Model construction method and device, model use method and device, electronic equipment and readable medium

Abstract

The invention discloses a model construction method and device, a model use method and device, electronic equipment and a readable medium, and the method comprises the steps: obtaining a target model in response to an interface parameter configured by a user on an interface provided by an integrated tool; in response to a processing statement input on the interface by a user, obtaining a to-be-processed data set; and obtaining a processing result according to the target model and the to-be-processed data set. It can be seen that the integration tool integrates model types suitable for various business scenes and automatically stores various types of target models obtained through training, a user can use simple and flexible interface parameter configuration options provided by the integration tool according to requirements, the target model corresponding to the requirements can be conveniently called, and the user experience is improved. And the to-be-processed data is provided for the integration tool based on the simple processing statement, and the integration tool can automatically process the to-be-processed data based on the target model, so that it is possible for a user to easily and accurately complete the construction and use of the machine learning model by means of the integration tool.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a model construction, a model use method, a device, an electronic device and a readable medium. Background Art

[0002] With the development of machine learning (ML) related theories, the explosive growth of online data, and the demand for low-cost computing, machine learning has become a commonly used or urgently needed tool in many industries. However, the construction and use of machine learning models are usually more professional and complex, requiring users to have relevant knowledge.

[0003] At present, in order to provide non-professional users with the convenience of building and using machine learning models, some encapsulated machine learning tools have emerged. However, when using these encapsulated machine learning tools, there will be problems such as multiple process modules, complex interfaces, and single functions. Non-professional users are required to have certain professional knowledge, which is not friendly enough for non-professional users.

[0004] It can be seen that the current packaged machine learning tools cannot provide convenience for building and using machine learning models. Summary of the invention

[0005] In order to solve the above technical problems, the present application provides a model building method, a model using method, an apparatus, an electronic device and a readable medium, which can facilitate users to build the required machine learning models and also facilitate users to use various trained machine learning models.

[0006] In order to achieve the above objectives, the technical solutions provided by this application are as follows:

[0007] In a first aspect, the present application provides a model use method, which may include:

[0008] In response to interface parameters configured by a user on an interface provided by the integration tool, a target model is obtained, wherein the interface parameters include model parameters corresponding to the target model, and the target model is a model trained in the integration tool;

[0009] In response to a processing statement input by the user on the interface, obtaining a data set to be processed, wherein the processing statement includes database parameters required for calling the data set to be processed;

[0010] A processing result is obtained according to the target model and the data set to be processed, and the processing result is used to guide a processing strategy for the data set to be processed.

[0011] In a possible implementation, the interface parameters further include input parameters and output parameters corresponding to the target model, and obtaining the target model includes:

[0012] In response to a selection statement input by a user on an interface provided by the integration tool, obtaining at least one initial model, the initial model being of the same type as the target model;

[0013] In response to the training statements and configured interface parameters input by the user on the interface provided by the integrated tool, a training data set is obtained;

[0014] The target model is obtained according to the training data set and the at least one initial model.

[0015] In another possible implementation, the model parameters include the name of the target model and a saving path of the target model, and obtaining the target model includes:

[0016] Acquire a model file corresponding to the name of the target model from the saving path of the target model;

[0017] The target model is obtained according to the model file.

[0018] The target model may be any of the following types of models integrated in the integration tool:

[0019] Attribution, regression, or classification models.

[0020] In a possible implementation, the processing statement is a Structured Query Language (SQL) statement.

[0021] In a second aspect, the present application also provides a model building method, which may include:

[0022] In response to a selection statement input by a user on an interface provided by the integration tool, obtaining at least one initial model belonging to a target type, the target type corresponding to the selection statement;

[0023] In response to the training statements and configured interface parameters input by the user on the interface provided by the integrated tool, obtaining a training data set;

[0024] Based on the training data set and the at least one initial model, a target model is obtained, where the target model is a trained model belonging to the target type.

[0025] In a possible implementation, the interface parameter includes an input parameter, and the input parameter is used to indicate an input rule of a training sample, and obtaining a training data set includes:

[0026] In response to a training statement input by the user on an interface provided by the integrated tool, obtaining a first data set;

[0027] In response to input parameters input by the user on the interface provided by the integrated tool, the first data set is processed to obtain the training data set, and the training data set complies with the input rules.

[0028] As an example, the interface parameters further include output parameters and model parameters, the output parameters are used to indicate output rules of the target model, and the model parameters are used to indicate storage rules of the target model, and the method further includes:

[0029] According to the output rule and the saving rule, the model file corresponding to the target model is saved.

[0030] In a possible implementation manner, obtaining a target model based on the training data set and the at least one initial model includes:

[0031] Training each of the at least one initial model based on the training data set to obtain at least one candidate model, where the at least one candidate model corresponds one-to-one to the at least one initial model;

[0032] Obtaining an evaluation result of each candidate model in the at least one candidate model, wherein the evaluation result is used to indicate an effect of executing machine learning of the corresponding candidate model;

[0033] Based on the evaluation result, the target model is selected from the at least one candidate model.

[0034] In one possible implementation, the input parameters in the interface parameters include a sample set division rule and a target value processing method, the sample set division rule includes random division by proportion or division by partition, and the target value processing method indicates: according to the discrete degree of the target value field value, the target value field value is processed through function mapping.

[0035] The target model is any one of the following types of models integrated in the integration tool:

[0036] Attribution, regression, or classification models.

[0037] In a possible implementation, the training statement is a SQL statement.

[0038] In a third aspect, the present application also provides a model using device, which may include:

[0039] A first obtaining unit, configured to obtain a target model in response to interface parameters configured by a user on an interface provided by the integration tool, wherein the interface parameters include model parameters corresponding to the target model, and the target model is a model trained in the integration tool;

[0040] A second obtaining unit, configured to obtain a data set to be processed in response to a processing statement input by the user on the interface, wherein the processing statement includes database parameters required for calling the data set to be processed;

[0041] The third obtaining unit is used to obtain a processing result according to the target model and the data set to be processed, wherein the processing result is used to guide a processing strategy for the data set to be processed.

[0042] In a possible implementation manner, the interface parameters further include input parameters and output parameters corresponding to the target model, and the first obtaining unit includes:

[0043] A first obtaining subunit, configured to obtain at least one initial model in response to a selection statement input by a user on an interface provided by the integration tool, wherein the initial model is of the same type as the target model;

[0044] A second obtaining subunit, for obtaining a training data set in response to a training statement and configured interface parameters input by a user on an interface provided by the integrated tool;

[0045] The third obtaining subunit is used to obtain the target model according to the training data set and the at least one initial model.

[0046] In another possible implementation, the model parameters include the name of the target model and a saving path of the target model, and the first obtaining unit includes:

[0047] A fourth obtaining subunit is used to obtain a model file corresponding to the name of the target model from a saving path of the target model;

[0048] The fifth obtaining subunit is used to obtain the target model according to the model file.

