A method, device, electronic device and storage medium for applying machine learning

Abstract

The present disclosure relates to a method, device, electronic device and storage medium for applying machine learning, directly connecting to business scenarios, accumulating data related to business scenarios and then exploring model solutions, ensuring that the data used for offline model solution exploration and the data used for the model online estimation service are of the same source, and achieving the homology of offline and online data. In order to avoid the problem of poor estimation effect due to inconsistency between the data obtained by online feature calculation and offline feature calculation after the offline model is directly deployed online, only the model solution is deployed online, and the offline model is not deployed online. After the model solution is deployed online, receiving the estimation request can obtain sample data with features and feedback, and then the sample data can be used for model self-learning. The model obtained by self-learning can be deployed online, ensuring that the data and feature engineering solutions used for model self-learning are respectively consistent with the data and feature engineering solutions used for the model online estimation service, and achieving consistency between the model self-learning effect and the model estimation effect.

Description

Technical Field

[0001] The embodiments of the present disclosure relate to the technical field of machine learning, and in particular to a method, device, electronic device, and storage medium for applying machine learning. Background Art

[0002] The application of machine learning may include but is not limited to: problem definition, machine learning model establishment (modeling for short), model online service, feedback information collection and model iterative update. At present, modeling is based on offline data exploration model, and then the model effect is determined based on offline evaluation. After the model effect meets the standard (that is, meets the preset requirements), IT personnel will deploy the model online and provide model online service.

[0003] However, the model that meets the offline performance requirements may not meet the online performance requirements. The inventors of the present application found that this is because the data used for modeling is inconsistent with the online data, and it is difficult to ensure the consistency of the features calculated during the modeling process. As a result, the model effects are very different online and offline, which does not meet expectations, making it difficult to put the model online.

[0004] The above description of the process of discovering the problem is only used to assist in understanding the technical solution of the present disclosure and does not constitute an admission that the above content is prior art. Summary of the invention

[0005] In order to solve at least one problem existing in the prior art, at least one embodiment of the present disclosure provides a method, device, electronic device and storage medium for applying machine learning.

[0006] In a first aspect, an embodiment of the present disclosure provides a method for applying machine learning, the method comprising:

[0007] Obtain relevant data streams for specified business scenarios online based on the data service interface;

[0008] accumulating data in the related data stream into a first database;

[0009] When the first preset condition is met, exploring a model solution based on the data in the first database; the model solution includes the following solution sub-items: feature engineering solution, model algorithm and model hyperparameters;

[0010] The model solution obtained through exploration is deployed online to provide a model online estimation service, wherein the model online estimation service is performed based on the relevant data stream of the specified business scenario obtained online by the data service interface.

[0011] In a second aspect, an embodiment of the present disclosure provides a device for applying machine learning, the device comprising:

[0012] A data management module, configured to obtain online relevant data streams of a specified business scenario based on a data service interface; and accumulate data in the relevant data streams into a first database;

[0013] A model solution exploration module, used for exploring a model solution based on the data in the first database when a first preset condition is met; the model solution includes the following solution sub-items: a feature engineering solution, a model algorithm, and a model hyperparameter;

[0014] The model online estimation service module is used to deploy the model solution obtained by the model solution exploration module online to provide a model online estimation service, wherein the model online estimation service is performed based on the relevant data flow of the specified business scenario obtained online by the data service interface.

[0015] In a third aspect, an embodiment of the present disclosure proposes an electronic device, comprising: a processor and a memory; the processor is used to execute the steps of the method of applying machine learning as described in the first aspect by calling a program or instruction stored in the memory.

[0016] In a fourth aspect, an embodiment of the present disclosure proposes a non-transitory computer-readable storage medium for storing a program or instruction, wherein the program or instruction enables a computer to execute the steps of the method for applying machine learning as described in the first aspect.

[0017] It can be seen that in at least one embodiment of the present disclosure, the business scenario is directly connected, the data related to the business scenario is accumulated and then the model solution is explored, the model solution and the offline model are obtained, and the data used for the offline model solution exploration and the data used for the model online estimation service are of the same source, so as to achieve the homology of offline and online data. In order to avoid the problem that the estimation effect of the deployed offline model is poor due to the inconsistency between the data obtained by online feature calculation and the data obtained by offline feature calculation after the offline model is directly deployed online, only the model solution is deployed online, and the offline model is not deployed online. After the model solution is deployed online, the estimation request (that is, the data of the request data stream) is received to obtain sample data with features and feedback, and then the sample data with features and feedback can be used for model self-learning. The model obtained by self-learning can be deployed online, ensuring that the data and feature engineering solutions used for model self-learning are respectively consistent with the data and feature engineering solutions used for the model online estimation service, so as to achieve consistency between the model self-learning effect and the model estimation effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present disclosure, 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 of the present disclosure, and a person skilled in the art can also obtain other drawings based on these drawings.

[0019] Figure 1 is an exemplary architecture diagram of a device for applying machine learning provided by an embodiment of the present disclosure;

[0020] Figure 2 is an exemplary architecture diagram of another device for applying machine learning provided by an embodiment of the present disclosure;

[0021] Figure 3 yes Figure 2 An exemplary process logic block diagram of an apparatus for applying machine learning is shown;

[0022] Figure 4 yes Figure 2 An exemplary data flow diagram of an apparatus for applying machine learning is shown;

[0023] Figure 5 is an exemplary architecture diagram of an electronic device provided by an embodiment of the present disclosure;

[0024] Figure 6 It is an exemplary flow chart of a method for applying machine learning provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0025] In order to more clearly understand the above-mentioned purposes, features and advantages of the present disclosure, the present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the described embodiments are part of the embodiments of the present disclosure, rather than all of the embodiments. The specific embodiments described herein are only used to explain the present disclosure, rather than to limit the present disclosure. Based on the described embodiments of the present disclosure, all other embodiments obtained by ordinary technicians in the field belong to the scope of protection of the present disclosure.

[0026] It should be noted that, in this document, relational terms such as “first” and “second” are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations.

