Deployment method and device for machine learning model prediction online service

Abstract

The present invention provides a method and device for deploying online services for machine learning model prediction, and relates to the field of computer application technology. The method comprises: firstly, using a feature processing operator to perform feature processing on pre-acquired offline data to generate a model training set, wherein the feature processing operator comprises a first online computing interface and a second online computing interface; then, using the model training set to train the operator of the machine learning algorithm to generate a trained machine learning model; then, according to the feature processing operator and the trained machine learning model, determining an operator link; finally, based on the first online computing interface and the second online computing interface in the operator link, generating a model online service, the model online service is used to process real-time requested data and generate model calculation results. The above method can replace manual code conversion, improve the efficiency of model development, and reduce the possibility of introducing errors and the cost of online model deployment.

Description

Technical Field

[0001] The present invention relates to the field of computer application technology, and in particular to a method and device for deploying a machine learning model to predict online services. Background Art

[0002] With the emergence of massive data, artificial intelligence technology has developed rapidly. Among them, machine learning technology is widely used to mine useful value from massive data records (for example, financial data, Internet data, etc.), and more and more machine learning models are deployed in the form of online services. In a standard machine learning development process, the original data is processed by features to form a feature wide table, and then the feature wide table is used for model training to obtain a machine learning model. When using this type of model, it is necessary to deploy the trained model as a model prediction service interface, request data to form a feature wide table through the same feature processing logic as in the development process, and then call the model prediction service interface to obtain the prediction result.

[0003] In the process of model building, Python or R is often used for model development, and then the feature calculation logic is converted into Java for deployment. Since different languages ​​are used to generate feature wide tables during model development and model deployment, code translation is inevitable. The current common solution is to use manual code conversion, which is not only time-consuming and labor-intensive, but also has the possibility of introducing bugs, increasing the cost of using artificial intelligence (AI) model technology to solve actual business problems. Summary of the invention

[0004] The purpose of the present invention is to provide a method and device for deploying online services using a machine learning model to predict the performance of an online service, so as to alleviate the technical problems in the prior art of time-consuming and labor-intensive manual code conversion and low development efficiency.

[0005] In order to achieve the above purpose, the technical solution adopted by the embodiment of the present invention is as follows:

[0006] In a first aspect, an embodiment of the present invention provides a method for deploying a machine learning model prediction online service, comprising: performing feature processing on pre-acquired offline data using a feature processing operator to generate a model training set; the feature processing operator includes a first online computing interface and a second online computing interface; training the operator of the machine learning algorithm using the model training set to generate a trained machine learning model; the operator of the trained machine learning model includes a first online computing interface; determining an operator link based on the feature processing operator and the trained machine learning model; the operator link includes the feature processing operator and the operator of the trained machine learning model; generating a model online service based on the first online computing interface and the second online computing interface in the operator link; the model online service is used to process real-time requested data and generate model calculation results.

[0007] In some possible implementations, before the step of performing feature processing on pre-acquired offline data to generate a model training set, it also includes: determining operators for feature processing and operators for machine learning algorithms; acquiring offline data for a modeling scenario, and determining an offline modeling training set.

[0008] In some possible implementations, the operator for the feature processing includes a first operator and a second operator, and the operator for the machine learning algorithm includes a first operator; the first operator includes a first online computing interface and a first offline computing interface, and the second operator includes a second online computing interface and a second offline computing interface.

[0009] In some possible implementations, the first operator generates a configuration file when performing offline calculations, and the configuration file is used to calculate the calculation logic of the first operator.

[0010] In some possible implementations, the above-mentioned feature processing includes: one or more of data cleaning, data filling and feature derivation; if the above-mentioned feature processing is the above-mentioned data cleaning or the above-mentioned data filling, the operator of the above-mentioned feature processing is the second operator; if the above-mentioned feature processing is feature derivation, the operator of the above-mentioned feature processing is the first operator, or the first operator and the second operator.

[0011] In some possible implementations, the operator link includes multiple first operators and multiple second operators; the step of generating a model online service based on the first online computing interface and the second online computing interface in the operator link includes: calling all the first operators and the second operators in the operator link; generating a model online service based on the first online computing interface of the first operator and the second online computing interface of the second operator.

[0012] In some possible implementations, the first operator is an Estimator type operator, and the second operator is a Transformer type operator.

