user experience, in one embodiment, the functional modules may include any one or any combination of the following: data source, data processing, feature engineering, classification algorithm, regression analysis, clustering model, evaluation test. The functional modules are not limited in the embodiments of the present invention, and are only used as examples. Those skilled in the art can determine the functional modules according to actual use requirements.

[0094] In order to improve the artificial intelligence modeling capability, efficiency and user experience, in one embodiment, the preset model library may include one or any combination of the following:

[0095] Data sources, data processing algorithms, feature engineering algorithms, classification algorithms, clustering models, regression analysis algorithms, and model evaluation algorithms;

[0096] The data processing algorithms include one or any combination of the following algorithms: removing null values, filling missing values, random shuffling, data deletion, data concatenation, data splitting, data merging, deleting duplicates, type conversion, and one-hot encoding.

[0097] Feature engineering algorithms include one or any combination of the following algorithms: principal component analysis, polynomial features, feature selection selectkbest, sliding window,

[0098] The classification algorithm includes one or any combination of the following algorithms: decision tree, logistic regression, random forest, support vector machine, gradient boosting decision tree,

[0099] The clustering model includes one of the following or any combination of algorithms: K-means, density clustering, fuzzy clustering, isolation forest,

[0100] The regression analysis algorithm includes one of the following or any combination of algorithms: linear regression, Lasso regression, ridge regression, elastic network regression Elastic Net,

[0101] The model evaluation algorithm includes one or any combination of the following algorithms: AUC evaluation, receiver operating characteristic curve, determination coefficient, mean square error.

[0102] The algorithm modules in the preset model library are not limited in the embodiments of the present invention. This is only an example. Those skilled in the art can determine the algorithm modules in the preset model library according to actual usage requirements.

[0103] In order to improve the artificial intelligence modeling capability, efficiency and user experience, in one embodiment, the development platform front end in the embodiment of the present invention may include any one or any combination of the following:

[0104] Preset model library display area, canvas and canvas function management area, function module parameter setting area, function module execution status real-time feedback area, artificial intelligence model calculation result preview area, artificial intelligence model calculation result display area, user customized code area; the function module execution status includes waiting, executing, error, and completed execution. The canvas is used to organize function modules to form an artificial intelligence model pipeline. The user customized code area is used to receive user customized code.

[0105] For example, the real-time feedback area of ​​the function module execution status supports displaying the status of the function module at runtime, including waiting, executing, error, completed execution, etc. This provides users with real-time feedback to help them better understand and track the operation of their AI pipelines.

[0106] The AI ​​model calculation result preview area allows users to preview the results directly from the UI interface after executing the algorithm module, which provides users with an intuitive way to understand the results of their machine learning models without leaving the Jupyterlab environment.

[0107] The development platform in the embodiment of the present invention actually provides users with an open coding environment that can be flexibly expanded. It has both a preset model library and allows users to customize functional modules in the pipeline graphics. Users execute code, view results and modify code in real time. They drag the customized code script directly into the canvas, connect the upper and lower nodes, and click Submit to run. In this way, users are allowed to use preset modules to quickly establish artificial intelligence model pipeline graphics while retaining a high degree of customization.

[0108] Specifically, if the preset model library does not have the algorithm module that the user wants, the user can customize the code as follows: Figure 3 This is a specific embodiment of the artificial intelligence modeling method in the embodiment of the present invention, referring to Figure 3 :

[0109] ① The user writes an executable Python script based on the examples provided by the platform;

[0110] ② Expose the parameters to be tested by writing a configuration file;

[0111] ③ The user drags the mouse to put the aforementioned script into the canvas on the right;

[0112] ④ The user selects the upper and lower functional modules of the pipeline in the preset model library (for example, the prediction algorithm model C developed by the user needs to be connected to data A and data preprocessing module B. A and B can be directly selected from the preset model library and connected into an ABC pipeline);

[0113] ⑤Click Submit and wait for the execution results to be displayed.