[0027] Different business scenarios in various industries have different business processing logics, but most of the different business scenarios require the application of machine learning to process business data, which can reduce the various problems caused by manual processing of business data, such as time-consuming, high labor costs, and inaccuracy. Figure 1An exemplary architecture diagram of a device for applying machine learning provided in an embodiment of the present disclosure, wherein the device for applying machine learning is applicable to supervised learning artificial intelligence modeling of various types of data, including but not limited to two-dimensional structured data, images, NLP (Natural Language Processing), voice, etc. The device for applying machine learning can be applied to a specified business scenario, wherein the information of the relevant data stream of the business scenario is predefined in the specified business scenario, wherein the relevant data stream may include but is not limited to: request data stream, display data stream, feedback data stream and business data stream, wherein the data of the display data stream is the data displayed by the specified business scenario based on the request data stream. Taking the short video application scenario as an example, after the request data is refreshed by the user by sliding or clicking the user terminal, a candidate video set will be screened out by the application background to form the request data that needs to be estimated by the model. The display data is what short videos the short video application actually shows to the user. The feedback data is, for example, whether the user clicks or watches the short video displayed by the short video application. The business data is, for example, the comment data and like data of the user when watching the short video, which are related to the business logic.

[0028] The information of the relevant data flow of the pre-defined business scenario can be understood as the fields included in the relevant data. For example, the relevant data flow is a request data flow. Accordingly, the information of the pre-defined request data flow can be understood as the fields included in the request data in the request data flow. The fields may be user ID, request content, request time, candidate material ID, etc.

[0029] After specifying the business scenario, you can Figure 1 The device for applying machine learning shown in FIG. 1 provides a model online estimation service. Figure 1 As shown, the device for applying machine learning may include but is not limited to: a data management module 100, a model solution exploration module 200, a model online estimation service module 300 and other components required for applying machine learning, such as an offline database, an online database, etc.

[0030] The data management module 100 is used to store and manage data from a specified business scenario and data produced by the model online estimation service module 300. The data from the specified business scenario is the relevant data stream obtained online by the data management module 100 based on the data service interface directly connecting to the specified business scenario. The data service interface is an application programming interface (API). In some embodiments, the data service interface is created by the data management module 100 based on the information of the relevant data stream of the specified business scenario predefined. In some embodiments, the data management module 100 can provide a user interface and receive information about the relevant data stream of the specified business scenario input by the user based on the user interface. In this embodiment, the user can be an operation and maintenance engineer of the specified business scenario. The data management module 100 can create a data service interface based on the information about the relevant data stream of the specified business scenario input by the user. In some embodiments, the data service interface is one-to-one with the relevant data stream, for example, the request data stream, the display data stream, the feedback data stream and the business data stream correspond to different data service interfaces respectively.

[0031] In some embodiments, the data management module 100 may accumulate data in the relevant data stream of the specified business scenario into a first database, wherein the first database is an offline database, for example, the offline database may be a distributed file storage system (HDFS, Hadoop Distributed File System), or may be other offline databases. In some embodiments, the data management module 100 may process the data of the request data stream to obtain sample data, wherein the processing method includes, for example, but is not limited to: using a filter for processing and flattening. The data management module 100 may accumulate the data of the request data stream, the sample data, the data of the feedback data stream, and the data of the business data stream into the first database. In some embodiments, after the data management module 100 obtains the data of the display data stream of the specified business scenario online based on the data service interface, the filter may be used to filter the data of the request data stream based on the data of the display data stream to obtain intersection data. For example, the display data stream has 10 data, the request data stream has 12 data, and the display data stream and the request data stream have 5 identical data, then the 5 identical data obtained by filtering by the filter are the intersection data, and the different data are filtered out. The data management module 100 may obtain sample data by flattening the intersection data (the five identical data). The data management module 100 may accumulate the data showing the data stream and the sample data obtained by filtering into a first database.

[0032] In some embodiments, the data management module 100 may receive data table attribute information input by a user through a user interface, wherein the data table attribute information describes the number of columns included in the data table and the data attributes of each column, such as user ID as a discrete field, request time as a time field, browsing duration as a numerical field, etc. The data management module 100 may receive a splicing scheme between data tables input by a user through a user interface, wherein the splicing scheme includes a splicing key for splicing different data tables, and a quantitative relationship, a temporal relationship, and an aggregation relationship between the same splicing keys between the primary and secondary tables. In some embodiments, the data management module 100 may maintain logical relationship information through the first database based on the data table attribute information and the splicing scheme; wherein the logical relationship information is information describing the relationship between different data tables, and the logical relationship information includes: the data table attribute information and the splicing scheme.

[0033] The model solution exploration module 200 is used to explore the model solution based on the data in the first database (for example, one or more of the logical relationship information, the data of the request data stream, the sample data, the data of the feedback data stream and the data of the business data stream, and the data of the display data stream) when the first preset condition is met. The first preset condition may include at least one of the data volume, time and manual trigger. For example, the first preset condition may be that the data volume in the first database reaches the preset data volume, or that the duration of data accumulation in the first database reaches the preset duration. The setting of the first preset condition can enable the model solution exploration module 200 to iteratively update the model solution. The model solution includes the following solution sub-items: feature engineering solution, model algorithm and model hyperparameters. The feature engineering solution is obtained based on the exploration of the logical relationship information. Therefore, the feature engineering solution has at least a table splicing function. It should be noted that the table splicing method of the feature engineering solution can be the same as or different from the table splicing solution input by the user. The feature engineering solution may also have other functions, such as extracting features from data for use by the model algorithm or model. The model algorithm may be a commonly used machine learning algorithm, such as a supervised learning algorithm, including but not limited to: LR (Logistic Regression), GBDT (Gradient Boosting Decision Tree), DeepNN (Deep Neural Network), etc. The hyperparameters of the model are parameters pre-set before machine learning to assist model training, such as the number of categories in a clustering algorithm, the step size of a gradient descent method, the number of layers of a neural network, the learning rate of training a neural network, etc.

[0034] In some embodiments, the model solution exploration module 200 may generate at least two model solutions when exploring model solutions. For example, at least two model solutions may be generated based on the logical relationship information maintained by the first database. Among them, at least one solution sub-item is different between different model solutions. In some embodiments, the model solution exploration module 200 uses the at least two model solutions to perform model training based on the data in the first database, and can obtain the parameters of the model itself, wherein the parameters of the model itself are, for example, weights in a neural network, support vectors in a support vector machine, coefficients in a linear regression or a logistic regression, etc. In some embodiments, the model solution exploration module 200 may evaluate the models trained by the at least two model solutions based on the machine learning model evaluation index, and then select from the at least two model solutions based on the evaluation results to obtain the explored model solution. The machine learning model evaluation index is, for example, an AUC (Area Under Curve) value, etc.

