A method, device, equipment and medium for generating a machine learning model

By obtaining user application scenario description information and using the development workflow and dataset information in the knowledge base to generate machine learning models, the problems of high operation threshold and low training efficiency when building models are solved, and efficient model generation is achieved and choose.

CN118674075BActive Publication Date: 2024-11-22PUTIAN EASTERN COMM GRP CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411156397.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Priority Date
2024-07-30
Filing Date
2024-08-22
Publication Date
2024-11-22
Estimated Expiration
2044-08-22

AI Technical Summary

Technical Problem

When building machine learning models, business personnel have difficulty understanding the internal operating mechanism of the model and the basis for output results, resulting in high operational threshold and low training efficiency.

Method used

By obtaining the application scenario description information of the machine learning model that users expect to build, using the pre-built knowledge base to match the corresponding development workflow and dataset information and algorithm information, generate machine learning models, lower the user's operational threshold, and pass test evaluation Reports help users choose the model that best suits their needs.

Benefits of technology

It effectively reduces the operation threshold for users to generate machine learning models through machine learning platforms, ensures the training effect of the model, and improves the user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118674075B_ABST
    Figure CN118674075B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of machine learning technology, and discloses a method, device, equipment and medium for generating a machine learning model, the method comprising: obtaining first application scenario description information corresponding to a machine learning model that a user expects to build; matching a target application scenario corresponding to the first application scenario description information in a knowledge base, the knowledge base comprising: when developing machine learning models in different application scenarios, corresponding development workflows, data set information and algorithm information of at least one available algorithm; based on the development workflows, data set information and algorithm information of at least one available algorithm corresponding to the target application scenario, generating first machine learning models corresponding to each available algorithm, the present invention generates a machine learning model by matching the development workflow and specific information corresponding to the application scenario description information input by the user in the knowledge base, thereby reducing the problems of high operation threshold and low training efficiency when users develop models.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of machine learning technology, and in particular to a method, device, equipment and medium for generating a machine learning model. Background Art

[0002] At present, machine learning has been widely used in various fields, and various machine learning models and algorithms have significantly improved production efficiency in various fields. For banking marketing business, machine learning models can also significantly improve the marketing efficiency of banks. Banks can generate machine learning models that meet the requirements through some existing machine learning platforms to meet actual business needs.

[0003] However, when building machine learning models through various current machine learning platforms, due to the relatively complex construction principles of the models, business personnel often find it difficult to understand the model's internal operating mechanism and the basis of the output results, resulting in high operating thresholds and low training efficiency when business personnel generate machine learning models. Summary of the invention

[0004] In view of this, the present invention provides a method, device, equipment and medium for generating a machine learning model to solve the problems of high operation threshold and low training efficiency when a machine learning platform generates a machine learning model.

[0005] In a first aspect, the present invention provides a method for generating a machine learning model, which is applied to a machine learning platform, wherein the machine learning platform comprises: a pre-built knowledge base, and the method comprises:

[0006] Obtaining first application scenario description information corresponding to the machine learning model that the user expects to build;

[0007] Matching a target application scenario corresponding to the first application scenario description information in the knowledge base, wherein the knowledge base includes: corresponding development workflow, data set information, and algorithm information of at least one available algorithm when developing machine learning models in different application scenarios;

[0008] Based on the development workflow, data set information and algorithm information of at least one available algorithm corresponding to the target application scenario, a first machine learning model corresponding to each available algorithm is generated.

[0009] In this method, the description information of the application scenario corresponding to the machine learning model that the user expects to build is obtained, and then the machine learning model is generated by matching the corresponding development workflow, data set information and algorithm information in the knowledge base. Since the user only needs to describe the application scenario of the model to be built and does not need to understand the specific principles of model construction, the operational threshold for users to generate machine learning models through the machine learning platform can be greatly reduced. At the same time, generating the model according to the development workflow in the knowledge base can also ensure the training effect of the model.

[0010] In an optional implementation, after constructing the first machine learning model corresponding to each available algorithm, the method further includes:

[0011] Acquire first test data, and test each of the first machine learning models using the first test data;

[0012] Generate a corresponding evaluation report based on the test results corresponding to each first machine learning model, so that the user can determine the target learning model from at least one first machine learning model based on the evaluation report.

[0013] In this embodiment, the machine learning models corresponding to each algorithm are tested and evaluated to generate an evaluation report, which explains the characteristics of each machine learning model to the user, thereby facilitating the user to select the machine learning model that best meets actual needs.

[0014] In an optional implementation, the knowledge base further includes: description information corresponding to different application scenarios;

[0015] The matching of the target application scenario corresponding to the first application scenario description information in the knowledge base includes:

[0016] Converting the first application scenario description information and description information corresponding to different application scenarios in the knowledge base into a vector format through a preset language model;

[0017] The first application scenario description information in the vector format is compared with description information corresponding to different application scenarios in the knowledge base to determine a target application scenario.

[0018] In this embodiment, by converting the application scenario description information of the machine learning model that the user expects to build and the description information corresponding to different application scenarios in the knowledge base into vector format, matching is performed according to the vectors, which can more accurately match the target application scenario corresponding to the user's needs and ensure the effectiveness of the subsequently constructed machine learning model.

[0019] In an optional implementation, the generating a first machine learning model corresponding to each available algorithm based on the development workflow corresponding to the target application scenario, the data set information, and the algorithm information of at least one available algorithm includes:

[0020] Generate development code of a first machine learning model corresponding to each available algorithm according to the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario;

[0021] According to the development code of the first machine learning model corresponding to each available algorithm, an initial model corresponding to each available algorithm is constructed, and the initial model is trained and adjusted to obtain the first machine learning model corresponding to each available algorithm.

[0022] In this implementation mode, the corresponding model development code is generated through the development workflow, specific algorithm information, and data set information in the knowledge base, and then the corresponding initial model is constructed according to the development code and the relevant parameters are adjusted to finally generate a machine learning model. The model is automatically constructed and trained through the machine learning platform, which lowers the usage threshold and improves the user experience.

[0023] In an optional implementation, the algorithm information in the knowledge base includes: hyper-parameter knowledge corresponding to the algorithm;

[0024] The training and parameter adjustment of the initial model to obtain a first machine learning model corresponding to each available algorithm includes:

[0025] Call the training data and test data of each available algorithm corresponding to the initial model;

[0026] Training the initial model using the training data, and adjusting model parameters of the initial model;

[0027] The trained initial models are tested using the test data to obtain evaluation results corresponding to the trained initial models;

[0028] Obtain the hyperparameter knowledge corresponding to each available algorithm;

[0029] Based on the hyper-parameter knowledge corresponding to each available algorithm and the evaluation results corresponding to each trained initial model, hyper-parameter optimization is performed on each trained initial model to obtain a first machine learning model corresponding to each available algorithm.