[0114] Among them, steps ① and ② do not involve front-end and back-end interactions, and users can write scripts in the environment provided by the embodiment of the present invention. ③ The front-end algorithm detects the "drag and drop" action (converted into operation instructions), initiates a network request to read the file to the back-end in the Jupyterlab framework, and the request is transmitted to the back-end algorithm of "reading custom scripts" through the Jupyter server framework according to the previously encapsulated route. The back-end algorithm returns the user-specified parameters read to the front-end, and then the front-end algorithm uses the interface for developing the right sidebar provided by the Jupyterlab front-end framework to develop the sidebar table using JS language according to the designed style, draws the parameters returned by the back-end into the table, and displays it on the interface. ⑤ After the front-end algorithm detects the user's "submit" action (converted into operation instructions), the front-end algorithm sends the machine learning pipeline information in the canvas (such as the front and back position information of the algorithm in the pipeline, and the configuration parameters in the algorithm, etc.) to the back-end through the http network request under the Jupyterlab framework, and the Jupyter server receives it. Since the back-end of the development platform has been connected to Jupyterlab before, Jupyter The corresponding request received by the server can be obtained by the backend of the development platform. The backend algorithm matches the module name, then calls the subprocess to run, passes the parameters passed by the frontend to the subprocess and executes the algorithm script. After running, the result is returned to the frontend. The frontend algorithm develops its own style on the interface layout provided by Jupyterlab and displays the results.

[0115] In the prior art, it usually involves the use of version control tools (such as Git) to collaborate with others for development, which is of limited use. In order to improve the user experience and the efficiency of artificial intelligence modeling in the embodiments of the present invention, in one embodiment, the development platform supports multiple users to share code and data in the JupyterLab file system for collaborative development and data analysis, which allows multiple users to work in a unified environment, view and change code and algorithms in real time, and complete various tasks, thereby improving the user experience and the efficiency of artificial intelligence modeling.

[0116] In one embodiment, the development platform supports compiling algorithms in a Notebook environment or a Python environment. During implementation, users can customize the environment according to their needs and preferences, and can compile algorithms in the Notebook environment or Python environment provided by JupyterLab. Furthermore, users can also choose different Python versions and package versions and manage them using images, which provides greater flexibility and precision, allowing users to more easily adapt to different development needs and preferences and ensure the reproducibility of the code.

[0117] Below, the artificial intelligence modeling method is described from the user's perspective: First, drag the algorithm module in the preset model library or the user-defined executable script into the canvas to form various functional modules, and then organize the order of each functional module and connect the upstream and downstream to form an artificial intelligence model pipeline; click the functional module, enter the setting parameters in the functional module parameters, and then submit the back-end to run the pipeline and train the artificial intelligence model. After each functional module is run, hover the mouse over the canvas module to preview the calculation results, or retrain and view the results through the parameter adjustment of the visual interface. Repeated experiments will finally obtain the optimal algorithm parameters. After each functional module is run, the corresponding result file is generated in the front-end file view bar, which can be double-clicked to open for display.

[0118] Step 1: Visual modeling task creation: For example, refer to Figure 4 , Figure 4 This is a specific embodiment of the artificial intelligence modeling method in the embodiment of the present invention. In the main interface of Jupyterlab, click the pipeline edit plug-in icon to start a new artificial intelligence model project, create a canvas for dragging and dropping tasks, and click the branch button in the toolbar at the top of the canvas to create a task name and create an experimental branch. You can select a previously built branch to quickly build and compare.

[0119] Step 2: Interactive construction of visual modeling process: Get the built-in machine learning algorithm module in the left interface; enter the algorithm module into the canvas area by dragging the mouse, fill in the model operation parameters, and connect the upper and lower functional modules to form a pipeline;

[0120] Step 3: Set the function module parameters: Figure 2 , Figure 2 The right side of the center shows the node parameter settings;

[0121] Step 4: View the visual modeling results: Save the AI ​​model pipeline data drawn in the current canvas, import and view the parameters of the training process, and the results after modeling training. Figure 5 , Figure 5This is a specific embodiment of the artificial intelligence modeling method in the embodiment of the present invention, and shows a preview of the artificial intelligence model calculation results.

[0122] The Jupyterlab plug-in technology adopted in the embodiment of the present invention is a complete plug-in technology including the front-end and the back-end. The front-end and the back-end can be integrated into Jupyterlab in the form of a large plug-in as a whole, providing platform functions with a lightweight carrier to realize rapid modeling of artificial intelligence models. Theoretical analysis and experimental results are shown in Table 3. Table 3 is a comparative diagram of the results of artificial intelligence modeling using traditional modeling methods and artificial intelligence modeling methods of the embodiment of the present invention. The modeling data used is real data of oilfield drilling accidents. In the task of leak detection, the modeling and deployment using the artificial intelligence modeling method of the embodiment of the present invention took only half a day, and the detection accuracy was improved to 92%. The detection accuracy and detection rate both reached a good level, which can meet the real-time and accuracy of leak detection tasks. Experiments have proved that the artificial intelligence modeling method of the embodiment of the present invention has achieved unexpected technical effects in oilfield AI applications such as complex early warning of drilling accidents, pipeline leakage detection, safe operation video monitoring, and abnormal foaming early warning of desulfurization solution systems. Figure 6 This is a schematic diagram of the oil field AI visualization modeling scenario in an embodiment of the present invention. Using the artificial intelligence modeling method in an embodiment of the present invention to perform visualization modeling greatly improves modeling time and modeling accuracy, and enhances the competitiveness of the enterprise.