[0035] The model online estimation service module 300 is used to deploy the model solution explored by the model solution exploration module 200 online to provide model online estimation services, wherein the model online estimation service is performed based on the relevant data stream of the specified business scenario obtained online by the data service interface. In some embodiments, the model online estimation service module 300 only deploys the model solution online, but does not deploy the offline model obtained during the exploration process of the model solution exploration module 200 online, which can avoid the problem that the estimation effect of the deployed offline model is poor due to the inconsistency between the data obtained by online feature calculation and offline feature calculation after the offline model is directly deployed online. In addition, since the model online estimation service module 300 only deploys the model solution online, but does not deploy the offline model online, no estimation result is generated when providing the model online estimation service. When the request data is received, the default estimation result is sent to the specified business scenario, and the specified business scenario ignores the default estimation result after receiving it. Therefore, Figure 1 The model solution exploration module 200 points to the model online estimation service module 300 with a dotted arrow, indicating that the model solution will not provide online estimation services, but will still feedback default estimation results. In some embodiments, when the model online estimation service module 300 deploys the model solution online, it also deploys the offline model obtained during the exploration process of the model solution exploration module 200 online. The offline model is trained based on the relevant data of the specified business scenario accumulated in the first database (i.e., the offline database), and after the offline model is deployed online, it provides estimation services for the relevant data of the specified business scenario. Therefore, although the data obtained by calculating the online and offline features may be inconsistent, the same source of online and offline data is still achieved.

[0036] In some embodiments, after the model online estimation service module 300 deploys the model solution online, the relevant data stream of the specified business scenario obtained by the data service interface can be stored in the second database, wherein the second database is an online database, such as a real-time feature storage engine (rtidb), which is a distributed feature database for AI hard real-time scenarios, with features such as efficient computing, read-write separation, high concurrency, and high-performance query; the second database can also be other online databases. When the model online estimation service module 300 receives the request data, based on the feature engineering solution in the deployed model solution, the data in the second database and the received request data are used to perform online real-time feature calculations to obtain feature data of the estimated sample. In some embodiments, when the model online estimation service module 300 receives the request data, based on the feature engineering solution in the deployed model solution, the data in the second database and the received request data are spliced ​​and online real-time feature calculations are performed to obtain wide table feature data, and the feature data of the estimated sample obtained is wide table feature data.

[0037] In some embodiments, the model online estimation service module 300 can obtain the feature data (or wide table feature data) of the estimated sample based on the model scheme deployed online, splice the feature data and feedback data to generate sample data with features and feedback, and the sample data may also include other data, such as timestamp data, etc.; the feedback data comes from the feedback data stream. In some embodiments, before splicing the feature data and feedback data, the model online estimation service module 300 splices the feature data and display data to obtain feature data with display data, and the display data comes from the display data stream; and then splices the feature data with display data and feedback data to generate sample data with display, features and feedback. In some embodiments, the model online estimation service module 300 returns the sample data with features and feedback to the first database for model self-learning, and the model obtained by self-learning can be deployed online to ensure that the data and feature engineering scheme used for model self-learning are consistent with the data and feature engineering scheme used by the model online estimation service, respectively, to achieve consistency between the model self-learning effect and the model estimation effect.

[0038] Based on the above description of the data management module 100, the model solution exploration module 200 and the model online estimation service module 300, Figure 1It can be seen that the data management module 100, the model solution exploration module 200 and the model online estimation service module 300 constitute a machine learning closed loop. Since the data used for the model solution exploration is the data in the first database, and the first database is an offline database, the data used for the model solution exploration can be understood as offline data, and the data used for the model online estimation service is online data, and the offline data and the online data are both obtained from the specified business scenario by the data service interface. Therefore, it can be ensured that the data used for the model solution exploration (referred to as offline data) and the data used for the model online estimation service (referred to as online data) are of the same origin, thereby achieving the homology of offline and online data.

[0039] Figure 2 Another device for applying machine learning is provided in an embodiment of the present disclosure. The device for applying machine learning includes Figure 1 The data management module 100, model solution exploration module 200 and model online estimation service module 300 shown also include a model self-learning module 400 and other components required for applying machine learning, such as offline databases, online databases, etc.

[0040] The model self-learning module 400 is used to perform model self-learning based on the sample data with features and feedback in the first database when the second preset condition is met. The second preset condition may include at least one of data volume, time and manual triggering. For example, the second preset condition may be that the data volume in the first database reaches a preset data volume, or that the duration of data accumulation in the first database reaches a preset duration. The setting of the second preset condition may enable the model self-learning module 400 to iteratively update the model.

[0041] In some embodiments, when the second preset condition is met, the model self-learning module 400 is trained based on the sample data with features and feedback through the model algorithm and the model hyperparameters in the model solution to obtain a machine learning model. In some embodiments, if the model online estimation service module 300 deploys the model solution online, it also deploys the initial model online, wherein the initial model is an offline model generated in the process of exploring the model solution by the model solution exploration module 200, then the model self-learning module 400 trains the initial model through the model algorithm and the model hyperparameters in the model solution, updates the parameter values ​​of the initial model itself, and obtains a machine learning model. In some embodiments, if the model online estimation service module 300 deploys the model solution online without deploying the initial model online, then the model self-learning module 400 trains a random model through the model algorithm and the model hyperparameters in the model solution to obtain a machine learning model, wherein the random model is a model generated based on the model algorithm, and the parameter values ​​of the model itself are random values.

[0042] The model online estimation service module 300 can deploy the model obtained by the model self-learning module 400 online to provide model online estimation services. In some embodiments, after the model online estimation service module 300 deploys the model obtained by the model self-learning module 400 online, when request data is received, it generates feature-based estimation samples based on the data in the second database and the received request data, and obtains the estimation results of the estimation samples through the deployed online model, which is different from the model solution in that the deployed online model can obtain the estimation results of the estimation samples. The model online estimation service module 300 can send the estimation results to the specified business scenario for use or reference by the business scenario.

[0043] In some embodiments, the model online estimation service module 300 may replace the machine learning model that has been deployed online with the model obtained by the model self-learning module 400; or, deploy the model obtained by the model self-learning module 400 online and provide model online estimation services together with the machine learning model that has been deployed online. In some embodiments, the model online estimation service module 300 may replace the model solution that has been deployed online with the model solution obtained by the model solution exploration module 200; or, deploy the model solution obtained by the model solution exploration module 200 online without taking the model solution that has been deployed online offline.