[0030] In this implementation, the model parameters are adjusted through training data, and the evaluation results corresponding to the trained model are obtained according to the test data. The model hyperparameters are optimized in combination with specific hyperparameter knowledge to obtain the final machine learning model, which can ensure the training effect of the machine learning model.

[0031] In an optional implementation, the development workflow includes: work content corresponding to each work node in the development workflow;

[0032] The step of generating development codes of the first machine learning model corresponding to each available algorithm according to the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario includes:

[0033] For each available algorithm, based on the algorithm information and data set information of the available algorithm and the work content corresponding to each work node in the development workflow, generate development code corresponding to each work node;

[0034] The development codes corresponding to the various working nodes are spliced ​​to obtain the development code of the first machine learning model corresponding to the available algorithm.

[0035] In this implementation mode, the development code corresponding to each work node is generated by the work content, specific algorithm information and data set information corresponding to each work node in the development workflow, and the codes of each work node are spliced ​​to obtain the development code corresponding to each machine learning model, thereby ensuring the accuracy of the code.

[0036] In an optional implementation, generating a corresponding evaluation report according to the test results corresponding to each first machine learning model includes:

[0037] Determine the test results of each first machine learning model under each preset indicator, wherein the preset indicators include: accuracy, precision, and recall;

[0038] Generate a corresponding evaluation report based on the test results corresponding to each preset indicator.

[0039] This implementation method, by determining the test results under each indicator and generating a corresponding evaluation report, can enable users to further understand the characteristics of each machine learning model, so that users can make choices that meet actual needs.

[0040] In a second aspect, the present invention provides a device for generating a machine learning model, the device comprising:

[0041] An information acquisition module, used to acquire first application scenario description information corresponding to the machine learning model that the user expects to build;

[0042] A scenario matching module, configured to match a target application scenario corresponding to the first application scenario description information in the knowledge base, wherein the knowledge base includes: corresponding development workflows, data set information, and algorithm information of at least one available algorithm when developing machine learning models in different application scenarios;

[0043] A model development module is used to generate a first machine learning model corresponding to each available algorithm based on the development workflow, data set information and algorithm information of at least one available algorithm corresponding to the target application scenario.

[0044] In a third aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor being communicatively connected to each other, the memory storing computer instructions, and the processor executing the method for generating a machine learning model of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0045] In a fourth aspect, the present invention provides a computer-readable storage medium having computer instructions stored thereon, the computer instructions being used to enable a computer to execute the method for generating a machine learning model of the first aspect or any corresponding embodiment thereof. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 is a flow chart of a method for generating a machine learning model according to an embodiment of the present invention;

[0048] Figure 2 is a flowchart of another method for generating a machine learning model according to an embodiment of the present invention;

[0049] Figure 3 is an example diagram of a process for executing machine learning model construction through a machine learning platform according to an embodiment of the present invention;

[0050] Figure 4 This is an example diagram of a knowledge base construction process according to an embodiment of the present invention;

[0051] Figure 5 is an example diagram of a process for generating model code by a machine learning platform according to an embodiment of the present invention;

[0052] Figure 6 is an example diagram of a process of executing model optimization by a machine learning platform according to this embodiment;

[0053] Figure 7 is an example diagram of a closed-loop full process generated by a machine learning model according to an embodiment of the present invention;

[0054] Figure 8 is a structural block diagram of a device for generating a machine learning model according to an embodiment of the present invention;

[0055] Fig. 9 It is a schematic diagram of the hardware structure of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0057] At present, machine learning has been widely used in various fields, and various machine learning models and algorithms have significantly improved production efficiency in various fields. For banking marketing business, machine learning models can also significantly improve the marketing efficiency of banks. Banks can generate machine learning models that meet the requirements through some existing machine learning platforms to meet actual business needs.

[0058] However, when building machine learning models through various current machine learning platforms, due to the relatively complex construction principles of the models, business personnel often find it difficult to understand the model's internal operating mechanism and the basis of the output results, resulting in high operating thresholds and low training efficiency when business personnel generate machine learning models.

[0059] To this end, an embodiment of the present invention provides a method for generating a machine learning model, by obtaining description information of the application scenario corresponding to the machine learning model that the user expects to build, and then generating the machine learning model according to the corresponding development workflow and data set information and algorithm information matched in the knowledge base. Since the user only needs to describe the application scenario of the model to be built and does not need to understand the specific principles of model construction, the operation threshold for users to generate machine learning models through a machine learning platform can be greatly reduced. At the same time, by generating the model according to the development workflow in the knowledge base, the training effect of the model can also be guaranteed.

[0060] According to an embodiment of the present invention, an embodiment of a method for generating a machine learning model is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in an order different from that shown here.

[0061] In this embodiment, a method for generating a machine learning model is provided, which is applied to a machine learning platform. The machine learning platform includes: a pre-built knowledge base, Figure 1 is a flow chart of a method for generating a machine learning model according to an embodiment of the present invention. Figure 1 As shown, the process includes the following steps:

[0062] Step S101, obtaining first application scenario description information corresponding to the machine learning model that the user expects to build.

[0063] When users use the machine learning platform to build a machine learning model that meets business needs, they can enter the specific application scenario of the machine learning model they want to build and the specific functional description in the scenario, that is, what scenario the machine learning model they want to build is applied to and what functions can be achieved in the scenario. This information is the corresponding application scenario description information.

[0064] For example, a user can enter: I want to build a machine learning model that can predict the churn of bank customers and the reasons for the churn. Similar content can be understood as the application scenario description information corresponding to the machine learning model that the user expects to build.

[0065] Step S102, matching the target application scenario corresponding to the first application scenario description information in the knowledge base, wherein the knowledge base includes: when developing machine learning models in different application scenarios, the corresponding development workflow, data set information, and algorithm information of at least one available algorithm.

[0066] In the machine learning platform, the knowledge base of the user's corresponding business field is pre-organized. The knowledge base includes: descriptive information corresponding to various application scenarios that may require the construction of machine learning models in the corresponding business field, as well as the corresponding development workflow, data set information and algorithm information when developing the corresponding machine learning model in each application scenario.