[0049] The target model may be any of the following types of models integrated in the integration tool:

[0050] Attribution, regression, or classification models.

[0051] In a possible implementation, the processing statement is a SQL statement.

[0052] It should be noted that the specific implementation method of the device and the technical effects achieved can refer to the relevant description of the method provided in the first aspect or any implementation method of the first aspect.

[0053] In a fourth aspect, the present application further provides a model building device, which may include:

[0054] A first obtaining unit, configured to obtain at least one initial model belonging to a target type in response to a selection statement input by a user on an interface provided by the integration tool, wherein the target type corresponds to the selection statement;

[0055] A second obtaining unit, configured to obtain a training data set in response to a training statement and configured interface parameters input by the user on an interface provided by the integrated tool;

[0056] The third obtaining unit is used to obtain a target model based on the training data set and the at least one initial model, where the target model is a trained model belonging to the target type.

[0057] In a possible implementation, the interface parameter includes an input parameter, and the input parameter is used to indicate an input rule of a training sample, and the second obtaining unit includes:

[0058] A first obtaining subunit, configured to obtain a first data set in response to a training sentence input by the user on an interface provided by the integrated tool;

[0059] The second obtaining subunit is used to process the first data set in response to the input parameters input by the user on the interface provided by the integrated tool to obtain the training data set, where the training data set complies with the input rules.

[0060] As an example, the interface parameters further include output parameters and model parameters, the output parameters are used to indicate the output rule of the target model, and the model parameters are used to indicate the saving rule of the target model. The device may further include:

[0061] A saving unit is used to save the model file corresponding to the target model according to the output rule and the saving rule.

[0062] In a possible implementation, the third obtaining unit includes:

[0063] A training subunit, configured to train each of the at least one initial model based on the training data set to obtain at least one candidate model, wherein the at least one candidate model corresponds one-to-one to the at least one initial model;

[0064] An evaluation subunit, used to obtain an evaluation result of each candidate model in the at least one candidate model, wherein the evaluation result is used to indicate an effect of executing machine learning of the corresponding candidate model;

[0065] A selection subunit is used to select the target model from the at least one candidate model based on the evaluation result.

[0066] In one possible implementation, the input parameters in the interface parameters include a sample set division rule and a target value processing method, the sample set division rule includes random division by proportion or division by partition, and the target value processing method indicates: according to the discrete degree of the target value field value, the target value field value is processed through function mapping.

[0067] The target model is any one of the following types of models integrated in the integration tool:

[0068] Attribution, regression, or classification models.

[0069] In a possible implementation, the training statement is a SQL statement.

[0070] It should be noted that the specific implementation method of the device and the technical effects achieved can refer to the relevant description of the method provided in the second aspect or any implementation method of the second aspect.

[0071] In a fifth aspect, the present application further provides an electronic device, the electronic device comprising: a processor and a memory;

[0072] The memory is used to store instructions or programs;

[0073] The processor is used to execute the instructions or programs in the memory so that the electronic device executes the method provided by the first aspect, any one of the implementations of the first aspect, the second aspect, or any one of the implementations of the second aspect.

[0074] In the sixth aspect, the present application also provides a readable medium, which stores instructions or programs. When the instructions or programs are run on a processor, the processor executes the method provided by the first aspect, any one of the implementations of the first aspect, the second aspect, or any one of the implementations of the second aspect.

[0075] Compared with the prior art, the embodiments of the present application have at least the following advantages:

[0076] In the technical solution provided by the present application, if the user wants to build a model suitable for a certain business scenario, the user needs to execute: first, input a selection statement on the interface provided by the integrated tool, and the integrated tool can automatically obtain at least one initial model, and the at least one initial model belongs to the target type and the target type corresponds to the selection statement; second, input a training statement and configure interface parameters on the interface provided by the integrated tool, and the integrated tool can automatically obtain a training data set; in this way, the integrated tool can obtain a target model based on the training data set and at least one initial model, and the target model is a trained model belonging to the target type. It can be seen that the integrated tool integrates model types suitable for various business scenarios, and the user can use the integrated tool to build a model that meets the needs of any business scenario. Moreover, the integrated tool provides simple and flexible configuration options for interface parameters, and is compatible with simple selection statement and training statement input. The user does not need to understand the model principle and related parameters, which reduces the user's code participation in the process of building the model. Only through simple configuration and statement input, the required model can be built, making it possible for users with zero code foundation to easily and accurately build a model that meets their needs.

[0077] In addition, the present application also provides a method for using a model. If the user wants to use a model of the target type, the user needs to perform the following operations: first, configure the interface parameters on the interface provided by the integration tool, and the integration tool can automatically obtain the target model. The interface parameters at least include the model parameters corresponding to the target model, and the target model is a trained model in the integration tool; second, enter the processing statement on the interface provided by the integration tool, and the integration tool can automatically obtain the data set to be processed. The processing statement includes the database parameters required to call the data set to be processed; in this way, the integration tool can obtain the processing result based on the target model and the data set to be processed, and the processing result is used to guide the processing strategy for the data set to be processed. In this way, the integration tool can save the training results of the integrated models suitable for various business scenarios. If the user has any business scenario requirements, he can conveniently call the target model corresponding to the requirement based on the simple and flexible interface parameter configuration options provided by the integration tool, and can also provide the data to be processed to the integration tool based on the simple processing statements compatible with the integration tool. The integration tool can automatically process the data to be processed based on the target model. The user does not need to understand the model principles and related parameters, which reduces the user's code participation in the process of using the model. Only simple configuration and statement input are required to use the trained model in the integration tool, making it possible for users with zero code foundation to easily and accurately use models that meet their needs. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0079] Figure 1 A schematic diagram of the structure of an integrated tool 100 provided in an embodiment of the present application;

[0080] Figure 2 A functional schematic diagram of an integrated tool 100 provided in an embodiment of the present application;

[0081] Figure 3 A schematic diagram of an interface of an integrated tool 100 provided in an embodiment of the present application;

[0082] Figure 4 A flow chart of a model building method provided in an embodiment of the present application;

[0083] Figure 5 A flowchart of a method for using a model provided in an embodiment of the present application;

[0084] Figure 6 A schematic diagram of a process of using an integrated tool provided in an embodiment of the present application;

[0085] Figure 7 A schematic diagram of the structure of a model building device 700 provided in an embodiment of the present application;

[0086] Figure 8 A schematic diagram of the structure of a model using device 800 provided in an embodiment of the present application;

[0087] Fig. 9 A schematic diagram of the structure of an electronic device 900 provided in an embodiment of the present application;

[0088] Fig.10 A schematic diagram of the structure of an integrated tool 1000 provided in an embodiment of the present application. DETAILED DESCRIPTION

[0089] A model is a mapping from input to output that is learned by optimizing an algorithm and determining a learning strategy in a specified hypothesis space. It can also be understood in layman's terms as a function learned from data that can achieve a specific function (mapping). As the core of artificial intelligence and data science, models have become a tool that many industries and scenarios are eager to use. However, the construction and use of models are usually complex and require personnel with model-related expertise, which greatly limits the promotion of models. It is urgent to design user-friendly technical solutions for building and using models.