[0123] Table 3 Comparison of the results of artificial intelligence modeling using the traditional modeling method and the artificial intelligence modeling method of the embodiment of the present invention

[0124]

[0125] In summary, the embodiment of the present invention develops a drag-and-drop low-code artificial intelligence model modeling development platform in the form of plug-ins under the JupyterLab framework. Based on this development platform, users can use the drag-and-drop method to drag algorithm modules into the canvas and connect them to form an artificial intelligence model pipeline, and users can more conveniently call other algorithms and functions required for artificial intelligence modeling based on JupyterLab, which saves users more time, better handles complex machine learning tasks, and can clearly see the task architecture of the pipeline; at the same time, the development platform can be flexibly expanded, allowing users to add custom code and expand a wider range of new machine learning algorithms; and users can customize the environment according to their specific needs and preferences, and can use the Notebook environment or python environment provided by JupyterLab to compile algorithms; the development platform also has front-end and back-end interaction and multi-user collaboration methods that are different from traditional web development. In general, the embodiments of the present invention provide a JupyterLab plug-in-based visual interactive artificial intelligence modeling method for the energy industry, which can simplify and speed up the artificial intelligence model development process and meet users' high-efficiency requirements for artificial intelligence model development. This method allows users to build and adjust models in an intuitive way, as well as visual programming modeling, quickly respond to user needs, reduce the difficulty of visual artificial intelligence modeling, improve artificial intelligence modeling capabilities and efficiency, and enhance user experience.

[0126] The present invention also provides an artificial intelligence modeling device, as described in the following embodiments. Since the principle of solving the problem by the device is similar to that of the artificial intelligence modeling method, the implementation of the device can refer to the implementation of the artificial intelligence modeling method, and the repeated parts will not be repeated.

[0127] Figure 7 Schematic diagram of an artificial intelligence modeling device in an embodiment of the present invention. Figure 7 As shown, the device comprises:

[0128] The visual modeling tool and preset model library construction module 701 is used to construct the visual modeling tool and preset model library in the form of a plug-in based on the Jupyterlab framework; the visual modeling tool is used to connect multiple functional modules on the canvas according to the user's operation instructions; the functional modules are used to represent the components of the artificial intelligence model, and the preset model library includes multiple preset machine learning algorithm toolkits;

[0129] The development platform building module 702 is used to build a development platform for artificial intelligence model modeling based on visual modeling tools and preset model libraries. The development platform includes a front-end and a back-end; wherein the front-end is used to receive user operation instructions, build an artificial intelligence model pipeline graph according to the operation instructions, and send a network request to the back-end according to the artificial intelligence model pipeline graph and the operation instructions; the back-end is used to receive network requests, schedule the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph according to the network request, and perform artificial intelligence model task management. Both the front-end and the back-end are developed based on Jupyterlab plug-in technology.

[0130] In one embodiment, the functional modules include any one or any combination of the following:

[0131] Data source, data processing, feature engineering, classification algorithm, regression analysis, clustering model, evaluation testing.

[0132] In one embodiment, the front end includes any one or any combination of the following:

[0133] Preset model library display area, canvas and canvas function management area, function module parameter setting area, function module execution status real-time feedback area, artificial intelligence model calculation result preview area, artificial intelligence model calculation result display area, user customized code area; the function module execution status includes waiting, executing, error, and completed execution. The canvas is used to organize function modules to form an artificial intelligence model pipeline. The user customized code area is used to receive user customized code.

[0134] In one embodiment, the preset model library includes one or any combination of the following:

[0135] Data sources, data processing algorithms, feature engineering algorithms, classification algorithms, clustering models, regression analysis algorithms, and model evaluation algorithms;

[0136] The data processing algorithms include one or any combination of the following algorithms: removing null values, filling missing values, random shuffling, data deletion, data concatenation, data splitting, data merging, deleting duplicates, type conversion, and one-hot encoding.