[0067] Exemplarily, the development workflow may include multiple specific work nodes and the work content corresponding to each work node; the data set information includes description information of the data set used in developing the corresponding machine learning model, such as the meaning of various feature columns, label definitions and other specific data description information; at the same time, for each application scenario, when developing a model that meets the requirements of the application scenario, there may be multiple related algorithms, so the knowledge base will record the algorithm information of at least one algorithm corresponding to the application scenario, and the algorithm information may include: the advantages and disadvantages of the algorithm, applicable data requirements, hyperparameter range and other information, each algorithm can be used to build a corresponding machine learning model that meets the application scenario.

[0068] An application scenario that is closest to the description information of the first application scenario is matched in the knowledge base and is used as the target application scenario.

[0069] Step S103: Generate a first machine learning model corresponding to each available algorithm based on the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario.

[0070] After determining the development workflow, data set information, and algorithm information corresponding to the target application scenario, based on the specific content of each work node in the development workflow and the specific descriptions in the data set information and algorithm information, this information can be used as a thinking chain for model development, namely COT. The machine learning platform can perform specific operations during model development based on the specific description information in the thinking chain, such as generating model code, calling related data for training, parameter optimization, and other specific operations, so as to obtain machine learning models corresponding to each algorithm. These machine learning models corresponding to each algorithm can be applied to the first application scenario corresponding to the machine learning model that the user expects to build.

[0071] The method for generating a machine learning model provided in this embodiment generates the corresponding model development code through the development workflow and specific algorithm information and data set information in the knowledge base, and then constructs the corresponding initial model according to the development code and adjusts the relevant parameters to finally generate a machine learning model. The model is automatically constructed and trained through a machine learning platform, which lowers the usage threshold and improves the user experience.

[0072] According to an embodiment of the present invention, another method for generating a machine learning model is provided, which is applied to a machine learning platform. The machine learning platform includes: a pre-built knowledge base, Figure 2 is a flowchart of another method for generating a machine learning model according to an embodiment of the present invention. Figure 2 As shown, the process includes the following steps:

[0073] Step S201, obtaining the first application scenario description information corresponding to the machine learning model that the user expects to build. Figure 1 Step S101 of the illustrated embodiment will not be described in detail here.

[0074] Step S202, matching the target application scenario corresponding to the first application scenario description information in the knowledge base, wherein the knowledge base includes: when developing machine learning models in different application scenarios, the corresponding development workflow, data set information, and algorithm information of at least one available algorithm.

[0075] Specifically, the knowledge base also includes: description information corresponding to different application scenarios. This description information can be understood as a specific description of the model application scenario when building models for different application scenarios, such as what functions the model needs to achieve, under what circumstances it is used, what type of problem the application scenario belongs to, what is the goal of the model, etc. For example, taking the model application scenario of bank customer streaming prediction in the field of bank marketing as an example, the information format of the corresponding scenario description information can be as follows:

[0076] "model_scene": "Bank Customer Churn Prediction";

[0077] "model_scene_description": "Bank customer churn prediction is a typical machine learning binary classification problem, which is mainly used in the financial services industry.

[0078] In this scenario, the bank, as a financial institution, aims to retain and enhance customer satisfaction as much as possible in order to maintain business stability and growth.

[0079] The main task of a churn prediction model is to identify those customers who are likely to stop using the bank's services in the future (churners) and those who are likely to continue to be loyal customers (non-churners)".

[0080] Through the above-mentioned scenario description information, the machine learning platform can understand the specific information corresponding to each model application scenario, so as to facilitate corresponding matching. At the same time, when subsequently generating a machine learning model corresponding to the scenario, the specific requirements of the model can be determined according to the specific scenario description information, so as to facilitate the machine learning platform to build the model.

[0081] Specifically, in step S202, matching the target application scenario corresponding to the first application scenario description information in the knowledge base includes:

[0082] Step S202-1: convert the first application scenario description information and description information corresponding to different application scenarios in the knowledge base into a vector format through a preset language model.

[0083] The first application scenario description information and the description information corresponding to different application scenarios in the knowledge base are all text information, and the description information corresponding to the application scenario can be converted into a vector format through some trained language models, such as BERT-base or RoBERTa. Specifically, when the knowledge base is pre-built, the application scenario description information in the knowledge base can be directly saved in a vector format, so that when matching, only the description information corresponding to the first application scenario needs to be converted into a vector format, which reduces time consumption.

[0084] Step S202-2: compare the first application scenario description information in vector format with the description information corresponding to different application scenarios in the knowledge base to determine the target application scenario.

[0085] After obtaining the application scenario description information in vector format, the vector distance calculation method can be used to quickly query the target application scenario corresponding to the first application scenario description information, thereby determining the data set information, algorithm information, and development workflow corresponding to the target application scenario in the knowledge base, so that the machine learning platform can generate a corresponding machine learning model through the thinking chain technology according to the specific description in the above information in the knowledge base.

[0086] Step S203: Generate a first machine learning model corresponding to each available algorithm based on the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario.

[0087] Specifically, in step S203, based on the development workflow corresponding to the target application scenario, the data set information, and the algorithm information of at least one available algorithm, a first machine learning model corresponding to each available algorithm is generated, including:

[0088] Step S203-1, based on the development workflow corresponding to the target application scenario, data set information, and algorithm information of at least one available algorithm, generate development code for the first machine learning model corresponding to each available algorithm.

[0089] It can be understood that the machine learning platform has the ability to generate code. After determining the development workflow, data set information, and algorithm information corresponding to the target application scenario, the machine learning platform can form a chain of thought for the development process based on the description of the specific work content corresponding to each link in the development workflow, combined with the specific data set information and the specific description in the algorithm information. Based on this chain of thought, the development code of the machine learning model corresponding to each algorithm is automatically generated.

[0090] Specifically, the development workflow includes: work content corresponding to each work node in the development workflow.

[0091] It can be understood that the development workflow simulates the workflow of real data scientists when developing machine learning models. It can be understood as a complete set of data analysis and algorithm development step guides. It includes a description of the specific work content corresponding to each step when developing a machine learning model. For example, the development workflow may include content descriptions corresponding to multiple development work links such as data preprocessing, feature engineering, model selection, parameter tuning, model evaluation, and result interpretation.