[0090] In order to provide convenience for non-professionals to build and use machine models, some encapsulated machine learning tools have emerged. However, when people use these encapsulated machine learning tools, there are problems such as multiple process modules, complex interfaces, and single functions. The efficiency of non-professional users using these encapsulated machine learning tools cannot be guaranteed, and non-professional users need to have certain professional knowledge. Specifically, the problems existing in the use of current encapsulated machine learning tools may include: (1) users call modules corresponding to the stages of data set construction, feature processing, model training, evaluation, saving, loading, and prediction according to the different stages of machine learning model training and use. Users participate more in the training and use of machine learning models, and the connection process is prone to errors; (2) users need to modify the data structure and pass in the limited parameters according to the interface requirements. The encapsulated machine learning tools have poor flexibility and inconsistent standards; (3) the machine learning model types provided by the encapsulated machine learning tools are single and cannot meet the user's needs for machine learning models in different business scenarios. When different types of machine learning models are needed, different encapsulated machine learning tools need to be called. The process of modifying the input parameters of different encapsulated machine learning tools in a targeted manner not only requires high professionalism from users, but also consumes a lot of time. In short, the current packaged machine learning tools are not convenient and friendly enough for users.

[0091] Based on this, in order to improve the universality and efficiency of machine learning models, the embodiment of the present application provides an integrated tool, which can solve the problems existing in the current technical solutions (including encapsulated machine learning tools), and the integrated tool has the following characteristics: (1) minimize the user's involvement in the code, so that the user can train and use the zero-code machine learning model without understanding the model principle; (2) provide a concise and flexible interface to avoid users from passing model-related parameters; (3) integrate machine learning model types that can cover multiple business scenarios, and integrate the modeling and use processes of different types of machine learning. In this way, when the user needs to build a certain type of machine learning model, he only needs to determine the machine learning model type, configure the interface parameters, and enter the training statement during the use of the integrated tool, and the integrated tool can automatically generate a trained machine learning model of this type. When the user needs to use a certain type of machine learning model, he only needs to configure the interface parameters and enter the processing statement during the use of the integrated tool, and the integrated tool can automatically call the trained machine learning model of this type to complete the processing of the data to be processed. It can be seen that this integrated tool can meet the different needs of users in different business scenarios and provide users with appropriate machine learning model training and usage services. For users, this comprehensive and easy-to-use integrated tool facilitates users to build the required machine learning models and also facilitates users to use various trained machine learning models.

[0092] In a possible implementation, the embodiment of the present application provides a model building method, in which if a user wants to build a model applicable to a certain business scenario, the user needs to execute: first, input a selection statement on the interface provided by the integrated tool, and the integrated tool can automatically obtain at least one initial model, and the at least one initial model belongs to the target type and the target type corresponds to the selection statement; second, input a training statement and configure interface parameters on the interface provided by the integrated tool, and the integrated tool can automatically obtain a training data set; in this way, the integrated tool can obtain a target model based on the training data set and at least one initial model, and the target model is a trained model belonging to the target type. It can be seen that the integrated tool integrates model types applicable to various business scenarios, and the user can use the integrated tool to build a model that meets the needs of any business scenario. Moreover, the integrated tool provides simple and flexible interface parameter configuration options, and is compatible with simple selection statement and training statement input, without the user understanding the model principle and related parameters, reducing the user's code participation in the process of building the model, and only through simple configuration and statement input, the required model can be built, making it possible for users with zero code foundation to easily and accurately build a model that meets their needs.

[0093] It should be noted that the subject implementing the model building method may be the model building device provided in the embodiment of the present application, and the model building device may include an integrated tool, or the model building device may be connected to the integrated tool. The model building device may be carried in an electronic device or a functional module of an electronic device.

[0094] In another possible implementation, the embodiment of the present application also provides a model usage method, in which if the user wants to use a model of the target type, the user needs to perform the following: first, configure interface parameters on the interface provided by the integration tool, and the integration tool can automatically obtain the target model, and the interface parameters at least include model parameters corresponding to the target model, and the target model is a model that has been trained in the integration tool; second, enter a processing statement on the interface provided by the integration tool, and the integration tool can automatically obtain the data set to be processed, and the processing statement includes the database parameters required to call the data set to be processed; in this way, the integration tool can obtain a processing result based on the target model and the data set to be processed, and the processing result is used to guide the processing strategy for the data set to be processed. In this way, the integration tool can save the training results of the integrated models suitable for various business scenarios. If the user has any business scenario requirements, he can conveniently call the target model corresponding to the requirement based on the simple and flexible interface parameter configuration options provided by the integration tool, and can also provide the data to be processed to the integration tool based on the simple processing statements compatible with the integration tool. The integration tool can automatically process the data to be processed based on the target model. The user does not need to understand the model principles and related parameters, which reduces the user's code participation in the process of using the model. Only simple configuration and statement input are required to use the trained model in the integration tool, making it possible for users with zero code foundation to easily and accurately use models that meet their needs.

[0095] It should be noted that the subject implementing the model use method may be the model use device provided in the embodiment of the present application, and the model use device may include an integrated tool, or the model use device may be connected to the integrated tool. The model use device may be carried in an electronic device or a functional module of an electronic device.

[0096] Before introducing the model building method and model using method provided in the embodiments of the present application, the structure and function of the integrated tool are first explained.

[0097] like Figure 1As shown, the integrated tool 100 may include, for example, an initialization module 110, a machine learning model training module 120, and a processing result output module 130. Among them, the initialization module 110 is used to receive the interface parameters input by the user through the interface 200, and the interface 200 is an interface provided by the integrated tool 100 for interacting with the user. The machine learning model training module 120 may include: a data loading unit 121, a feature processing unit 122, a sample set division unit 123, a model training unit 124, a model evaluation unit 125, and a model preservation unit 126. To call the machine learning model training module 120 in the integrated tool 100, the user only needs to enter an SQL statement for indicating training (referred to as a training statement in the embodiment of the present application) on the interface 200 to trigger the units 121 to 126 in the machine learning model training module 120 to automatically perform model training and obtain a trained machine learning model. The processing result output module 130 may include: a model loading unit 131, an algorithm prediction unit 132, and an output result unit 133. To call the processing result output module 130 in the integrated tool 100, the user only needs to input an SQL statement (referred to as a processing statement in the embodiment of the present application) for indicating processing on the interface 200 to trigger the units 131 to 133 in the processing result output module 130 to automatically call and use the model, and obtain the processing results of the called machine learning model on the data to be processed.