[0137] Feature engineering algorithms include one or any combination of the following algorithms: principal component analysis, polynomial features, feature selection selectkbest, sliding window,

[0138] The classification algorithm includes one or any combination of the following algorithms: decision tree, logistic regression, random forest, support vector machine, gradient boosting decision tree,

[0139] The clustering model includes one of the following or any combination of algorithms: K-means, density clustering, fuzzy clustering, isolation forest,

[0140] The regression analysis algorithm includes one of the following or any combination of algorithms: linear regression, Lasso regression, ridge regression, elastic network regression Elastic Net,

[0141] The model evaluation algorithm includes one or any combination of the following algorithms: AUC evaluation, receiver operating characteristic curve, determination coefficient, mean square error.

[0142] In one embodiment, the development platform supports multiple users to share code and data under the JupyterLab file system.

[0143] In one embodiment, the development platform supports compiling algorithms in a Notebook environment or a Python environment.

[0144] In one embodiment, the visual modeling tool is started in the Lancher component of Jupyterlab, and the visual modeling tool is rendered in the Shell component of Jupyterlab.

[0145] An embodiment of the present invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned artificial intelligence modeling method when executing the computer program.

[0146] An embodiment of the present invention also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned artificial intelligence modeling method is implemented.

[0147] An embodiment of the present invention also provides a computer program product, which includes a computer program, and when the computer program is executed by a processor, the above-mentioned artificial intelligence modeling method is implemented.

[0148] In an embodiment of the present invention, a visual modeling tool and a preset model library are constructed in the form of plug-ins based on the Jupyterlab framework; the visual modeling tool is used to connect multiple functional modules on the canvas according to the user's operating instructions; the functional modules are used to represent the components of the artificial intelligence model, and the preset model library includes multiple preset machine learning algorithm toolkits; based on the visual modeling tool and the preset model library, a development platform for artificial intelligence model modeling is built, and the development platform includes a front end and a back end; wherein, the front end: is used to receive the user's operating instructions, build an artificial intelligence model pipeline graph according to the operating instructions, and send a network request to the back end according to the artificial intelligence model pipeline graph and the operating instructions; the back end: is used to receive network requests, and according to the network requests, schedules the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph, and performs artificial intelligence model task management, and both the front end and the back end are developed based on Jupyterlab plug-in technology. In an embodiment of the present invention, a visual modeling tool and a preset model library are constructed in the form of a plug-in based on the Jupyterlab framework. According to the visual modeling tool and the preset model library, a development platform for artificial intelligence model modeling is built. Both the front end and the back end are developed based on the Jupyterlab plug-in technology to realize rapid modeling of artificial intelligence models. Since the pluggable structure of Jupyterlab is used for development, the development cost is low and the development difficulty is reduced. Moreover, based on the development platform, users can call other components and services of Jupyterlab at will, which greatly reduces the difficulty of user use, reduces the difficulty of visual artificial intelligence modeling, improves the ability of visual artificial intelligence modeling and the efficiency of artificial intelligence modeling, and improves the user's sense of use experience.

[0149] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0150] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0151] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0152] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0153] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An artificial intelligence modeling method, characterized in that: include: Based on the Jupyterlab framework, build visual modeling tools and preset model libraries in the form of plug-ins; The visual modeling tool is used to connect multiple functional modules on the canvas according to the user's operation instructions; The functional module is used to represent the components of the artificial intelligence model, and the preset model library includes a plurality of preset machine learning algorithm toolkits; According to the visual modeling tools and preset model library, a development platform for artificial intelligence model modeling is built, and the development platform includes a front-end and a back-end; wherein, the front-end is used to receive the user's operation instructions, build the artificial intelligence model pipeline graph according to the operation instructions, and send network requests to the back-end according to the artificial intelligence model pipeline graph and the operation instructions; the back-end is used to receive network requests, and according to the network requests, schedule the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph, and perform artificial intelligence model task management. Both the front-end and the back-end are developed based on Jupyterlab plug-in technology.

2. The method according to claim 1, characterized in that The functional modules include any one or any combination of the following: Data source, data processing, feature engineering, classification algorithm, regression analysis, clustering model, evaluation testing.

3. The method according to claim 1, characterized in that The front end includes any one or any combination of the following: Preset model library display area, canvas and canvas function management area, function module parameter setting area, function module execution status real-time feedback area, artificial intelligence model calculation result preview area, artificial intelligence model calculation result display area, user customized code area; the function module execution status includes waiting, executing, error, and completed execution. The canvas is used to organize function modules to form an artificial intelligence model pipeline. The user customized code area is used to receive user customized code.