[0013] In a second aspect, an embodiment of the present invention provides a deployment device for a machine learning model prediction online service, comprising: a feature processing module, used to perform feature processing on pre-acquired offline data using a feature processing operator to generate a model training set; the above-mentioned feature processing operator includes a first online computing interface and a second online computing interface; a training module, used to train the operator of the machine learning algorithm using the above-mentioned model training set to generate a trained machine learning model; the operator of the above-mentioned trained machine learning model includes a first online computing interface; a determination module, used to determine an operator link according to the feature processing operator and the trained machine learning model; the operator link includes the feature processing operator and the operator of the trained machine learning model; an online service generation module, used to generate a model online service based on the first online computing interface and the second online computing interface in the operator link; the model online service is used to process real-time requested data and generate model calculation results.

[0014] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program executable on the processor, and when the processor executes the computer program, the steps of any one of the methods described in the first aspect are implemented.

[0015] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores machine-executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions cause the processor to execute any method described in the first aspect above.

[0016] The present invention provides a method and device for deploying online services for predicting machine learning models, the method comprising: firstly, using a feature processing operator to perform feature processing on pre-acquired offline data to generate a model training set, wherein the feature processing operator comprises a first online computing interface and a second online computing interface; then, using the model training set to train the operator of the machine learning algorithm to generate a trained machine learning model; then, according to the feature processing operator and the trained machine learning model, determining an operator link; finally, based on the first online computing interface and the second online computing interface in the operator link, generating a model online service, the model online service is used to process real-time requested data and generate model calculation results. Through the above method, the process of manual code conversion can be replaced, the efficiency of model development can be improved, the possibility of introducing errors can be reduced, and the cost of online model deployment can be reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the specific implementation methods of the present invention or the technical solutions in the prior art, the drawings required for use in the specific implementation methods or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some implementation methods of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A flow chart of a method for deploying a machine learning model to predict online services provided by an embodiment of the present invention;

[0019] Figure 2 A schematic diagram of the development process of two operators in a method for deploying a machine learning model prediction online service provided by an embodiment of the present invention;

[0020] Figure 3 A schematic diagram of the structure of a deployment device for a machine learning model prediction online service provided by an embodiment of the present invention;

[0021] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0023] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] In a standard machine learning development process, the original data is processed by features to form a feature wide table, and then the feature wide table is used for model training to obtain a machine learning model. When using this type of model, the trained model needs to be deployed as a model prediction service interface, the request data is processed by the same feature processing logic as in the development process to form a feature wide table, and then the model prediction service interface is called to obtain the prediction result. The execution efficiency of the language used for model development (such as python, R, etc.) is not enough, and the existing development process cannot meet the online business scenarios with timeliness requirements. For example, online credit review generally requires feedback results in seconds; the advertising column recommendation position in the application requires millisecond response, and the recommendation result can be returned when the customer opens the application. It is inconvenient to directly develop models in languages ​​such as java and scal that can meet timeliness requirements. Therefore, in most cases, python and R languages ​​are used for model development. When deploying the model, the feature calculation logic is converted into java language and then deployed. The actual prediction workflow is to execute the java feature processing logic in real time and then call the model service.

[0025] In the process of model building, Python or R is often used for model development, and then the feature calculation logic is converted into Java for deployment. Since different languages ​​are used to generate feature wide tables during model development and model deployment, code translation is inevitable. The current common solution is to use manual code conversion, which is not only time-consuming and labor-intensive, but also has the possibility of introducing bugs, increasing the cost of using artificial intelligence (AI) model technology to solve actual business problems.

[0026] Based on this, the embodiment of the present invention provides a method and device for deploying a machine learning model prediction online service. To facilitate understanding of the present embodiment, a method for deploying a machine learning model prediction online service disclosed in the embodiment of the present invention is first described in detail. Figure 1 The flowchart of a method for deploying a machine learning model prediction online service is shown. The method can be executed by an electronic device and mainly includes the following steps S110 to S140:

[0027] S110: performing feature processing on the pre-acquired offline data using a feature processing operator to generate a model training set; the feature processing operator includes a first online computing interface and a second online computing interface;

[0028] Feature processing includes: one or more of data cleaning, data filling and feature derivation; if the feature processing is data cleaning or data filling, the operator of the feature processing is the second operator; if the feature processing is feature derivation, the operator of the feature processing is the first operator, or the first operator and the second operator.