[0098] It should be noted that the integration tool 100 can integrate various types of machine learning models. Specifically, it can integrate multiple machine learning models commonly used in the industry or field targeted by the integration tool 100. For example, in the e-commerce industry, there are usually requirements for target prediction, ranking recommendation, and label precipitation. Figure 2 As shown, the integrated tool 100 suitable for the e-commerce industry may include: a regression model 1, a classification model 2 and an attribution model 3, wherein the regression model 1 is used to achieve target prediction, the classification model 2 is used to achieve ranking recommendation, and the attribution model 3 is used to achieve interpretable label precipitation.

[0099] The integrated tool 100 can be used alone in the process of model building, then the user can use the initialization module 110 and the machine learning model training module 120 in the integrated tool 100, and the trained target model is saved by the integrated tool 100, which is convenient for the subsequent direct use of the target model. Alternatively, the integrated tool 100 can also be used alone in the process of model use, then the user can use the initialization module 110 and the processing result output module 130 in the integrated tool 100, and the used, trained target model can be directly called from the trained model saved by the integrated tool 100, which is convenient for the user to use the target model. Alternatively, the integrated tool 100 can also be used in the process of model building + model use, then the user can use the initialization module 110, the machine learning model training module 120 and the processing result output module 130 in the integrated tool 100, and directly use the target model to be processed after training. In this process, in order to facilitate the subsequent user's call to the target model, the integrated tool 100 will also save the target model.

[0100] It should be noted that after the integrated tool 100 completes a training of a certain model, it will save the trained target model, which can be used by the user all the time. Alternatively, the integrated tool 100 can also set a usage period (such as one month) for the saved target model. When the target model is used for a period of time, an updated target model can be obtained by retraining. In this way, by regularly updating the target model, the accuracy of the target model can be improved to a certain extent. The user can obtain more accurate processing results by using the target model, which improves the user's experience of using the target model, and further improves the user's experience of using the integrated tool 100.

[0101] It can be seen that the integrated tool 100 includes processes such as data processing, model training, model evaluation, model preservation, model loading, model prediction and result output. The standardized module design enables the integrated tool 100 to realize the integrated encapsulation of the construction and use processes of multiple types of machine learning models, reducing the user's participation in the intermediate process of model construction and use.

[0102] In addition, the integrated tool 100 is also designed with a modular interface, and the user can configure interface parameters on the interface 200 provided by the integrated tool 100. According to the process of building and / or using the machine learning model, the interface parameters may include at least one of the following parameters: input parameters, model parameters, and output parameters, wherein the input parameters are used to indicate the input rules of the training samples, the output parameters are used to indicate the output rules of the model, and the model parameters are used to indicate the saving rules of the model. For example, see Figure 3As shown, the configuration interface for configuring interface parameters on the interface 200 can be designed in the form of a table, wherein the input parameters, model parameters and output parameters correspond to the input table, model table and output table respectively. The input table may include, for example: the primary key field name, the target field name, the sample set division rule and the target value processing method, wherein the primary key field name may carry, for example, the merchant identifier (Identifier, ID) or the product ID, etc.; the target field name may carry, for example, the sales volume or the number of visits, etc.; the sample set division rule may include a proportional random division rule or a partition division rule; the target value processing method may indicate that the target field value is processed through function mapping according to the discrete degree of the target field value. The model table may include, for example: the custom model name and the save path of the custom model. The output table may include, for example: primary key field name, processing result field name, algorithm field name, output table name and business time, wherein the primary key field name and processing result field name correspond to the primary key field name and target field name in the input table respectively, the algorithm field name is used to carry the algorithm ID corresponding to the trained target model, the output table name is used to indicate the name of the table obtained after the output data is filled into the output table, and the business date may indicate the time of the obtained table. It should be noted that the initialization parameters on the interface 200 provided by the integration tool 100 may adopt the default values ​​for parameter transmission. Figure 3 The interface design shown is designed for business considerations on the one hand, and flexibility and simplicity on the other hand. The primary key of the input table is not determined by a single field in the business. The interface 200 provides an interface for inputting all fields for determining the primary key, which are automatically spliced ​​and mapped to the primary key of the output table by the processing result output module 130. In addition, users can also flexibly select sample set division rules and target value processing methods according to their needs, which improves the user experience brought to users by the integrated tool 100.

[0103] It can be seen that, for example Figure 3 The design of the configuration interface on the interface 200 shown provides the user with a simple, convenient and flexible interface. The user can easily use the interface 200 provided by the integrated tool 100 to complete the configuration of the interface parameters, making it more convenient and faster to build and use the model.

[0104] The following is an introduction to a model building method provided in an embodiment of the present application.

[0105] Figure 4 A schematic diagram of a model building method provided in an embodiment of the present application. The method can be applied to a model building device, which can be, for example, Figure 7 The model building device 700 shown in FIG. 1 may include an integrated tool, or the model building device may be connected to the integrated tool, or the method may be applied to the integrated tool. The relevant description of the integrated tool can be found in the above Figure 1 to Figure 3 Related description.

[0106] like Figure 4 As shown, the method may include, for example, the following S101 to S103:

[0107] S101, in response to a selection statement input by a user on an interface provided by an integration tool, obtaining at least one initial model belonging to a target type, wherein the target type corresponds to the selection statement.

[0108] It is understandable that the integrated tool includes multiple types of models, and each type of model corresponds to at least one algorithm belonging to this type. For example, the model type that the integrated model may include is at least one of the following types: attribution model, regression model or classification model, the attribution model may correspond to at least one of the following algorithms: occlusion-based attribution algorithm, layer-by-layer back-propagation-based attribution algorithm and gradient-based attribution algorithm, the regression model may correspond to at least one of the following algorithms: eXtreme Gradient Boosting (XGboost), Generalized Linear Model (GLM), Gradient Boosting Decision Tree (GBDT) or Distributed Random Forest (DRF), and the classification model may correspond to at least one of the following algorithms: Bayesian algorithm, decision tree algorithm, K-Nearest Neighbor (KNN) algorithm.

[0109] In a possible implementation, the interface provided by the integration tool has multiple selection statements for users to copy, each selection statement corresponds to a model type, when the user copies a selection statement on the interface provided by the integration tool and pastes the copied selection statement into the input position of "Select Model Type", the model type corresponding to the selection statement is the target type, and the integration tool determines that the user needs to build a target model of the target type. In this way, the user does not need to write code to "tell" the integration tool what type of model he wants to build, but only needs to select a selection statement and perform simple operations such as copy + paste to let the integration tool determine the type of model to be built. The integration tool is very user-friendly.

[0110] After determining the target type, the integration tool can match at least one initial model belonging to the target type from the integrated multiple initial models. Each of the at least one initial model can participate in the model building process to obtain the corresponding trained model, and then determine the target model from the model trained by the initial model.

[0111] For example, when the selection statement input by the user indicates that the target type corresponds to a classification model, the integration tool can obtain an initial model belonging to the classification model based on S101, which may include at least one of the following models: Bayesian algorithm, decision tree algorithm, KNN algorithm.