4. The method according to claim 1, characterized in that The preset model library includes one or any combination of the following: Data sources, data processing algorithms, feature engineering algorithms, classification algorithms, clustering models, regression analysis algorithms, and model evaluation algorithms; The data processing algorithms include one or any combination of the following algorithms: removing null values, filling missing values, random shuffling, data deletion, data concatenation, data splitting, data merging, deleting duplicates, type conversion, and one-hot encoding. Feature engineering algorithms include one or any combination of the following algorithms: principal component analysis, polynomial features, feature selection selectkbest, sliding window, The classification algorithm includes one or any combination of the following algorithms: decision tree, logistic regression, random forest, support vector machine, gradient boosting decision tree, The clustering model includes one of the following or any combination of algorithms: K-means, density clustering, fuzzy clustering, isolation forest, The regression analysis algorithm includes one of the following or any combination of algorithms: linear regression, Lasso regression, ridge regression, elastic network regression Elastic Net, The model evaluation algorithm includes one or any combination of the following algorithms: AUC evaluation, receiver operating characteristic curve, determination coefficient, mean square error.

5. The method according to claim 1, characterized in that The development platform supports multiple users to share code and data under the Jupyterlab file system.

6. The method according to claim 1, characterized in that The development platform supports compiling algorithms in a Notebook environment or a Python environment.

7. The method according to claim 1, characterized in that The visual modeling tool is started in the Lancher component of Jupyterlab, and the visual modeling tool is rendered in the Shell component of Jupyterlab.

8. An artificial intelligence modeling device, characterized in that: include: A visual modeling tool and a preset model library construction module, which is used to build a visual modeling tool and a preset model library in the form of a plug-in based on the Jupyterlab framework; the visual modeling tool is used to connect multiple functional modules on the canvas according to the user's operation instructions; The functional module is used to represent the components of the artificial intelligence model, and the preset model library includes a plurality of preset machine learning algorithm toolkits; The development platform building module is used to build a development platform for artificial intelligence model modeling based on visual modeling tools and preset model libraries. The development platform includes a front-end and a back-end; wherein, the front-end is used to receive user operation instructions, build an artificial intelligence model pipeline graph according to the operation instructions, and send network requests to the back-end according to the artificial intelligence model pipeline graph and the operation instructions; the back-end is used to receive network requests, and according to the network requests, schedule the execution of algorithm tasks corresponding to the artificial intelligence model pipeline graph, and perform artificial intelligence model task management. Both the front-end and the back-end are developed based on Jupyterlab plug-in technology.

9. The device according to claim 8, characterized in that The functional modules include any one or any combination of the following: Data source, data processing, feature engineering, classification algorithm, regression analysis, clustering model, evaluation testing.

10. The device according to claim 8, characterized in that The front end includes any one or any combination of the following: Preset model library display area, canvas and canvas function management area, function module parameter setting area, function module execution status real-time feedback area, artificial intelligence model calculation result preview area, artificial intelligence model calculation result display area, user customized code area; the function module execution status includes waiting, executing, error, and completed execution. The canvas is used to organize function modules to form an artificial intelligence model pipeline. The user customized code area is used to receive user customized code.

11. The device according to claim 8, characterized in that The preset model library includes one or any combination of the following: Data sources, data processing algorithms, feature engineering algorithms, classification algorithms, clustering models, regression analysis algorithms, and model evaluation algorithms; The data processing algorithms include one or any combination of the following algorithms: removing null values, filling missing values, random shuffling, data deletion, data concatenation, data splitting, data merging, deleting duplicates, type conversion, and one-hot encoding. Feature engineering algorithms include one or any combination of the following algorithms: principal component analysis, polynomial features, feature selection selectkbest, sliding window, The classification algorithm includes one or any combination of the following algorithms: decision tree, logistic regression, random forest, support vector machine, gradient boosting decision tree, The clustering model includes one of the following or any combination of algorithms: K-means, density clustering, fuzzy clustering, isolation forest, The regression analysis algorithm includes one of the following or any combination of algorithms: linear regression, Lasso regression, ridge regression, elastic network regression Elastic Net, The model evaluation algorithm includes one or any combination of the following algorithms: AUC evaluation, receiver operating characteristic curve, determination coefficient, mean square error.

12. The device according to claim 8, characterized in that The development platform supports multiple users to share code and data under the Jupyterlab file system.

13. The device according to claim 8, characterized in that The development platform supports compiling algorithms in a Notebook environment or a Python environment.

14. The device according to claim 8, characterized in that The visual modeling tool is started in the Lancher component of Jupyterlab, and the visual modeling tool is rendered in the Shell component of Jupyterlab.

15. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

16. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

17. A computer program product, characterized in that The computer program product comprises a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images