[0092] For example, in the application scenario of predicting customer churn in banking marketing business, when building its corresponding machine learning model, the information format of the work content of each work node in the corresponding development workflow can be as follows:

[0093] "model_workflow": [

[0094] {

[0095] "workflow_status": "Requirement Analysis",

[0096] "workflow_content": "- Identify business objectives: Understand what the bank wants to achieve through predictive models, such as improving customer retention, reducing customer acquisition costs, etc.\n- Define the problem: Make it clear whether the problem is to predict whether a customer will churn, or to determine what factors lead to churn.\n- Data collection: Understand the characteristics and limitations of the existing dataset, and what additional data may be needed to supplement it."

[0097] },

[0098] {

[0099] "workflow_status": "Feature Engineering",

[0100] "workflow_content": "- Data cleaning: missing value processing, outlier processing, to ensure data quality.\n- Data transformation: encoding categorical variables (such as One-Hot encoding), standardizing or normalizing numerical features.\n- Feature engineering: creating new features such as interaction terms, time series features, etc. based on business knowledge and model requirements."

[0101] },

[0102] {

[0103] "workflow_status": "Model training",

[0104] "workflow_content": "- Select an appropriate algorithm, such as logistic regression, random forest, decision tree, or XGBoost, based on the problem characteristics (e.g., data size, number of features, nonlinear relationships, etc.).\n- Divide the dataset: divide the dataset into training set, validation set, and test set.\n- Model training: train the model using the training set and adjust the parameters to optimize the performance."

[0105] },

[0106] {

[0107] "workflow_status": "Model Evaluation",

[0108] "workflow_content": "- Use the validation set to evaluate the model performance, such as accuracy, precision, recall, F1 score, AUC-ROC curve, etc.\n- Choose appropriate evaluation metrics, because accuracy may not be the best metric for imbalanced data."

[0109] },

[0110] {

[0111] "workflow_status": "Hyperparameter optimization",

[0112] "workflow_content": "- Find the optimal model parameter combination through grid search method."

[0113] }, ]

[0115] Specifically, in step S203-1, according to the development workflow corresponding to the target application scenario, the data set information, and the algorithm information of at least one available algorithm, the development code of the first machine learning model corresponding to each available algorithm is generated, including:

[0116] Step S203-1-1: for each available algorithm, based on the algorithm information and data set information of the available algorithm and the work content corresponding to each work node in the development workflow, generate the development code corresponding to each work node.

[0117] Step S203-1-2, splicing the development codes corresponding to each working node to obtain the development code of the first machine learning model corresponding to the available algorithm.

[0118] For step S203-1-1, it can be understood that when generating the development code of each algorithm corresponding to the machine learning model, the machine learning platform may need to generate the development code required for the work content corresponding to each work node based on the specific work content in each work node in the development workflow, the specific description of the data set in the data set information under the target application scenario in the knowledge base, the specific description of the algorithm in the algorithm information of each algorithm, and the relevant information about the data set and algorithm in the knowledge base.

[0119] For example, the data set information may include: a detailed description of the basic information of the data set in the data set used to train the model, as well as the meaning of each feature column in the data set, label definition, and the correlation analysis results between features and labels. Taking the bank customer streaming prediction scenario as an example, the specific information format of the corresponding data set information may be as follows:

[0120] "dataset_description": "Based on the bank customer churn prediction dataset, predict whether customers will churn to help financial institutions develop retention strategies.\nI have a bank customer churn prediction dataset, which contains features divided into the following three dimensions:\n(1) Personal information (feature columns: age, gender, geographic location)\n(2) Financial information (feature columns: customer years, balance, salary estimate)\n(3) Historical purchase records (feature columns: number of products purchased, whether there is a credit card, customer activity)".

[0121] "datase_column": [

[0122] {

[0123] "feature_name": "CreditScore",

[0124] "feature_description": "Record number, corresponds to the record (row) number, has no effect on the output"

[0125] },

[0126] …

[0127] {

[0128] "feature_name": "EstimatedSalary",

[0129] "feature_description": "Salary estimates show that people with lower salaries are more likely to leave the bank than those with higher salaries."

[0130] },

[0131] {

[0132] "label_name": "Exited",

[0133] "label_description": "Whether the customer has left the bank, label column."

[0134] } ]

[0136] Exemplarily, the specific information format of the algorithm information may be as follows:

[0137] "algorithm_description": "Whether bank customers will churn or not is a typical machine learning binary classification task.",

[0138] "algorithms":[

[0139] {

[0140] "algorithm_name": "K-Nearest Neighbors (KNN)",

[0141] "algorithm_advantages": "- Simple in theory, no training process required.",

[0142] "algorithm_disadvantages": "- The prediction process requires large amount of calculations, is sensitive to noise data, and has poor results.",

[0143] "algorithm_scene": "- The algorithm is simple, suitable for small data sets, with high computational complexity and high requirements for data preprocessing.",

[0144] "algorithm_hyper_opt": "- n_neighbors: number of neighbors, integer, range [3, 20]"

[0145] },

[0146] … ]

[0148] The above content only describes the information corresponding to an algorithm by way of example. In an actual knowledge base, for each application scenario, there may be multiple available algorithms under it. For each algorithm, its corresponding algorithm information will be sorted out in the above format to obtain the algorithm information corresponding to each application scenario.

[0149] When the machine learning platform develops the corresponding code according to the work content corresponding to each work node in the development workflow, it can combine the algorithm information and data set information, as well as the specific description information of the application scenario in the knowledge base, to form a thinking chain of the machine learning platform model development process, thereby generating the development code corresponding to each work node for subsequent model generation.

[0150] For step S203-1-2, it can be understood that, for each machine learning model corresponding to an algorithm, after generating the development code of each corresponding working link, the machine learning platform can splice these codes to obtain the development code of the machine learning model corresponding to each algorithm.

[0151] Step S203-2: According to the development code of the first machine learning model corresponding to each available algorithm, construct an initial model corresponding to each available algorithm, and train and adjust the parameters of the initial model to obtain the first machine learning model corresponding to each available algorithm.

[0152] After obtaining the development code of each first machine learning model, the initial model corresponding to each available algorithm can be obtained based on the development code. The development code corresponding to each initial model contains codes corresponding to specific processes such as data calling, training, hyperparameter optimization, etc. in all links. The initial model corresponding to each algorithm is trained and the parameters are adjusted through the code. Finally, the first machine learning models corresponding to each algorithm can be applied to the first application scenario.

[0153] Specifically, the algorithm information in the knowledge base includes: the hyperparameter knowledge corresponding to the algorithm. For example, the hyperparameter type corresponding to each algorithm when building a machine learning model, and the adjustment range of each type of hyperparameter.