[0112] It can be seen that through S101, the integrated tool can determine the initial model to be trained in the model building process with the help of the user's input of the selection statement on the interface, and prepare for the model building.

[0113] S102, obtaining a training data set in response to the training statements and configured interface parameters input by the user on the interface provided by the integrated tool.

[0114] The training statements can be statements that users can learn and write, such as SQL statements. The training statements are used to indicate the storage location of the data set involved in the training. Based on the instructions of the training statements, the integrated tool can obtain the data set involved in the training. In this way, users do not need to understand complex programming knowledge, but only need to be able to enter training statements such as SQL on the interface provided by the integrated tool, which reduces the threshold for users to build models.

[0115] Interface parameters may include at least input parameters, for example, see Figure 3 In the input table shown, the input parameters are used to indicate the input rules of the training samples. The input parameters may include, for example: primary key field name, target field name, sample set division rule and target value processing method.

[0116] As an example, S102 may include: in response to a training statement input by a user on an interface provided by the integrated tool, obtaining a first data set; in response to an input parameter input by a user on the interface provided by the integrated tool, processing the first data set to obtain a training data set that complies with the input rules indicated by the input parameters.

[0117] Corresponding to Figure 1 In the structure of the integrated tool 100 shown, S102 may correspond to the following S1021 to S1023:

[0118] S1021, the data loading unit 121 performs data loading. For example, in an e-commerce scenario, in order to load massive e-commerce big data, the Spark computing engine is used to read the database table and automatically convert the data into a data type that can be read by the initial model.

[0119] S1022, the feature processing unit 122 determines the data type in the first data set that is successfully read, aligns features, and performs automated feature engineering.

[0120] Among them, the interpretation of the data type in the first data set may include: if the data type is numerical, data cleaning is performed; if the data type is non-numerical, data cleaning is performed first, and then processed by label encoding (LabelEncoding). Among them, data cleaning may include, for example: filling the missing values ​​of the corresponding features with numerical fixed values, and / or eliminating non-demand features. Considering that the features used for machine learning models in e-commerce business are usually high-dimensional, the encoding method can mainly be One-Hot Encoding, which generates a large number of sparse matrices to expand the feature dimension. Moreover, the machine learning model in the embodiment of the present application can be mainly established based on a tree model, which is insensitive to the order between feature values.

[0121] Aligning features and performing automated feature engineering can be understood as automatically reorganizing the data order in the first data set to avoid different feature orders between the training set and the prediction set and to make the feature orders of the training set and the prediction set consistent.

[0122] S1023, the sample set division unit 123 divides the features processed in S1022 into sample sets.

[0123] In order to evaluate the effect of the trained model, it is necessary to divide a part of the data from the training data set (also called the training sample set) for testing. In the embodiment of the present application, two sample set division modes are designed for business scenarios: proportional random division rules and partition division rules. Among them, since the classification and attribution scenarios focus on exploring the logical relationship between input features and output features, and do not pay attention to feature changes in the time dimension, the proportional random division rule is mainly used in classification and attribution scenarios; since the regression scenario focuses on exploring the relationship between features in a certain time period in the current or history and target features in a certain time period in the future, the partition division rule is mainly used in regression scenarios. The user can select the sample set division rule in the configuration interface parameters on the interface 200 provided by the integrated tool 100, enter the specific sample set division rule, and support the implementation of S1023.

[0124] It can be seen that through S102, the integrated tool can determine the training data set with the help of a small amount of input from the user on the interface, thereby providing a data basis for this model construction.

[0125] S103, obtaining a target model based on the training data set and the at least one initial model, wherein the target model is a trained model belonging to the target type.

[0126] In some possible implementations, S103 may, for example, include: the integration tool trains at least one initial model based on the training data set to obtain a target model, where the target model is a trained model of the target type, and the target model may be any one of the following types of models integrated in the integration tool: an attribution model, a regression model, or a classification model.

[0127] As an example, S103 may include: training each of the at least one initial model based on the training data set to obtain at least one candidate model, the at least one candidate model corresponding to the at least one initial model one-to-one; obtaining an evaluation result of each of the at least one candidate model, the evaluation result is used to indicate the effect of the corresponding candidate model in performing machine learning; based on the evaluation result, selecting the target model from the at least one candidate model. Wherein, in the process of training each of the at least one initial model, the initial model can be trained based on the target value processing method in the input parameter, and the target value processing method can indicate: according to the discrete degree of the target field value, the target field value (i.e., the target value) is processed through function mapping.

[0128] It should be noted that the interface parameters may also include output parameters and model parameters, the output parameters are used to indicate the output rules of the target model, and the model parameters are used to indicate the saving rules of the target model. After S103, the method may further include: saving the model file corresponding to the target model according to the output rules and the saving rules. This is to prepare for the subsequent calling of the target model, so that the user can use the target model later.

[0129] Corresponding to Figure 1 In the structure of the integrated tool 100 shown, S103 may correspond to the following S1031 to S1033:

[0130] S1031, the model training unit 124 automatically performs training and tuning of the machine learning model, which may include: automatic processing of target fields, reading of features and target values, model training and automatic parameter tuning.

[0131] According to the instructions of the target value processing method in the input parameters, the target value is processed by function mapping according to the discrete degree of the target field value. In order to build an algorithm model with better effect, the hyperparameters of the machine learning model can be automatically tuned by random grid search in the embodiment of the present application, that is, by randomly selecting a certain number of hyperparameter combinations in the hyperparameter space for evaluation, and selecting the hyperparameter combination with the best evaluation result, the advantage is that it is possible to quickly find a better machine learning model hyperparameter combination suitable for the current data. Among them, hyperparameters can be understood as model parameters such as nodes, branches, number of branches, probabilities, weights, etc. of the model. In addition, the embodiment of the present application can also give default values ​​for hyperparameter initialization, so that users can use it without understanding the relevant knowledge of the machine learning model.

[0132] S1032, the model evaluation unit 125 evaluates the trained machine learning model (i.e., the candidate model mentioned above), and may use a test sample set independent of the training data set to perform offline model evaluation and output the evaluation result.

[0133] Among them, the evaluation indicators may include but are not limited to: mean square error (MSE), area under the receiver operating characteristic curve (ROC) and the coordinate axis (Area Under Curve, AUC) or logarithmic loss value (LogLoss). Batch training and fitting are performed for various models, and the test results are compared. The evaluation indicators of each model may include at least one of MSE, AUC or LogLoss. When the number of evaluation indicators is greater than 1, the values ​​of multiple evaluation indicators can also be weighted averaged, added, averaged, etc., to obtain the corresponding evaluation score or evaluation grade as the final evaluation result. The evaluation result can be recorded in the log. The integrated tool 100 can determine the best candidate model based on the evaluation results of each candidate model, and save it as the target model for subsequent use.

[0134] S1033, after completing the model evaluation in S1032, the model saving unit 126 saves the target model.