[0029] As a specific example, the first operator is an Estimator operator, and the second operator is a Transformer operator. The Estimator operator includes an offline computing interface and an online computing interface, and the Estimator operator also includes an offline computing interface and an online computing interface.

[0030] Feature processing generally includes multiple steps, some of which use Transformer operators and some use Estimator operators. The specific type of operator used is determined by the calculation logic of the calculation step. Generally speaking, Transformer operators are used in most cases of data cleaning and data filling; feature derivation is more complicated. Depending on the calculation logic, a considerable number of feature derivations use Estimator operators, while others use Transformer operators.

[0031] The main difference between the two operators is that when feature processing logic needs to be executed during model prediction, the Transformer operator is only related to the calculation logic itself, while the Estimator operator is also related to the parameters generated by feature processing execution during model training.

[0032] S120: Using the model training set to train the operator of the machine learning algorithm to generate a trained machine learning model;

[0033] Among them, the operator of the trained machine learning model includes a first online computing interface;

[0034] S130: Determine an operator link according to the feature processing operator and the trained machine learning model; the operator link includes the feature processing operator and the trained machine learning model operator;

[0035] S140: Generate a model online service based on the first online computing interface and the second online computing interface in the operator link; the model online service is used to process real-time requested data and generate model calculation results.

[0036] In one embodiment, before the above method uses the feature processing operator to perform feature processing on the pre-acquired offline data to generate a model training set, it is generally necessary to determine the feature processing operator and the operator of the machine learning algorithm, and then obtain the offline data of the modeling scenario to determine the offline modeling training set.

[0037] The operator of feature processing includes a first operator and a second operator, the operator of the machine learning algorithm includes the first operator; the first operator includes a first online computing interface and a first offline computing interface, and the second operator includes a second online computing interface and a second offline computing interface. When the first operator performs offline computing, a configuration file is generated, and the configuration file is used to calculate the computing logic of the first operator.

[0038] As a specific example, feature processing generally includes multiple operators, and in most cases includes both Estimator-type and Transformer-type operators; the operator of the machine learning algorithm may be one or multiple, all of which are Estimator-type operators. Since the operator chain generated in the above step S130 includes operators for feature processing and operators for trained machine learning models, the operator chain includes multiple first operators and multiple second operators, that is, the operator chain includes multiple first online computing interfaces and multiple second online computing interfaces.

[0039] Then, the step of generating the model online service based on the first online computing interface and the second online computing interface in the operator link may include: first calling all the first operators and the second operators in the operator link; then generating the model online service based on the first online computing interface of the first operator and the second online computing interface of the second operator. Among them, the first operator is an Estimator type operator, and the second operator is a Transformer type operator. This method packages feature processing and machine learning model prediction into an operator link, that is, the feature processing and model prediction functions are packaged and released, reducing the workload of going online.

[0040] The present invention provides a method and device for deploying online services for predicting machine learning models, the method comprising: firstly, using a feature processing operator to perform feature processing on pre-acquired offline data to generate a model training set, wherein the feature processing operator comprises a first online computing interface and a second online computing interface; then, using the model training set to train the operator of the machine learning algorithm to generate a trained machine learning model; then, according to the feature processing operator and the trained machine learning model, determining an operator link; finally, based on the first online computing interface and the second online computing interface in the operator link, generating a model online service, the model online service is used to process real-time requested data and generate model calculation results. Through the above method, the process of manual code conversion can be replaced, the efficiency of model development can be improved, the possibility of introducing errors can be reduced, and the cost of online model deployment can be reduced.

[0041] Taking the most common process of using Python language to develop AI models as an example, an embodiment includes the following process:

[0042] First, we need to build algorithm operators commonly used in AI modeling business. Common operators include feature processing and machine learning algorithms. Feature processing includes data cleaning, data filling, feature derivation, etc. These operators can be divided into two categories: Transformer type and Estimator.

[0043] Transformer operators: The output results of this type of operator are only related to a single piece of input data; for example: data filling, comprehensive operations on multiple fields, string segmentation, etc. The calculation logic of this type of operator is exactly the same during offline model development and online prediction. The calculation interface can be implemented by directly calling Python code in Java or Scala language, such as Figure 2 As shown in part (A).