[0154] Specifically, in step S203-2, the initial model is trained and adjusted to obtain a first machine learning model corresponding to each available algorithm, including:

[0155] Step S203-2-1, calling the training data and test data of the initial model corresponding to each available algorithm.

[0156] Since the algorithms corresponding to different initial models are different, the corresponding training data may also be different. The original data of these training data can be the same. The specific description information of the algorithm involves the format requirements of the training data when training the model corresponding to the algorithm.

[0157] Therefore, the machine learning platform preprocesses the raw data according to the specific description information corresponding to each algorithm, so the training data corresponding to different initial models may be different. For test data, the format is the same as the training data, but the content is different, so as to check the specific effect of the trained machine learning model.

[0158] Step S203-2-2, training the initial model using training data, and adjusting model parameters of the initial model.

[0159] The machine learning platform can perform corresponding training based on the specific development code corresponding to the model, and adjust the relevant parameters of the model during the training process.

[0160] Step S203-2-3, testing the trained initial models using test data to obtain evaluation results corresponding to each trained initial model.

[0161] After a certain number of rounds of training, the trained initial models can be tested to obtain the corresponding evaluation results of each trained initial model. The specific training status of the model can be understood based on the evaluation results.

[0162] Step S203-2-4, obtain the hyper-parameter knowledge corresponding to each available algorithm.

[0163] The algorithm information in the knowledge base includes the corresponding hyper-parameter knowledge. The machine learning platform can call the corresponding hyper-parameter knowledge for subsequent hyper-parameter adjustments.

[0164] Step S203-2-5, based on the hyper-parameter knowledge corresponding to each available algorithm and the evaluation results corresponding to each trained initial model, perform hyper-parameter optimization on each trained initial model to obtain the first machine learning model corresponding to each available algorithm.

[0165] After determining the evaluation results and hyperparameter knowledge corresponding to each initial model, the machine learning platform can develop specific content for the hyperparameter optimization part of the workflow, generate corresponding hyperparameter optimization code to automatically perform the operations required for hyperparameter optimization, and obtain the best hyperparameter combination corresponding to each trained initial model. The best hyperparameter combination is combined in the trained initial model, and corresponding iterative training is performed to obtain the first machine learning model corresponding to each available algorithm.

[0166] Step S204, obtaining first test data, and testing each first machine learning model using the first test data;

[0167] Generate corresponding evaluation reports based on the test results corresponding to each first machine learning model, so that the user can determine the target learning model from at least one first machine learning model based on the evaluation report.

[0168] It can be understood that after obtaining the first machine learning models corresponding to each available algorithm, since the machine learning models corresponding to different algorithms are bound to be different from each other, different models can be tested through test data to generate an evaluation report, which may include the advantages and disadvantages of each first machine learning model, as well as information used to evaluate the characteristics of the model such as performance under different indicators. Users can select the first machine learning model that best suits the actual application scenario based on the evaluation of the advantages and disadvantages of each first machine learning model in the evaluation report.

[0169] Specifically, in step S204, a corresponding evaluation report is generated according to the test results corresponding to each first machine learning model, including:

[0170] Determine the test results of each first machine learning model under each preset indicator, where the preset indicators include: accuracy, precision, and recall;

[0171] Generate a corresponding evaluation report based on the test results corresponding to each preset indicator.

[0172] It can be understood that when evaluating each first machine learning model, its performance under the three indicators of accuracy, precision and recall can be mainly considered, so as to generate an evaluation report corresponding to each model according to the corresponding test performance under each indicator. Specifically, different weights can be set for different indicators, so as to give the most recommended first machine learning model according to the specific performance of each indicator.

[0173] The method for generating a machine learning model provided in this embodiment obtains description information of the application scenario corresponding to the machine learning model that the user expects to build and converts it into a vector format to ensure the matching effect, and then matches the corresponding development workflow and data set information and at least one algorithm information in the knowledge base to generate at least one machine learning model, and at the same time generates an evaluation report corresponding to each machine learning model, so that the user can select the machine learning model that best meets the actual needs. Since the user only needs to describe the application scenario of the model to be built and does not need to understand the specific principles of model construction, the operation threshold for the user to generate a machine learning model through a machine learning platform can be greatly reduced. At the same time, the model can be generated according to the development workflow in the knowledge base, which can also ensure the training effect of the model.

[0174] In order to help understand the specific content of the above embodiments, for example, Figure 3 As shown, it is an example diagram of a process of executing machine learning model construction through a machine learning platform according to an embodiment of the present invention.

[0175] For step S1 in this method flow, an expert knowledge base for dataset recommendation, algorithm selection, and development workflow is constructed. Figure 4 , is an example diagram of a knowledge base construction process according to an embodiment of the present invention.

[0176] The constructed knowledge base covers a wealth of financial data sets and includes various classic machine learning algorithms, so that the machine learning platform can perform expert-level data set recommendations and algorithm selection.

[0177] First, collect description information of model scenarios that may be designed in a certain business field, as well as the corresponding data set information, machine learning algorithm knowledge and development workflow when data scientists develop machine learning models that meet the corresponding scenarios.

[0178] After obtaining these specific information, sort them out accordingly. For specific sorting methods, please refer to the above Figure 2 In the illustrated embodiment, the exemplary information format corresponding to the application scenario description information, data set information, algorithm information and development workflow in the bank customer streaming prediction scenario in the banking marketing field is given, which will not be repeated here.

[0179] After sorting out this information, we obtain the descriptive information corresponding to each application scenario, as well as the data set information, algorithm information and development workflow under each scenario. These contents are stored in the knowledge base as knowledge documents as expert experience. At the same time, the application scenario description information of the model is vectorized to generate a model scenario vector library. Each model scenario vector in the model scenario vector library is associated with the expert experience under the scenario, that is, the data set information, algorithm information and development workflow, so that when matching through the application scenario description information, the corresponding data set information, algorithm information and development workflow can be obtained in the knowledge base.

[0180] By structured combing of expert knowledge on data sets and algorithms, an expert knowledge document is formed that includes scenario model description, data set recommendation, algorithm selection, and data scientist workflow. Then, with the help of advanced pre-trained language models (such as BERT-base or RoBERTa), the scenario model description part is converted into a vector.

[0181] Users can quickly query the model scenario knowledge document content by entering questions or keywords and using the vector distance calculation method to obtain the expert knowledge most relevant to their needs. After retrieving relevant information, the machine learning platform combines the Chain of Thought (CoT) technology of the large model to provide in-depth interpretation and customized financial scenario model solutions.