[0135] For example, the model saving unit 126 can serialize the target model, upload it to the target location according to the model file named according to the model name defined by the model parameters in the interface parameters input by the user, and define the model saving path according to the model parameters. The target location can be, for example, a storage location corresponding to the saving path of the model in the distributed object storage system.

[0136] It can be seen that through S103, the integrated tool can automatically perform operations such as training, testing and evaluation on at least one initial model based on the training data set to obtain the target model, that is, the result of this model construction.

[0137] In some implementations, if the user needs to directly use the target model after obtaining the target model, the integrated tool can automatically obtain the data set to be processed by only requiring the user to enter a processing statement on the interface provided by the integrated tool, and obtain the processing result according to the target model and the data set to be processed. For the specific implementation of the process of using the target model after S103, see the relevant description of S202 to S203 below.

[0138] It can be seen that in this method, by integrating model types suitable for various business scenarios, users can use the integrated tool to build a model that meets any business scenario requirements. Moreover, the integrated tool provides simple and flexible interface parameter configuration options, and is compatible with simple selection statements and training statement inputs. Users do not need to understand the model principles and related parameters, which reduces the user's code participation in the model building process. Only simple configuration and statement input are required to build the required model, making it possible for users with zero code foundation to easily and accurately build models that meet their needs.

[0139] The following is an introduction to a method for using a model provided in an embodiment of the present application.

[0140] Figure 5 A flow chart of a model use method provided in an embodiment of the present application. The method can be applied to a model use device, which can be used, for example, to Figure 8 The model using device 800 is shown. The model using device may include an integrated tool, or the model using device may be connected to the integrated tool, or it may be understood that the method is applied to the integrated tool. The relevant description of the integrated tool can be found in the above Figure 1 to Figure 3 Related description.

[0141] like Figure 5 As shown, the method may include, for example, the following S201 to S203:

[0142] S201, in response to interface parameters configured by a user on an interface provided by an integration tool, a target model is obtained, wherein the interface parameters include model parameters corresponding to the target model, and the target model is a model that has been trained in the integration tool.

[0143] The target model is any one of the following types of models integrated in the integration tool: an attribution model, a regression model or a classification model.

[0144] When the user wants to use a target type of machine learning model to analyze and process data, the user can perform relevant operations on the interface provided by the integrated tool to obtain the target model.

[0145] As an example, the user may first train a model of the target type, and record the trained model as the target model. S201 may include: S201a1, in response to a selection statement input by the user on the interface provided by the integrated tool, obtaining at least one initial model, wherein the initial model is of the same type as the target model; S201a2, in response to a training statement and configured interface parameters input by the user on the interface provided by the integrated tool, obtaining a training data set; S201a3, obtaining the target model based on the training data set and the at least one initial model. It should be noted that the specific implementation of S201a1 to S201a3 can be found in the above. Figure 4 Relevant description of S101 to S103 in the method shown.

[0146] As another example, if the target model is pre-trained and saved in the integrated tool, the model parameters may include the name of the target model and the save path of the target model, then S201 may include: S201b1, obtaining the model file corresponding to the name of the target model from the save path of the target model; S201b2, obtaining the target model according to the model file.

[0147] In this example, it corresponds to Figure 1 The structure of the integrated tool 100 shown, S201 may correspond to: the model loading unit 131 loads the target model, obtains a model file named with a custom model name from a user-specified saving path, deserializes the model file, and extracts the target model.

[0148] It should be noted that in order to prevent the program from reporting an error due to an error in the user-defined save path, or the non-existence of the file specified in the save path, in the embodiment of the present application, it is possible to first determine whether there is a file named with the custom model name in the user-defined save path. If not, a prompt message is generated. The prompt message is used to prompt the user to rebuild the model of the target type and save it to the custom save path with the custom model name.

[0149] It can be seen that obtaining the target model to be used through S201 provides a processing tool for subsequent processing (such as prediction) of the data to be processed.

[0150] S202, in response to a processing statement input by the user on the interface, obtaining a data set to be processed, wherein the processing statement includes database parameters required for calling the data set to be processed.

[0151] The processing statement can be a statement that is easy for the user to learn and write, such as an SQL statement. The processing statement is used to indicate the storage location of the data set to be processed. Based on the instructions of the processing statement, the integrated tool can obtain the data set to be processed. In this way, the user does not need to understand complex programming knowledge, but only needs to be able to enter processing statements such as SQL on the interface provided by the integrated tool, which reduces the threshold for users to use the model.

[0152] Corresponding to Figure 1 The structure of the integrated tool 100 shown, taking the data set to be processed as a prediction data set as an example, S202 may correspond to: the algorithm prediction unit 132 reads the data set to be predicted from the database, processes the data to be predicted into a resilient distributed data set (RDD) by calling the data loading unit 121, and automatically calculates the number of partitions for distributed prediction execution based on the data to be predicted, and, by calling the feature processing unit 122, processes the data to be predicted processed by the data loading unit 121 for consistency with the training data set, thereby obtaining the data set to be predicted.

[0153] S203: Obtain a processing result according to the target model and the data set to be processed, wherein the processing result is used to guide a processing strategy for the data set to be processed.

[0154] In some implementations, S203 may include: the integration tool automatically inputs the data set to be processed into the target model, and the target model outputs the processing result. For example, if the target model is a classification model, then the processing result includes the classification result of the data set to be processed, and the processing result may also include a ranking recommendation generated based on the data set to be processed; for another example, if the target model is a regression model, then the processing result includes the prediction result of the data set to be processed; for another example, if the target model is an attribution model, then the processing result includes the label precipitation result of the interpretability of the data set to be processed.

[0155] Corresponding to Figure 1 The structure of the integrated tool 100 shown, taking the data set to be processed as a prediction data set as an example, S203 may correspond to: the algorithm prediction unit 132 performs algorithm prediction, for example, adopts a distributed prediction method to achieve rapid prediction of the data set to be predicted (such as large sample data); after the prediction is completed, the output result unit 133 inputs the prediction result into a database form with a user-specified name (such as the output table indicated by the output parameter). For example, the user-defined input table name and field name are automatically spliced ​​into an SQL statement, and the output result is stored in a standardized format.

[0156] For some data tables where the primary key is determined by multiple fields, the integrated tool can automatically identify and concatenate the primary keys of the input table to ensure the consistency of the primary keys of the output table obtained by using the target model and the input table used to train the target model, and determine the primary key of the output table based on the results of identification and concatenation. In addition, the integrated tool can also determine the target value processing method in the training process when building the model, and perform corresponding inverse function mapping on the processing results (i.e., target values).

[0157] For the regression model, the processing result (i.e., the target value) can be the standard prediction value of the business indicator; for the classification model, the processing result (i.e., the target value) can be a value between 0 and 1, which can be used as the recommendation weight for the sorting business scenario; for the attribution model, the processing result (i.e., the target value) can be the importance evaluation value of each single feature or feature combination, which can be used as a reference for feature selection.