[0044] Estimator type operators: The output results of this type of operator are not only related to a single piece of input data, but also to the entire training sample set. For example: normalization, standardization, model prediction and other operators. Normalization requires the calculation of normalized transformation parameters based on the entire training sample set before it can be executed. Model prediction requires training a model before it can be predicted based on the input data. When this type of operator is calculated offline, a configuration file will be generated to record the calculation logic of the operator. When making online predictions, it is necessary to load the configuration file and python code at the same time in the java or scala language environment to obtain the output result based on a single piece of input data, such as Figure 2 As shown in part (B).

[0045] By building the above two types of operators, we wrote support interfaces for Scala and Python languages, so that the corresponding codes can be served online. When the Estimator operator is used for online prediction, it is necessary to load the configuration file and Python code in the Java or Scala language environment at the same time to obtain the output result based on the input single data. And the machine learning-specific computing logic is componentized. When the Estimator operator is used for online prediction, it is necessary to load the configuration file and Python code in the Java or Scala language environment at the same time to obtain the output result based on the input single data. Therefore, corresponding components need to be developed for the Estimator-type computing logic to achieve rapid deployment.

[0046] Then read the offline data of any modeling scenario as the training set for offline modeling; perform feature engineering on the read offline data, calculate the required input features, and generate a model training set; then input the model training set into the machine learning algorithm operator for model training.

[0047] After training is completed, the operators and corresponding model files used above are input into the model compiler to generate a model online service. The calculation logic of the model compiler is: connect the feature processing operator and the machine learning algorithm operator in the order used in offline calculation (that is, first the feature processing operator and then the machine learning algorithm operator) to generate an operator link, and call the real-time online calculation interface of the operator in sequence. Finally, the real-time requested data can be input into the above model online service, and the model calculation results can be output in real time.

[0048] The deployment method of the machine learning model prediction online service provided in the embodiment of the present application integrates the development and deployment of the machine learning model. After the development is completed, the deployment code can be automatically generated and deployed online, saving a lot of development manpower. The feature development operator is divided into two types: Transformer and Estimator. Different processing methods are used to ensure the consistency of the feature development code in both training and deployment states, avoiding logical bugs that may occur during code rewriting.

[0049] The embodiment of the present invention also provides a deployment device for machine learning model prediction online service, see Figure 3 , the device comprises:

[0050] The feature processing module 310 is used to perform feature processing on the pre-acquired offline data using a feature processing operator to generate a model training set; the feature processing operator includes a first online computing interface and a second online computing interface;

[0051] A training module 320 is used to train the operator of the machine learning algorithm using the model training set to generate a trained machine learning model; the operator of the trained machine learning model includes a first online computing interface;

[0052] A determination module 330 is used to determine an operator chain according to the operator of feature processing and the trained machine learning model; the operator chain includes the operator of feature processing and the operator of the trained machine learning model;

[0053] The online service generation module 340 is used to generate a model online service based on the first online computing interface and the second online computing interface in the operator link; the model online service is used to process the real-time requested data and generate model calculation results.

[0054] The deployment device of the machine learning model prediction online service provided in the embodiment of the present application can be specific hardware on the device or software or firmware installed on the device, etc. The implementation principle and technical effect of the device provided in the embodiment of the present application are the same as those of the aforementioned method embodiment. For the sake of brief description, the parts not mentioned in the device embodiment can refer to the corresponding contents in the aforementioned method embodiment. Technical personnel in the relevant field can clearly understand that for the convenience and conciseness of description, the specific working processes of the systems, devices and units described above can all refer to the corresponding processes in the aforementioned method embodiments, and will not be repeated here. The deployment device of the machine learning model prediction online service provided in the embodiment of the present application has the same technical features as the deployment method of the machine learning model prediction online service provided in the above embodiment, so it can also solve the same technical problems and achieve the same technical effects.

[0055] An embodiment of the present application further provides an electronic device. Specifically, the electronic device includes a processor and a storage device. The storage device stores a computer program, and when the computer program is executed by the processor, it executes the method described in any one of the above-mentioned embodiments.