[0044] Based on the above description of the model self-learning module 400 and the model online estimation service module 300, combined with Figure 2It can be seen that the data management module 100, the model self-learning module 400 and the model online estimation service module 300 constitute a machine learning closed loop. Since the sample data with features and feedback used by the model self-learning module 400 to train the model is generated online based on the data in the second database (that is, the online database) and the received request data after the model solution is deployed online, and the model online estimation service module 300 also provides estimation services based on the data in the second database after deploying the model trained by the model self-learning module 400 online, it is ensured that the data and feature engineering solutions used for model self-learning are respectively consistent with the data and feature engineering solutions used by the model online estimation service, thereby achieving consistency between the model self-learning effect and the model estimation effect.

[0045] In some embodiments, the division of each module in the device for applying machine learning is only a logical function division. There may be other division methods in actual implementation. For example, at least two modules of the data management module 100, the model solution exploration module 200, the model online estimation service module 300 and the model self-learning module 400 can be implemented as one module; the data management module 100, the model solution exploration module 200, the model online estimation service module 300 or the model self-learning module 400 can also be divided into multiple sub-modules. It is understandable that each module or sub-module can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application.

[0046] Figure 3 yes Figure 2 An exemplary process logic block diagram of an apparatus for applying machine learning is shown, as Figure 3 As shown, in the definition 301 of the specified business scenario, the user can input the information of the relevant data flow of the specified business scenario through the user interface. The user can also input data table attribute information and table assembly scheme through the user interface during the process of model scheme exploration 303. Figure 3 In the figure, data management 302, model self-learning 305, and model online estimation service 304 form a small closed loop; data management 302, model solution exploration 303, and model online estimation service 304 form a large closed loop. Among them, the small closed loop ensures that the data and feature engineering solutions used in model self-learning 305 are respectively consistent with the data and feature engineering solutions used in model online estimation service 304, so as to achieve consistency between model self-learning effect and model estimation effect. The large closed loop ensures that the data used in model solution exploration 303 (referred to as offline data) and the data used in model online estimation service 304 (referred to as online data) are of the same source, thus achieving the same source of offline and online data.

[0047] Figure 4 for Figure 2 An exemplary data flow diagram of an apparatus for applying machine learning is shown. Figure 4 The explanations of the English words are as follows:

[0048] GW is the gateway for the specified business scenario;

[0049] retain-mixer is used to implement the function of accumulating data in the relevant data stream of the specified business scenario into the first database in the data management module 100;

[0050] Trial1-mixer and trial2-mixer can be understood as two parallel model online estimation service modules 300;

[0051] HDFS is the first database;

[0052] rtidb1 and rtidb2 are two second databases;

[0053] AutoML is a model solution exploration module 200;

[0054] self-learn1 and self-learn2 are two model self-learning modules 400;

[0055] request is request data; impression is display data; action is feedback data; BOes is business data; enrich1 and enrich2 are wide table feature data; viewlog1 and viewlog2 are wide table feature data with feedback;

[0056] fedb1 and fedb2 can be understood as feature engineering solutions in the model solution.

[0057] Based on the above English word descriptions, the data flow of the device applying machine learning is described as follows:

[0058] Retain-mixer obtains request, impression, action and BOes from the specified business scenario based on the data service interface, and adds eventTime or ingestionTime to request, impression and action respectively, so that the data management module 100 can maintain the data timing relationship information in the logical relationship information. Adding eventTime belongs to the data management function of the data management module 100.

[0059] retain-mixer accumulates the request in HDFS for subsequent operation and maintenance. retain-mixer adds ingestionTime to impression, action and BOes respectively, obtains impression', action' and BOes', and accumulates impression', action' and BOes' in HDFS. Adding ingestionTime belongs to the data management function of the data management module 100.

[0060] retain-mixer processes request and impression through filter operation to obtain intersection data. For example, impression has 10 data and request has 12 data. Request and impression have 5 identical data. Then, through filter operation, these 5 identical data are obtained, i.e., intersection data, and different data are filtered out. Then, through flatten operation, the intersection data (these 5 identical data) is processed to obtain flatten_req (sample data). retain-mixer accumulates flatten_req in HDFS.

[0061] AutoML can explore model solutions based on flatten_req, impression', action', and BOes' in HDFS.

[0062] After trial1-mixer and trial2-mixer deploy different model solutions online, impression', action' and BOes' are accumulated in rtidb1 and rtidb2, and the user's historical data, such as user behavior data, can be synchronized to rtidb1 and rtidb2.

[0063] After trial1-mixer and trial2-mixer deploy different model solutions online, each time a request data is obtained, the accumulated data is obtained from rtidb1 and rtidb2 through fedb1 and fedb2 for feature engineering to obtain enrich1 and enrich2.

[0064] trial1-mixer and trial2-mixer join enrich1 and enrich2 with impression and action respectively and flatten them to get viewlog1 and viewlog2. trial1-mixer and trial2-mixer accumulate viewlog1 and viewlog2 into HDFS.

[0065] Self-learn1 and self-learn2 perform model self-learning based on viewlog1 and viewlog2 respectively to obtain machine learning models. Trial1-mixer and trial2-mixer deploy the machine learning models obtained by self-learn1 and self-learn2 respectively online to provide model online estimation services.

[0066] Depend on Figure 4 It can be seen that retain-mixer has the same data source as trial1-mixer and trial2-mixer, and accumulates data in HDFS, ensuring that the data used by AutoML and the data used after the model solution is deployed online are of the same source, achieving the same source of online and offline data. In addition, the data and feature engineering solutions used by self-learn1 and self-learn2 are consistent with the data and feature engineering solutions used after the model is deployed online, achieving consistency between the model self-learning effect and the model estimation effect.

[0067] It can be seen that the device for applying machine learning disclosed in this embodiment does not need to rely on importing historical offline data from other databases, and can collect data from scratch.