[0182] Step S1 can be understood as the preparation work of the machine learning platform. After receiving the user's model building requirements, the steps corresponding to step S2 are executed. In step S2, when the knowledge base is searched based on user needs, the expert thinking chain is obtained, and the tool is scheduled, executed, and feedback is given.

[0183] Specific reference Figure 5 , which is an example diagram of the process of generating model code by a machine learning platform according to an embodiment of the present invention. After receiving the user's demand information, the closest application scenario in the knowledge base is matched according to the specific description of the model application scenario in the demand information, and the development workflow, data set information, and algorithm information under the scenario are called.

[0184] Since the development workflow, dataset information, and algorithm information all contain specific description information, the development workflow can be combined with the dataset information and algorithm information as an expert workflow thinking chain, namely COT.

[0185] Taking the streaming prediction scenario of bank customers as an example, the machine learning platform performs specific scheduling, execution and feedback based on the thinking chain, and generates the code corresponding to the data set acquisition API, the code corresponding to the data preprocessing API, the algorithm code and the evaluation code.

[0186] The code corresponding to the generated data set acquisition API can be understood as the machine learning platform using the large model information extraction capability to obtain request parameters including data set keywords, feature column name list, label column name and other information, and output the data set reading code block, which is used for the data source reading module of the subsequent algorithm code.

[0187] For the code corresponding to the generated data preprocessing API, it can be understood that the machine learning platform calls the data preprocessing API according to the requirements of the feature engineering part in the development workflow, extracts feature engineering keywords including missing value processing, outlier processing, One-Hot encoding, standardization, normalization, etc., and outputs feature engineering code blocks for the feature engineering module of subsequent algorithm codes.

[0188] For generating algorithm code, it can be understood that the machine learning platform uses the code generation capability of the large model based on the corresponding algorithm information in the knowledge base and the requirements of model training in the development work, including machine learning algorithm functions and the initial hyperparameters of the algorithm.

[0189] As for the generation of evaluation code, it can be understood that the machine learning platform generates model evaluation code based on the corresponding algorithm information in the knowledge base, the requirements for model evaluation in the development workflow, and general knowledge of machine learning. For example, commonly used evaluation indicators for machine learning classification tasks are accuracy, precision, recall, F1 score, etc., and commonly used evaluation indicators for machine learning regression tasks are MSE, MAE, and R-squared.

[0190] These codes are spliced ​​and integrated to generate training code, and the code verification capabilities of the large model, including syntax checking, code completion, and function verification, are used to ensure the integrity and executable nature of the training code.

[0191] After obtaining the training code corresponding to each model, execute step S3, model training, and large model analysis training evaluation indicators to trigger hyperparameter optimization. For the specific process, refer to Figure 6 , which is an example diagram of the process of performing model optimization on a machine learning platform according to this embodiment.

[0192] As shown in the figure, the model training code corresponding to each model is used to train the model concurrently, and after training, the model is evaluated to obtain the evaluation results corresponding to each model. It should be noted that the number of models corresponds to the number of algorithms in the corresponding application scenarios in the knowledge base. The machine learning platform calls the machine learning hyperparameter optimization knowledge corresponding to the target scenario in the expert knowledge base, and generates the corresponding hyperparameter optimization prompt words according to the specific training evaluation results of each model, performs the scheduling, execution and feedback related to the hyperparameter optimization based on the above-mentioned thinking chain, generates the code related to the hyperparameter optimization for hyperparameter optimization, and evaluates the machine learning model after the hyperparameter optimization to obtain an evaluation report for users to select the best machine learning model. For example, based on the machine learning expertise of the large model and the evaluation results of the hyperparameter optimization, the user algorithm final selection and the optimal hyperparameter combination can be given.

[0193] Specifically, when performing hyperparameter optimization, optimization interfaces for hyperparameter grid search of machine learning algorithms supported by various platforms can be prefabricated. According to the preset machine learning algorithm and the parameters given in the algorithm selection-hyperparameter optimization part in step S1, as well as the parameter type and parameter grid search range, the large model agent in the machine learning platform dispatches the corresponding API and extracts the input parameter field information to achieve hyperparameter optimization of the algorithm and output the hyperparameter combination conclusion.

[0194] Finally, in step S4, the big model generates a model training report. For the entire scenario model development process, according to the requirements of the data scientist workflow in step S1, including demand analysis, feature engineering, model training, model evaluation, hyperparameter optimization and other nodes, as well as the execution of each node in step S2 and step S3, the training report generation prompt words are summarized, and the analysis and summary capabilities of the big model are used to generate an overall training workflow report.

[0195] The present invention aims to propose an enhanced closed-loop automatic learning system that integrates thinking chains and knowledge bases. Users use natural language to input the training requirements of machine learning models in financial scenarios, obtain CoT from the expert knowledge base through RAG retrieval technology, and realize model training, model evaluation and hyper-parameter optimization through the planning capabilities of large models, and finally output a model training report to the user.

[0196] For the above Figure 3 The exemplary embodiment shown in FIG. 1 may be used as follows: Figure 7 As shown, this is an example diagram of a closed-loop full process generated by a machine learning model according to an embodiment of the present invention.

[0197] As shown in the figure, the user inputs the model requirements, that is, the specific application scenario of the model. The machine learning platform searches through RAG and matches the corresponding development workflow, data set information, and algorithm information in the expert knowledge base. The matched development workflow, data set information, and algorithm information are used as the thinking chain COT of the model development process. Model training and hyperparameter optimization are performed according to the thinking chain. Finally, the machine learning model is trained and evaluated, and the final model training report is generated and sent to the user, so that the user can determine the best machine learning model.

[0198] In the embodiment of the present invention, a knowledge base including data sets, algorithm knowledge and data scientist workflow thinking chains is established in the machine learning platform to realize the automatic generation and optimization of intelligent agents. This not only improves the development efficiency, but also ensures the accuracy and practicality of the model, and makes full use of the expert experience of data scientists. The data scientist workflow thinking chain CoT (Continual On-Task Learning) and the knowledge base retrieval enhancement technology RAG (Retrieval-Augmented Generation) are effectively integrated. CoT allows large models to make continuous decisions, execution, and feedback in specific tasks, while RAG is good at information retrieval. This combination enables the system to quickly locate information while improving the quality and personalization of answers through deep understanding and expert-level recommendations. In addition, when the system develops the model, it adopts a closed-loop automatic process to develop it, and provides real-time model optimization strategies and evaluation feedback during the development process. This real-time feedback mechanism enables the system to continuously optimize performance and improve the flexibility and accuracy of the entire information processing process.