[0158] It can be seen that in this method, the integration tool can save the training results of the integrated models suitable for various business scenarios. If the user has any business scenario requirements, he can conveniently call the target model corresponding to the requirement based on the simple and flexible interface parameter configuration options provided by the integration tool, and can also provide the data to be processed to the integration tool based on the simple processing statements compatible with the integration tool. The integration tool can automatically process the data to be processed based on the target model, without the need for the user to understand the model principles and related parameters, reducing the user's code participation in the process of using the model. Only simple configuration and statement input are required to use the trained model in the integration tool, making it possible for users with zero code foundation to easily and accurately use models that meet their needs.

[0159] In order to make the method provided in the embodiment of the present application clearer, the method is described below in conjunction with the process of a user using an integrated tool.

[0160] like Figure 6 As shown, when the user constructs the target model, the process may include the following S1 to S7, the process of using the target model may include the following S1 to S9, or the process of using the target model may include the following S5, S6, S8 and S9:

[0161] S1, configure environment variables;

[0162] S2, get the Docker image;

[0163] S3, importing an algorithm package corresponding to at least one initial model of the target type;

[0164] S4, input training SQL statement;

[0165] S5, configure interface parameters;

[0166] S6, call the initialization function;

[0167] S7, calling the model training function;

[0168] S8, input processing SQL statement;

[0169] S9, call the processing output function.

[0170] Among them, S6, S7 and S9 are respectively equivalent to the integration tool 100 calling the initialization module 110, the machine learning model training module 120 and the processing result output module 130.

[0171] Among them, S4, S5 and S8 are equivalent to the steps on the integrated tool that require user-defined input ( Figure 6 The other steps can be understood as standardized steps using integrated tools.

[0172] It can be seen that based on the integrated tool designed in the embodiment of the present application, the user only needs to call a few statements and execute three customized steps to use the integrated tool to build and use any desired model, which provides convenience for users to build and use machine learning models and makes the promotion and application of machine learning models possible.

[0173] Accordingly, the present embodiment also provides a model building device 700, see Figure 7 As shown, the apparatus 700 may include:

[0174] A first obtaining unit 701 is configured to obtain at least one initial model belonging to a target type in response to a selection statement input by a user on an interface provided by the integration tool, wherein the target type corresponds to the selection statement;

[0175] A second obtaining unit 702, configured to obtain a training data set in response to a training statement and configured interface parameters input by the user on an interface provided by the integrated tool;

[0176] The third obtaining unit 703 is used to obtain a target model based on the training data set and the at least one initial model, where the target model is a trained model belonging to the target type.

[0177] In a possible implementation, the interface parameter includes an input parameter, and the input parameter is used to indicate an input rule of a training sample. The second obtaining unit 702 includes:

[0178] A first obtaining subunit, configured to obtain a first data set in response to a training sentence input by the user on an interface provided by the integrated tool;

[0179] The second obtaining subunit is used to process the first data set in response to the input parameters input by the user on the interface provided by the integrated tool to obtain the training data set, where the training data set complies with the input rules.

[0180] As an example, the interface parameters further include output parameters and model parameters, the output parameters are used to indicate the output rule of the target model, and the model parameters are used to indicate the saving rule of the target model. The apparatus 700 may further include:

[0181] A saving unit is used to save the model file corresponding to the target model according to the output rule and the saving rule.

[0182] In a possible implementation, the third obtaining unit 703 includes:

[0183] A training subunit, configured to train each of the at least one initial model based on the training data set to obtain at least one candidate model, wherein the at least one candidate model corresponds one-to-one to the at least one initial model;

[0184] An evaluation subunit, used to obtain an evaluation result of each candidate model in the at least one candidate model, wherein the evaluation result is used to indicate an effect of executing machine learning of the corresponding candidate model;

[0185] A selection subunit is used to select the target model from the at least one candidate model based on the evaluation result.

[0186] In one possible implementation, the input parameters in the interface parameters include a sample set division rule and a target value processing method, the sample set division rule includes random division by proportion or division by partition, and the target value processing method indicates: according to the discrete degree of the target value field value, the target value field value is processed through function mapping.

[0187] The target model is any one of the following types of models integrated in the integration tool:

[0188] Attribution, regression, or classification models.

[0189] In a possible implementation, the training statement is a SQL statement.

[0190] It should be noted that the specific implementation method of the device 700 and the technical effects achieved can be found in Figure 4 Description of the method shown.

[0191] Accordingly, the present application embodiment also provides a model using device 800, see Figure 8As shown, the apparatus 800 may include:

[0192] A first obtaining unit 801 is used to obtain a target model in response to interface parameters configured by a user on an interface provided by an integration tool, wherein the interface parameters include model parameters corresponding to the target model;

[0193] A second obtaining unit 802 is used to obtain a data set to be processed in response to a processing statement input by the user on the interface, wherein the processing statement includes database parameters required for calling the data set to be processed;

[0194] The third obtaining unit 803 is used to obtain a processing result according to the target model and the data set to be processed, wherein the processing result is used to guide a processing strategy for the data set to be processed.

[0195] In a possible implementation, the interface parameters further include input parameters and output parameters corresponding to the target model, and the first obtaining unit 801 includes:

[0196] A first obtaining subunit, configured to obtain at least one initial model in response to a selection statement input by a user on an interface provided by the integration tool, wherein the initial model is of the same type as the target model;

[0197] A second obtaining subunit, for obtaining a training data set in response to a training statement and configured interface parameters input by a user on an interface provided by the integrated tool;

[0198] The third obtaining subunit is used to obtain the target model according to the training data set and the at least one initial model.

[0199] In another possible implementation, the model parameters include the name of the target model and the saving path of the target model, and the first obtaining unit 801 includes:

[0200] A fourth obtaining subunit is used to obtain a model file corresponding to the name of the target model from a saving path of the target model;

[0201] The fifth obtaining subunit is used to obtain the target model according to the model file.

[0202] The target model may be any of the following types of models integrated in the integration tool:

[0203] Attribution, regression, or classification models.

[0204] In a possible implementation, the processing statement is a SQL statement.

[0205] It should be noted that the specific implementation method of the device 800 and the technical effects achieved can be found in Figure 5 Description of the method shown.

[0206] In addition, an embodiment of the present application also provides an electronic device, which includes a processor and a memory: the memory is used to store instructions or computer programs; the processor is used to execute the instructions or computer programs in the memory, so that the electronic device executes any implementation of the comment display method provided in the embodiment of the present application.