[0068] Figure 5 Schematic diagram of the structure of an electronic device provided by an embodiment of the present disclosure. Figure 5 As shown, the electronic device includes: at least one processor 501, at least one memory 502 and at least one communication interface 503. The various components in the electronic device are coupled together through a bus system 504. The communication interface 503 is used to transmit information between external devices. It can be understood that the bus system 504 is used to achieve connection and communication between these components. In addition to the data bus, the bus system 504 also includes a power bus, a control bus and a status signal bus. However, for the sake of clarity, Figure 5 Various buses are labeled as bus system 504 .

[0069] It can be understood that the memory 502 in this embodiment can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories.

[0070] In some implementations, the memory 502 stores the following elements, executable units or data structures, or a subset thereof, or an extended set thereof: an operating system and application programs.

[0071] The operating system includes various system programs, such as a framework layer, a core library layer, a driver layer, etc., which are used to implement various basic services and process hardware-based tasks. The application includes various application programs, such as a media player (Media Player), a browser (Browser), etc., which are used to implement various application services. The program that implements the method of applying machine learning provided in the embodiment of the present disclosure can be included in the application.

[0072] In the embodiment of the present disclosure, the processor 501 calls the program or instructions stored in the memory 502, specifically, the program or instructions stored in the application program, and the processor 501 is used to execute the steps of each embodiment of the method for applying machine learning provided in the embodiment of the present disclosure.

[0073] The method for applying machine learning provided in the embodiment of the present disclosure can be applied to the processor 501, or implemented by the processor 501. The processor 501 can be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in the processor 501 or an instruction in the form of software. The above-mentioned processor 501 can be a general-purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), a field programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc.

[0074] The steps of the method for applying machine learning provided in the embodiment of the present disclosure can be directly embodied as being executed by a hardware decoding processor, or being executed by a combination of hardware and software units in the decoding processor. The software unit can be located in a mature storage medium in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 502, and the processor 501 reads the information in the memory 502 and completes the steps of the method in combination with its hardware.

[0075] Figure 6 An exemplary flow chart of a method for applying machine learning provided in an embodiment of the present disclosure. The execution subject of the method is an electronic device. For ease of description, the process of the method for applying machine learning is described in the following embodiments using an electronic device as the execution subject.

[0076] In some embodiments, the electronic device may provide a user interface, based on which information about relevant data streams of a specified business scenario is received from a user input, wherein the relevant data streams include, but are not limited to, request data streams, display data streams, feedback data streams, and business data streams. The information about the relevant data streams of the specified business scenario may be understood as fields included in the relevant data. The electronic device then creates a data service interface based on the information about the relevant data streams of the specified business scenario, for example, the request data stream, display data stream, feedback data stream, and business data stream correspond to different data service interfaces, respectively.

[0077] In some embodiments, the electronic device may receive data table attribute information input by the user based on the user interface, wherein the data table attribute information describes the number of columns included in the data table and the data attributes of each column. The electronic device may also receive a splicing scheme between data tables input by the user through the user interface, wherein the splicing scheme includes splicing keys for splicing different data tables, and the quantity relationship, time sequence relationship and aggregation relationship of the same splicing keys between the main and sub tables. In some embodiments, the electronic device may maintain logical relationship information through the first database based on the data table attribute information and the splicing scheme; wherein the logical relationship information is information describing the relationship between different data tables, and the logical relationship information includes: data table attribute information and the splicing scheme.

[0078] In step 601, the electronic device obtains the relevant data stream of the specified business scenario online based on the data service interface. For example, the electronic device can obtain the display data stream of the specified business scenario online based on the data service interface, wherein the display data is the data displayed by the specified business scenario based on the request data stream.

[0079] In step 602, the electronic device accumulates the data in the related data stream into a first database. The first database is an offline database. In some embodiments, the electronic device processes the data of the request data stream to obtain sample data; and then accumulates the data of the request data stream, the sample data, the data of the feedback data stream, and the data of the business data stream into the first database. The processing method includes, for example, but is not limited to: processing using a filter and flattening. In some embodiments, the electronic device uses a filter to filter the data of the request data stream based on the data of the display data stream to obtain intersection data; and then obtains sample data by flattening the intersection data. The electronic device accumulates the display data and the sample data obtained by filtering into the first database.

[0080] In step 603, when the first preset condition is met, the electronic device explores a model solution based on the data in the first database (for example, one or more of the logical relationship information, the data of the request data stream, the sample data, the data of the feedback data stream and the data of the business data stream, and the data of the display data stream); the model solution includes the following solution sub-items: feature engineering solution, model algorithm and model hyperparameters. The feature engineering solution is obtained based on the exploration of the logical relationship information. Therefore, the feature engineering solution has at least a table splicing function. It should be noted that the table splicing method of the feature engineering solution can be the same as or different from the table splicing solution input by the user. The feature engineering solution can also have other functions, such as extracting features from the data for use by the model algorithm or model. In some embodiments, the first preset condition may include at least one of the data volume, time and manual triggering. For example, the first preset condition may be that the data volume in the first database reaches the preset data volume, or that the duration of data accumulation in the first database reaches the preset duration.

[0081] In some embodiments, when a first preset condition is met, the electronic device generates at least two model solutions. For example, at least two model solutions can be generated based on the logical relationship information maintained by the first database, wherein at least one solution sub-item is different between different model solutions; then, based on the data in the first database, the at least two model solutions are respectively used to perform model training; then, based on the machine learning model evaluation index, the models trained by the at least two model solutions are evaluated; finally, based on the evaluation result, a selection is made from the at least two model solutions to obtain the explored model solution.

[0082] In step 604, the electronic device deploys the model solution obtained through exploration online to provide a model online estimation service, wherein the model online estimation service is performed based on the relevant data stream of the specified business scenario obtained online by the data service interface. In some embodiments, the electronic device only deploys the model solution online, but does not deploy the offline model obtained during the model solution exploration process online, so as to avoid the problem that the estimation effect of the deployed offline model is poor due to the inconsistency between the data obtained by online feature calculation and offline feature calculation after the offline model is directly deployed online. In addition, since only the model solution is deployed online and the offline model is not deployed online, no estimation result is generated when providing the model online estimation service. When the request data is received, the default estimation result is sent to the specified business scenario, and the specified business scenario ignores the default estimation result after receiving it. In some embodiments, when the electronic device deploys the model solution online, it also deploys the offline model obtained in the model solution exploration process online. The offline model is trained based on the relevant data of the specified business scenario accumulated in the first database (i.e., the offline database), and after the offline model is deployed online, it provides estimation services for the relevant data of the specified business scenario. Therefore, although the data obtained by calculating the online and offline features may be inconsistent, the same source of online and offline data is still achieved.