[0199] The present invention can be used in the field of intelligent marketing for banks, and combined with large language model intelligent agent technology, it can provide banks with marketing decision support tools to overcome the limitations of the current mainstream closed-loop automated learning system. The current closed-loop automated learning system mainly focuses on low-code applications, and has high requirements for the user's basic knowledge of machine learning, thus failing to achieve a truly low technical entry threshold. By introducing automatic machine learning technology and large model intelligent agent technology, it aims to lower barriers to entry and eliminate strict restrictions on user expertise, so that it can seamlessly accept a wider user group, and realize financial applications that can be enabled by artificial intelligence through human-computer interaction regardless of the user's artificial intelligence technical background.

[0200] In this embodiment, a device for generating a machine learning model is also provided, which is used to implement the above-mentioned embodiments and preferred implementation modes, and will not be repeated here. As used below, the term "module" can implement a combination of software and / or hardware for a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0201] This embodiment provides a device for generating a machine learning model, such as Figure 8 As shown, including:

[0202] The information acquisition module 401 is used to obtain the first application scenario description information corresponding to the machine learning model that the user expects to build.

[0203] The scenario matching module 402 is used to match the target application scenario corresponding to the first application scenario description information in the knowledge base, and the knowledge base includes: when developing machine learning models in different application scenarios, the corresponding development workflow, data set information and algorithm information of at least one available algorithm.

[0204] The model development module 403 is used to generate a first machine learning model corresponding to each available algorithm based on the development workflow corresponding to the target application scenario, data set information, and algorithm information of at least one available algorithm.

[0205] In some optional implementations, the model development module 403, after constructing the first machine learning model corresponding to each available algorithm, is further used to obtain first test data and test each first machine learning model through the first test data;

[0206] Generate corresponding evaluation reports based on the test results corresponding to each first machine learning model, so that the user can determine the target learning model from at least one first machine learning model based on the evaluation report.

[0207] In some optional implementations, the knowledge base also includes: description information corresponding to different application scenarios;

[0208] The scenario matching module 402, when matching the target application scenario corresponding to the first application scenario description information in the knowledge base, includes:

[0209] The first application scenario description information and description information corresponding to different application scenarios in the knowledge base are converted into a vector format through a preset language model;

[0210] The first application scenario description information in the vector format is compared with description information corresponding to different application scenarios in the knowledge base to determine the target application scenario.

[0211] In an optional implementation, the model development module 403, when generating a first machine learning model corresponding to each available algorithm based on the development workflow corresponding to the target application scenario, the data set information, and the algorithm information of at least one available algorithm, includes:

[0212] Generate development code of a first machine learning model corresponding to each available algorithm according to a development workflow corresponding to a target application scenario, data set information, and algorithm information of at least one available algorithm;

[0213] According to the development code of the first machine learning model corresponding to each available algorithm, an initial model corresponding to each available algorithm is constructed, and the initial model is trained and adjusted to obtain the first machine learning model corresponding to each available algorithm.

[0214] In an optional implementation, the algorithm information in the knowledge base includes: hyper-parameter knowledge corresponding to the algorithm;

[0215] The model development module 403, when training and adjusting the parameters of the initial model to obtain the first machine learning model corresponding to each available algorithm, includes:

[0216] Call the training data and test data of each available algorithm corresponding to the initial model;

[0217] Train the initial model through the training data and adjust the model parameters of the initial model;

[0218] The trained initial models are tested using test data to obtain evaluation results corresponding to each trained initial model;

[0219] Obtain the hyperparameter knowledge corresponding to each available algorithm;

[0220] Based on the hyper-parameter knowledge corresponding to each available algorithm and the evaluation results corresponding to each trained initial model, each trained initial model is hyper-parameter optimized to obtain the first machine learning model corresponding to each available algorithm.

[0221] In an optional implementation, the development workflow includes: work content corresponding to each work node in the development workflow;

[0222] The model development module 403, when generating development codes of the first machine learning model corresponding to each available algorithm according to the development workflow corresponding to the target application scenario, the data set information, and the algorithm information of at least one available algorithm, includes:

[0223] For each available algorithm, based on the algorithm information of the available algorithm, the data set information, and the work content corresponding to each work node in the development workflow, generate the development code corresponding to each work node;

[0224] The development codes corresponding to each working node are spliced ​​together to obtain the development code of the first machine learning model corresponding to the available algorithm.

[0225] In an optional implementation, the model development module 403, when generating a corresponding evaluation report according to the test results corresponding to each first machine learning model, includes:

[0226] Determine the test results of each first machine learning model under each preset indicator, where the preset indicators include: accuracy, precision, and recall;

[0227] Generate a corresponding evaluation report based on the test results corresponding to each preset indicator.

[0228] The further functional description of each of the above modules and units is the same as that of the above corresponding embodiments and will not be repeated here.

[0229] The device for generating the machine learning model in this embodiment is presented in the form of a functional unit, where the unit refers to an ASIC (Application Specific Integrated Circuit) circuit, a processor and memory that executes one or more software or fixed programs, and / or other devices that can provide the above-mentioned functions.

[0230] The embodiment of the present invention also provides a computer device having the above Figure 8 The generation device of the machine learning model shown.

[0231] See also Fig. 9 , Fig. 9 is a schematic diagram of the structure of a computer device provided by an optional embodiment of the present invention, such as Fig. 9As shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including high-speed interfaces and low-speed interfaces. Various components are connected to each other using different buses for communication, and can be installed on a common mainboard or installed in other ways as needed. The processor can process the instructions executed in the computer device, including instructions stored in or on the memory to display the graphical information of the GUI on an external input / output device (such as, a display device coupled to the interface). In some optional embodiments, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a group of blade servers, or a multi-processor system). Fig. 9 A processor 10 is taken as an example.

[0232] The processor 10 may be a central processing unit, a network processor or a combination thereof. The processor 10 may further include a hardware chip. The hardware chip may be a dedicated integrated circuit, a programmable logic device or a combination thereof. The programmable logic device may be a complex programmable logic device, a field programmable gate array, a general purpose array logic or any combination thereof.