[0207] See also Fig. 9 , which shows a schematic diagram of the structure of an electronic device 900 suitable for implementing the embodiment of the present disclosure. The terminal device in the embodiment of the present disclosure may include but is not limited to mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle terminals (such as vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Fig. 9 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0208] like Fig. 9 As shown, the electronic device 900 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 901, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 902 or a program loaded from a storage device 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the electronic device 900 are also stored. The processing device 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0209] Typically, the following devices may be connected to the I / O interface 905: an input device 906 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 907 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 908 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 909. The communication device 909 may allow the electronic device 900 to communicate with other devices wirelessly or by wire to exchange data. Although Fig. 9 The electronic device 900 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0210] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device 909, or installed from a storage device 908, or installed from a ROM 902. When the computer program is executed by the processing device 901, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0211] The electronic device provided by the embodiment of the present disclosure and the method provided by the above embodiment belong to the same inventive concept, and the technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0212] The present application embodiment also provides an integrated tool 1000, see Fig.10 The integrated tool 1000 may include a model building unit 1001 and a model using unit 1002, wherein the model building unit 1001 is used to execute Figure 4 The method shown in the figure, the model using unit 1002 is used to execute Figure 5 The method shown.

[0213] It should be noted that the model building unit 1001 may correspond to the above-mentioned model building device 700 , and the model using unit 1002 may correspond to the above-mentioned model using device 800 .

[0214] The embodiment of the present application also provides a computer-readable medium, in which instructions or computer programs are stored. When the instructions or computer programs are executed on a device, the device executes any implementation of the comment display method provided in the embodiment of the present application.

[0215] It should be noted that the computer-readable medium disclosed above may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, in which a computer-readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0216] In some embodiments, the client and the server may communicate using any currently known or future developed network protocol such as HTTP (Hyper Text Transfer Protocol), and may be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0217] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.

[0218] The computer-readable medium carries one or more programs, and when the one or more programs are executed by the electronic device, the electronic device can execute the method.

[0219] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages ​​or a combination thereof, including, but not limited to, object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0220] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0221] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware, wherein the name of a unit / module does not, in some cases, constitute a limitation on the unit itself.

[0222] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.

[0223] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0224] It should be noted that the various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the system or device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0225] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0226] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0227] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0228] It should be noted that in the embodiments of the present application, user sensitive information is not involved, and user-related information is obtained, used and determined after user authorization. In one example, before obtaining user-related information, the corresponding interface displays prompt information related to obtaining data use authorization. The prompt information is based on relevant laws and regulations in an appropriate manner to inform the user of the type, scope of use, usage scenarios, etc. of the personal information involved in this disclosure, so that the user can determine whether to agree to the authorization based on the prompt information. It can be understood that the above notification and the process of obtaining user authorization are only illustrative and do not constitute a limitation on the implementation method of the present disclosure. Other methods that meet relevant laws and regulations may also be applied to the implementation method of the present disclosure.

[0229] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for using a model, characterized in that: include: In response to interface parameters configured by a user on an interface provided by the integration tool, a target model is obtained, wherein the interface parameters include model parameters corresponding to the target model, and the target model is a model trained in the integration tool; In response to a processing statement input by the user on the interface, obtaining a data set to be processed, wherein the processing statement includes database parameters required for calling the data set to be processed; A processing result is obtained according to the target model and the data set to be processed, and the processing result is used to guide a processing strategy for the data set to be processed.

2. The method according to claim 1, characterized in that The interface parameters also include input parameters and output parameters corresponding to the target model, and obtaining the target model includes: In response to a selection statement input by a user on an interface provided by the integration tool, obtaining at least one initial model, the initial model being of the same type as the target model; In response to the training statements and configured interface parameters input by the user on the interface provided by the integrated tool, a training data set is obtained; The target model is obtained according to the training data set and the at least one initial model.

3. The method according to claim 1, characterized in that The model parameters include the name of the target model and the saving path of the target model, and obtaining the target model includes: Acquire a model file corresponding to the name of the target model from the saving path of the target model; The target model is obtained according to the model file.

4. A model building method, characterized in that: include: In response to a selection statement input by a user on an interface provided by the integration tool, obtaining at least one initial model belonging to a target type, the target type corresponding to the selection statement; In response to the training statements and configured interface parameters input by the user on the interface provided by the integrated tool, obtaining a training data set; Based on the training data set and the at least one initial model, a target model is obtained, where the target model is a trained model belonging to the target type.

5. The method according to claim 4, characterized in that The interface parameters include input parameters, and the input parameters are used to indicate input rules of training samples. The step of obtaining a training data set includes: In response to a training statement input by the user on an interface provided by the integrated tool, obtaining a first data set; In response to input parameters input by the user on the interface provided by the integrated tool, the first data set is processed to obtain the training data set, and the training data set complies with the input rules.

6. The method according to claim 5, characterized in that The interface parameters further include output parameters and model parameters, the output parameters are used to indicate output rules of the target model, and the model parameters are used to indicate storage rules of the target model. The method further includes: According to the output rule and the saving rule, the model file corresponding to the target model is saved.

7. The method according to claim 4, characterized in that The step of obtaining a target model based on the training data set and the at least one initial model comprises: Training each of the at least one initial model based on the training data set to obtain at least one candidate model, where the at least one candidate model corresponds one-to-one to the at least one initial model; Obtaining an evaluation result of each candidate model in the at least one candidate model, wherein the evaluation result is used to indicate an effect of executing machine learning of the corresponding candidate model; Based on the evaluation result, the target model is selected from the at least one candidate model.

8. The method according to any one of claims 4 to 7, characterized in that: The input parameters in the interface parameters include sample set division rules and target value processing methods. The sample set division rules include random division by proportion or division by partition. The target value processing method indicates: according to the discrete degree of the target field value, the target field value is processed through function mapping.

9. The method according to any one of claims 1 to 8, characterized in that: The target model is any one of the following types of models integrated in the integration tool: Attribution, regression, or classification models.

10. A model using device, characterized in that: include: A first obtaining unit, configured to obtain a target model in response to interface parameters configured by a user on an interface provided by the integration tool, wherein the interface parameters include model parameters corresponding to the target model, and the target model is a model trained in the integration tool; A second obtaining unit, configured to obtain a data set to be processed in response to a processing statement input by the user on the interface, wherein the processing statement includes database parameters required for calling the data set to be processed; The third obtaining unit is used to obtain a processing result according to the target model and the data set to be processed, wherein the processing result is used to guide a processing strategy for the data set to be processed.

11. A model building device, characterized in that: include: A first obtaining unit, configured to obtain, in response to a selection statement input by a user on an interface provided by the integration tool, at least one initial model belonging to a target type, the target type corresponding to the selection statement; A second obtaining unit, configured to obtain a training data set in response to a training statement and configured interface parameters input by the user on an interface provided by the integrated tool; The third obtaining unit is used to obtain a target model based on the training data set and the at least one initial model, where the target model is a trained model belonging to the target type.

12. An electronic device, characterized in that: The electronic device comprises: a processor and a memory; The memory is used to store instructions or programs; The processor is used to execute the instructions or programs in the memory so that the electronic device executes the method according to any one of claims 1 to 9.

13. A readable medium, characterized in that: The readable medium stores instructions or programs, and when the instructions or programs are executed on a processor, the processor is enabled to execute the method according to any one of claims 1 to 9.

Images