[0083] In some embodiments, after the electronic device deploys the model solution obtained through exploration online, the data of the relevant data stream is stored in a second database, wherein the second database is an online database. When the electronic device receives the request data, based on the feature engineering solution in the deployed online model solution, the data in the second database and the received request data are used to perform online real-time feature calculations to obtain feature data of the estimated sample. In some embodiments, after the electronic device deploys the model solution obtained through exploration online, when the electronic device receives the request data, based on the feature engineering solution in the deployed online model solution, the data in the second database and the received request data are pieced together and online real-time feature calculations are performed to obtain wide table feature data, and the feature data of the estimated sample obtained is wide table feature data.

[0084] In some embodiments, the electronic device obtains feature data (or wide table feature data) of the estimated sample based on the model solution deployed online, and splices the feature data and feedback data to generate sample data with features and feedback, and the sample data may also include other data, such as timestamp data, etc.; the feedback data comes from the feedback data stream. In some embodiments, before splicing the feature data and feedback data, the electronic device splices the feature data and display data to obtain feature data with display data, and the display data comes from the display data stream; and then splices the feature data with display data and feedback data to generate sample data with display, features and feedback.

[0085] In some embodiments, the electronic device returns the sample data with features and feedback to the first database, and when a second preset condition is met, the electronic device performs model self-learning based on the sample data with features and feedback in the first database. The second preset condition may include at least one of data volume, time, and manual triggering. For example, the second preset condition may be that the data volume in the first database reaches a preset data volume, or that the duration of data accumulation in the first database reaches a preset duration.

[0086] In some embodiments, when the second preset condition is met, the electronic device will, based on the sample data with features and feedback, train through the model algorithm and the model hyperparameters in the model scheme to obtain a machine learning model. In some embodiments, if the electronic device deploys the model scheme online, it also deploys the initial model online, wherein the initial model is an offline model generated in the process of exploring the model scheme, then the electronic device trains the initial model through the model algorithm and the model hyperparameters in the model scheme, updates the parameter values ​​of the initial model itself, and obtains a machine learning model. In some embodiments, if the electronic device deploys the model scheme online without deploying the initial model online, then the electronic device trains a random model through the model algorithm and the model hyperparameters in the model scheme to obtain a machine learning model, wherein the random model is a model generated based on the model algorithm, and the parameter values ​​of the model itself are random values.

[0087] In some embodiments, the electronic device deploys the machine learning model online to provide an online model estimation service. In some embodiments, after the electronic device deploys the machine learning model online, when request data is received, an estimation sample with features is generated based on the data in the second database and the received request data, and the estimation result of the estimation sample is obtained through the deployed online model. The difference from the model solution is that the deployed online model can obtain the estimation result of the estimation sample. The electronic device can send the estimation result to the specified business scenario for use or reference by the business scenario.

[0088] In some embodiments, the electronic device replaces the machine learning model that has been deployed online with the model obtained by self-learning of the model; or, the model obtained by self-learning of the model is deployed online, and together with the machine learning model that has been deployed online, provides a model online estimation service. In some embodiments, the electronic device replaces the model solution that has been deployed online with the model solution obtained by exploration; or, the model solution obtained by exploration is deployed online, and the model solution that has been deployed online is not taken offline.

[0089] It can be seen that the method of applying machine learning disclosed in the above embodiment, since the data used for model solution exploration is the data in the first database, and the first database is an offline database, the data used for model solution exploration can be understood as offline data, and the data used for the model online estimation service is online data, and the offline data and online data are both obtained from the specified business scenario by the data service interface. Therefore, it can be ensured that the data used for model solution exploration (referred to as offline data) and the data used for the model online estimation service (referred to as online data) are of the same origin, thereby achieving the homology of offline and online data.

[0090] In addition, the method of applying machine learning disclosed in the above embodiments, since the sample data with features and feedback used for model self-learning is generated online based on the data in the second database (i.e., the online database) and the received request data after the model solution is deployed online, and the model obtained by the model self-learning module also provides estimation services based on the data in the second database after it is deployed online, it is ensured that the data and feature engineering solutions used for model self-learning are respectively consistent with the data and feature engineering solutions used for the model online estimation service, thereby achieving consistency between the model self-learning effect and the model estimation effect.

[0091] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all described as a series of action combinations, but those skilled in the art can understand that the embodiments of the present disclosure are not limited by the described action sequence, because according to the embodiments of the present disclosure, some steps can be performed in other sequences or simultaneously. In addition, those skilled in the art can understand that the embodiments described in the specification are all optional embodiments.

[0092] The embodiments of the present disclosure also propose a non-transitory computer-readable storage medium, which stores programs or instructions. The programs or instructions enable a computer to execute the steps of each embodiment of the method for applying machine learning. To avoid repeated description, they are not repeated here.

[0093] It should be noted that, in this article, the terms "include", "comprises" or any other variations 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, an element defined by the sentence "includes..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0094] Those skilled in the art will appreciate that although some embodiments described herein include certain features included in other embodiments but not other features, the combination of features from different embodiments is meant to be within the scope of the present disclosure and to form different embodiments.

[0095] Those skilled in the art will appreciate that the description of each embodiment has its own emphasis, and for parts not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0096] Although the embodiments of the present disclosure have been described in conjunction with the accompanying drawings, those skilled in the art may make various modifications and variations without departing from the spirit and scope of the present disclosure, and such modifications and variations are all within the scope defined by the appended claims.

Claims

1. A method for applying machine learning, characterized in that: The method comprises: Obtain relevant data flows of a specified business scenario online based on a data service interface, wherein the relevant data flows at least include request data flows; accumulating data in the related data stream into a first database; When the first preset condition is met, exploring a model solution based on the data in the first database; the model solution includes the following solution sub-items: feature engineering solution, model algorithm and model hyperparameters; The model solution obtained through exploration is deployed online to provide an online model estimation service, wherein the online model estimation service is performed based on the relevant data stream of the specified business scenario obtained online through the data service interface; After the explored model solution is deployed online, the method further includes: Storing the data of the related data stream in a second database; the second database supports online real-time feature calculation; When the request data is received, based on the feature engineering solution in the deployed model solution, online real-time feature calculation is performed using the data in the second database and the received request data to obtain feature data of the estimated sample, so as to perform model self-learning based on the feature data, and the model obtained by the model self-learning is deployed online to provide model online estimation service; Wherein, the model self-learning includes: An initial model or a random model is trained by the model algorithm and the model hyperparameters in the model scheme to obtain a machine learning model, wherein the initial model is a model generated in the process of exploring the model scheme, and the random model is a model generated based on the model algorithm, and the parameter values ​​of the model itself are random values.