[0233] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0234] The memory 20 may include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function; the data storage area may store data created according to the use of the computer device, etc. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some optional embodiments, the memory 20 may optionally include a memory remotely arranged relative to the processor 10, and these remote memories may be connected to the computer device via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0235] The memory 20 may include a volatile memory, such as a random access memory; the memory may also include a non-volatile memory, such as a flash memory, a hard disk or a solid state drive; the memory 20 may also include a combination of the above types of memory.

[0236] The computer device also includes an input device 30 and an output device 40. The processor 10, the memory 20, the input device 30 and the output device 40 may be connected via a bus or other means. Fig. 9 The example of connecting through bus is taken in the following.

[0237] The input device 30 can receive input digital or character information, and generate key signal input related to the user settings and function control of the computer device, such as a touch screen, a keypad, a mouse, a track pad, a touch pad, an indicator bar, one or more mouse buttons, a trackball, a joystick, etc. The output device 40 may include a display device, an auxiliary lighting device (e.g., an LED) and a tactile feedback device (e.g., a vibration motor), etc. The above-mentioned display device includes but is not limited to a liquid crystal display, a light emitting diode, a display and a plasma display. In some optional embodiments, the display device may be a touch screen.

[0238] The embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or can be implemented as a computer code that can be recorded in a storage medium, or can be implemented as a computer code that is originally stored in a remote storage medium or a non-temporary machine-readable storage medium and will be stored in a local storage medium through a network download, so that the method described herein can be stored in such software processing on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only storage memory, a random access memory, a flash memory, a hard disk or a solid-state hard disk, etc.; further, the storage medium can also include a combination of the above types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by a computer, a processor, or hardware, the method shown in the above embodiment is implemented.

[0239] Although the embodiments of the present invention have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for generating a machine learning model, applied to a machine learning platform, the machine learning platform comprising: A pre-built knowledge base, characterized in that the method comprises: Obtaining first application scenario description information corresponding to the machine learning model that the user expects to build; Matching a target application scenario corresponding to the first application scenario description information in the knowledge base, wherein the knowledge base includes: corresponding development workflow, data set information, and algorithm information of at least one available algorithm when developing machine learning models in different application scenarios; Based on the development workflow corresponding to the target application scenario, the data set information, and the algorithm information of at least one available algorithm, generate a first machine learning model corresponding to each available algorithm; The step of generating a first machine learning model corresponding to each available algorithm based on the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario includes: Generate development code of a first machine learning model corresponding to each available algorithm according to the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario; According to the development code of the first machine learning model corresponding to each available algorithm, construct an initial model corresponding to each available algorithm, and train and adjust the parameters of the initial model to obtain the first machine learning model corresponding to each available algorithm; The development workflow includes: work content corresponding to each work node in the development workflow; The step of generating development codes of the first machine learning model corresponding to each available algorithm according to the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario includes: For each available algorithm, based on the algorithm information and data set information of the available algorithm and the work content corresponding to each work node in the development workflow, generate development code corresponding to each work node; The development codes corresponding to the various working nodes are spliced ​​to obtain the development code of the first machine learning model corresponding to the available algorithm.

2. The method according to claim 1, characterized in that After constructing the first machine learning model corresponding to each available algorithm, the method further includes: Acquire first test data, and test each of the first machine learning models using the first test data; Generate a corresponding evaluation report based on the test results corresponding to each first machine learning model, so that the user can determine the target learning model from at least one first machine learning model based on the evaluation report.

3. The method according to claim 1, characterized in that The knowledge base also includes: description information corresponding to different application scenarios; The matching of the target application scenario corresponding to the first application scenario description information in the knowledge base includes: Converting the first application scenario description information and description information corresponding to different application scenarios in the knowledge base into a vector format through a preset language model; The first application scenario description information in the vector format is compared with description information corresponding to different application scenarios in the knowledge base to determine a target application scenario.

4. The method according to claim 1, characterized in that: The algorithm information in the knowledge base includes: super-parameter knowledge corresponding to the algorithm; The training and parameter adjustment of the initial model to obtain a first machine learning model corresponding to each available algorithm includes: Call the training data and test data of each available algorithm corresponding to the initial model; Training the initial model using the training data, and adjusting model parameters of the initial model; The trained initial models are tested using the test data to obtain evaluation results corresponding to the trained initial models; Obtain the hyper-parameter knowledge corresponding to each available algorithm; Based on the hyper-parameter knowledge corresponding to each available algorithm and the evaluation results corresponding to each trained initial model, hyper-parameter optimization is performed on each trained initial model to obtain a first machine learning model corresponding to each available algorithm.

5. The method according to claim 2, characterized in that: The generating a corresponding evaluation report according to the test results corresponding to each first machine learning model includes: Determine the test results of each first machine learning model under each preset indicator, wherein the preset indicators include: accuracy, precision, and recall; Generate a corresponding evaluation report based on the test results corresponding to each preset indicator.

6. A device for generating a machine learning model, applied to a machine learning platform, the machine learning platform comprising: A pre-built knowledge base, characterized in that the device comprises: An information acquisition module, used to acquire first application scenario description information corresponding to the machine learning model that the user expects to build; A scenario matching module, configured to match a target application scenario corresponding to the first application scenario description information in the knowledge base, wherein the knowledge base includes: corresponding development workflows, data set information, and algorithm information of at least one available algorithm when developing machine learning models in different application scenarios; A model development module, configured to generate a first machine learning model corresponding to each available algorithm based on a development workflow corresponding to the target application scenario, data set information, and algorithm information of at least one available algorithm; The step of generating a first machine learning model corresponding to each available algorithm based on the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario includes: Generate development code of a first machine learning model corresponding to each available algorithm according to the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario; According to the development code of the first machine learning model corresponding to each available algorithm, construct an initial model corresponding to each available algorithm, and train and adjust the parameters of the initial model to obtain the first machine learning model corresponding to each available algorithm; The development workflow includes: work content corresponding to each work node in the development workflow; The step of generating development codes of the first machine learning model corresponding to each available algorithm according to the development workflow, data set information, and algorithm information of at least one available algorithm corresponding to the target application scenario includes: For each available algorithm, based on the algorithm information and data set information of the available algorithm and the work content corresponding to each work node in the development workflow, generate development code corresponding to each work node; The development codes corresponding to the various working nodes are spliced ​​to obtain the development code of the first machine learning model corresponding to the available algorithm.

7. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for generating a machine learning model according to any one of claims 1 to 5 by executing the computer instructions.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for generating a machine learning model according to any one of claims 1 to 5.