2. The method according to claim 1, characterized in that Before the step of acquiring the relevant data flow of the specified business scenario online based on the data service interface, the method further includes: Providing a user interface, and receiving information about relevant data flows of the specified business scenario input by a user based on the user interface; The data service interface is created based on the information about the relevant data flow of the specified business scenario.

3. The method according to claim 1, characterized in that The relevant data flows include: request data flow, feedback data flow and business data flow.

4. The method according to claim 3, characterized in that The accumulating the data in the related data stream into the first database comprises: Processing the data of the request data stream to obtain sample data; The data of the request data stream, the sample data, the data of the feedback data stream and the data of the service data stream are accumulated in the first database.

5. The method according to claim 4, characterized in that The related data stream also includes a display data stream, wherein the data in the display data stream is data displayed by the specified business scenario based on the request data stream; Accordingly, processing the data of the request data stream to obtain sample data includes: filtering the data of the request data stream based on the data of the display data stream to obtain intersection data; processing the intersection data to obtain sample data; Accordingly, the data of the presentation data stream and the sample data are accumulated in the first database.

6. The method according to any one of claims 1 to 5, characterized in that: The data exploration model solution based on the first database includes: Generate at least two model solutions, wherein at least one solution sub-item is different between different model solutions; Based on the data in the first database, respectively use the at least two model schemes to perform model training; Based on the machine learning model evaluation index, the models trained by the at least two model schemes are evaluated; A selection is made from the at least two model solutions based on the evaluation result to obtain an explored model solution.

7. The method according to claim 6, characterized in that The method further comprises: Receive data table attribute information and table assembly plan input by the user based on the user interface; Based on the data table attribute information and the table splicing scheme, logical relationship information is maintained through the first database; the logical relationship information is information describing the relationship between different data tables; Accordingly, generating at least two model solutions includes: At least two model solutions are generated based on the logical relationship information.

8. The method according to claim 7, characterized in that The table splicing scheme includes splicing keys, time series relationships and aggregation relationships for splicing different data tables; The logical relationship information includes: data table attribute information and the table splicing scheme.

9. The method according to claim 1, characterized in that: Performing online real-time feature calculation using the data in the second database and the received request data includes: Based on the feature engineering solution in the deployed model solution, the data in the second database and the received request data are combined and online real-time feature calculation is performed to obtain wide table feature data; Accordingly, the characteristic data of the estimated sample is wide-table characteristic data.

10. The method according to claim 1, characterized in that The method further comprises: After the model solution obtained through exploration is deployed online, when request data is received and the model is not deployed online, a default estimation result is sent to the specified business scenario.

11. The method according to claim 1, characterized in that: The deployment of the explored model solution to provide online model estimation services includes: Obtain feature data of the estimated samples based on the deployed model solution; Splicing the feature data and feedback data to generate sample data with features and feedback; the feedback data comes from a feedback data stream; Returning the sample data with features and feedback to the first database; When the second preset condition is met, performing model self-learning based on the sample data with features and feedback in the first database; The model obtained by self-learning of the model is deployed online to provide model online estimation service.

12. The method according to claim 11, characterized in that Before splicing the feature data and the feedback data, the method further includes: splicing the feature data and the display data to obtain the feature data with the display data; the display data is derived from the display data stream; Accordingly, the feature data and feedback data with the display data are spliced ​​to generate sample data with display, feature and feedback.

13. The method according to claim 11, characterized in that The self-learning of the model based on the sample data with features and feedback in the first database includes: Based on the sample data with features and feedback, the machine learning model is obtained by training the model algorithm and the hyperparameters of the model in the model scheme.

14. The method according to claim 13, characterized in that The method of training the model algorithm and the hyperparameters of the model in the model scheme to obtain a machine learning model includes: The initial model is trained by the model algorithm and the model hyperparameters in the model scheme to obtain a machine learning model; when the model scheme obtained through exploration is deployed online, the initial model is also deployed online.

15. The method according to claim 13, characterized in that The method of training the model algorithm and the hyperparameters of the model in the model scheme to obtain a machine learning model includes: A random model is trained by using the model algorithm and the model hyperparameters in the model scheme to obtain a machine learning model; when the model scheme obtained through exploration is deployed online, the initial model is not deployed online.

16. The method according to claim 11, characterized in that Deploying the model obtained by self-learning of the model online to provide the model online estimation service includes: After the model obtained by self-learning of the model is deployed online, when request data is received, an estimated sample with features is generated based on the data in the second database and the received request data, and an estimated result of the estimated sample is obtained through the deployed online model; The estimation result is sent to the specified business scenario.

17. The method according to claim 11, characterized in that Deploying the model obtained by self-learning of the model online includes: The model obtained by self-learning of the model replaces the machine learning model that has been deployed online; or, the model obtained by self-learning of the model is deployed online, and together with the machine learning model that has been deployed online, a model online estimation service is provided; The deployment of the explored model solution includes: The model solution obtained by the exploration replaces the model solution that has been deployed online; or, the model solution obtained by the exploration is deployed online, and the model solution that has been deployed online does not go offline.

18. The method according to claim 11, characterized in that The first preset condition and the second preset condition include: data volume, time and / or manual triggering.

19. A device for applying machine learning, characterized in that: The device comprises: A data management module, configured to obtain online relevant data streams of a specified business scenario based on a data service interface, wherein the relevant data streams at least include request data streams; and accumulate data in the relevant data streams into a first database; A model solution exploration module, used for exploring a model solution based on the data in the first database when a first preset condition is met; the model solution includes the following solution sub-items: a feature engineering solution, a model algorithm, and a model hyperparameter; A model online estimation service module, used to deploy the model solution obtained by the model solution exploration module online to provide a model online estimation service, wherein the model online estimation service is performed based on the relevant data stream of the specified business scenario obtained online by the data service interface; The model online estimation service module is also used for: After the model solution obtained by the model solution exploration module is deployed online, the data of the relevant data stream is stored in a second database; the second database supports online real-time feature calculation; When the request data is received, based on the feature engineering solution in the deployed model solution, online real-time feature calculation is performed using the data in the second database and the received request data to obtain feature data of the estimated sample, so as to perform model self-learning based on the feature data, and the model obtained by the model self-learning is deployed online to provide model online estimation service; Wherein, the model self-learning includes: An initial model or a random model is trained by the model algorithm and the model hyperparameters in the model scheme to obtain a machine learning model, wherein the initial model is a model generated in the process of exploring the model scheme, and the random model is a model generated based on the model algorithm, and the parameter values ​​of the model itself are random values.

20. The device according to claim 19, characterized in that The data management module is also used for: Providing a user interface, and receiving information about relevant data flows of the specified business scenario input by a user based on the user interface; The data service interface is created based on the information about the relevant data flow of the specified business scenario.

21. The device according to claim 19, characterized in that The relevant data flows include: request data flow, feedback data flow and business data flow.

22. The device according to claim 21, characterized in that The data management module accumulating the data in the related data stream into the first database comprises: Processing the data of the request data stream to obtain sample data; The data of the request data stream, the sample data, the data of the feedback data stream and the data of the service data stream are accumulated in the first database.

23. The device according to claim 22, characterized in that The related data stream also includes a display data stream, wherein the data in the display data stream is data displayed by the specified business scenario based on the request data stream; The data management module processes the data of the request data stream to obtain sample data, including: filtering the data of the request data stream based on the data of the display data stream to obtain intersection data; processing the intersection data to obtain sample data; The data management module is further used to accumulate the data of the display data stream and the sample data into the first database.

24. The device according to any one of claims 19 to 23, characterized in that The model solution exploration module is used to: When the first preset condition is met, at least two model solutions are generated, wherein at least one solution sub-item is different between different model solutions; Based on the data in the first database, respectively use the at least two model schemes to perform model training; Based on the machine learning model evaluation index, the models trained by the at least two model schemes are evaluated; A selection is made from the at least two model solutions based on the evaluation result to obtain an explored model solution.

25. The device according to claim 24, characterized in that The data management module is also used for: Receive data table attribute information and table assembly plan input by the user based on the user interface; Maintaining logical relationship information through the first database based on the data table attribute information and the table assembly scheme; The logical relationship information is information describing the relationship between different data tables; Correspondingly, the model solution exploration module generates at least two model solutions, including: generating at least two model solutions based on the logical relationship information.

26. The device according to claim 25, characterized in that The table splicing scheme includes splicing keys, time series relationships and aggregation relationships for splicing different data tables; The logical relationship information includes: data table attribute information and the table splicing scheme.

27. The device according to claim 19, characterized in that The model online estimation service module uses the data in the second database and the received request data to perform online real-time feature calculation, including: Based on the feature engineering solution in the deployed model solution, the data in the second database and the received request data are combined and online real-time feature calculation is performed to obtain wide table feature data; Accordingly, the characteristic data of the estimated sample is wide-table characteristic data.

28. The device according to claim 19, characterized in that The model online estimation service module is also used for: After the model solution obtained by the model solution exploration module is deployed online, when request data is received and the model is not deployed online, a default estimation result is sent to the specified business scenario.

29. The device according to claim 19, characterized in that The model online estimation service module is further used to: obtain feature data of the estimated sample based on the deployed model solution; splice the feature data and feedback data to generate sample data with features and feedback, wherein the feedback data comes from the feedback data stream; return the sample data with features and feedback to the first database; The device further comprises a model self-learning module, which is used to perform model self-learning based on the sample data with features and feedback in the first database when a second preset condition is met; The model online estimation service module is also used to deploy the model obtained by the model self-learning module online to provide model online estimation service.

30. The device according to claim 29, characterized in that Before the model online estimation service module splices the feature data and the feedback data, it is also used to: splicing the feature data and the display data to obtain the feature data with the display data; the display data is derived from the display data stream; Correspondingly, the model online estimation service module splices the feature data and feedback data with the display data to generate sample data with display, features and feedback.

31. The device according to claim 29, characterized in that The model self-learning module is used to: When the second preset condition is met, the machine learning model is obtained by training the model algorithm and the hyperparameters of the model in the model scheme based on the sample data with features and feedback.

32. The device according to claim 31, characterized in that The model self-learning module is trained by the model algorithm and the model hyperparameters in the model scheme to obtain a machine learning model, including: The initial model is trained by the model algorithm and the model hyperparameters in the model solution to obtain a machine learning model; when the model online estimation service module deploys the model solution obtained by the model solution exploration module online, the initial model is also deployed online.

33. The device according to claim 31, characterized in that The model self-learning module is trained by the model algorithm and the model hyperparameters in the model scheme to obtain a machine learning model, including: A random model is trained by using the model algorithm and the model hyperparameters in the model solution to obtain a machine learning model; wherein when the model online estimation service module deploys the model solution obtained by the model solution exploration module online, the initial model is not deployed online.

34. The device according to claim 29, characterized in that The model online estimation service module deploys the model obtained by the model self-learning module online to provide the model online estimation service, including: After the model obtained by the model self-learning module is deployed online, when request data is received, an estimated sample with features is generated based on the data in the second database and the received request data, and an estimated result of the estimated sample is obtained through the deployed online model; The estimation result is sent to the specified business scenario.

35. The device according to claim 29, characterized in that The model online estimation service module deploys the model obtained by the model self-learning module online, including: The model obtained by the model self-learning module replaces the machine learning model that has been deployed online; or, the model obtained by the model self-learning module is deployed online, and together with the machine learning model that has been deployed online, a model online estimation service is provided; The model online estimation service module deploys the model solution obtained by the model solution exploration module online, including: The model solution obtained by the model solution exploration module replaces the model solution that has been deployed online; or, the model solution obtained by the model solution exploration module is deployed online, and the model solution that has been deployed online does not go offline.

36. The device according to claim 29, characterized in that The first preset condition and the second preset condition include: data volume, time and / or manual triggering.

37. An electronic device, characterized in that: include: Processor and memory; The processor is used to execute the steps of the method according to any one of claims 1 to 18 by calling the program or instruction stored in the memory.

38. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores a program or instruction, which enables a computer to execute the steps of the method according to any one of claims 1 to 18.

Images