[0056] Figure 4 A structural schematic diagram of an electronic device provided in an embodiment of the present application, the electronic device 400 includes: a processor 40, a memory 41, a bus 42 and a communication interface 43, wherein the processor 40, the communication interface 43 and the memory 41 are connected via the bus 42; the processor 40 is used to execute an executable module stored in the memory 41, such as a computer program.

[0057] The memory 41 may include a high-speed random access memory (RAM), and may also include a non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is realized through at least one communication interface 43 (which may be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. may be used.

[0058] The bus 42 may be an ISA bus, a PCI bus, or an EISA bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0059] Among them, the memory 41 is used to store programs, and the processor 40 executes the program after receiving the execution instruction. The method executed by the device for flow process definition disclosed in any embodiment of the above-mentioned embodiment of the present invention can be applied to the processor 40 or implemented by the processor 40.

[0060] The processor 40 may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the hardware integrated logic circuit or software instructions in the processor 40. The above processor 40 can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present invention can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor. The steps of the method disclosed in the embodiments of the present invention can be directly embodied as a hardware decoding processor to execute, or can be executed by a combination of hardware and software modules in the decoding processor. The software module can be located in a mature storage medium in the field 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 41, and the processor 40 reads the information in the memory 41 and completes the steps of the above method in combination with its hardware.

[0061] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0062] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0063] In addition, each functional unit in the embodiments provided in the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0064] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, electronic device, or network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0065] It should be noted that similar numbers and letters represent similar items in the accompanying drawings. Therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. In addition, the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance.

[0066] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present invention, which are used to illustrate the technical solutions of the present invention rather than to limit them. The protection scope of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the aforementioned embodiments, those of ordinary skill in the art should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed by the present invention, or can easily conceive of changes, or make equivalent replacements for some of the technical features therein. Such modifications, changes or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the protection scope of the present invention.

Claims

1. A method for deploying a machine learning model to predict online services, characterized in that: include: Use feature processing operators to process the pre-acquired offline data and generate a model training set; The operator of the feature processing includes a first online computing interface and a second online computing interface; The operator of the machine learning algorithm is trained using the model training set to generate a trained machine learning model; the operator of the algorithm of the trained machine learning model includes a first online computing interface; the operator of the feature processing includes a first operator and a second operator, and the operator of the machine learning algorithm includes a first operator; the first operator includes a first online computing interface and a first offline computing interface, and the second operator includes a second online computing interface and a second offline computing interface; the first operator is an Estimator type operator, and the second operator is a Transformer type operator; Determine an operator chain according to the feature processing operator and the trained machine learning model; the operator chain includes the feature processing operator and the trained machine learning model operator; Based on the first online computing interface and the second online computing interface in the operator link, a model online service is generated; the model online service is used to process real-time requested data and generate model calculation results.

2. The method for deploying a machine learning model prediction online service according to claim 1, characterized in that: Before the step of processing the features of the pre-acquired offline data to generate the model training set, the following steps are also included: Determine the operators for feature processing and the operators for machine learning algorithms; Obtain offline data for the modeling scenario and determine the offline modeling training set.

3. The method for deploying online service of machine learning model prediction according to claim 1, characterized in that: When performing offline calculation, the first operator generates a configuration file, where the configuration file is used to calculate the calculation logic of the first operator.

4. The method for deploying a machine learning model prediction online service according to claim 3, characterized in that: The feature processing includes: one or more of data cleaning, data filling and feature derivation; If the feature processing is the data cleaning or the data filling, the operator of the feature processing is the second operator; If the feature processing is feature derivation, the operator of the feature processing is the first operator, or the first operator and the second operator.

5. The method for deploying a machine learning model prediction online service according to claim 2, characterized in that: The operator chain includes a plurality of first operators and a plurality of second operators; The step of generating a model online service based on the first online computing interface and the second online computing interface in the operator link includes: Calling all the first operators and the second operators in the operator chain; A model online service is generated based on a first online computing interface of the first operator and a second online computing interface of the second operator.

6. A deployment device for predicting online services using a machine learning model, characterized in that: include: A feature processing module, used to perform feature processing on pre-acquired offline data using a feature processing operator to generate a model training set; the feature processing operator includes a first online computing interface and a second online computing interface; A training module, used to train the operator of the machine learning algorithm using the model training set to generate a trained machine learning model; the operator of the trained machine learning model includes a first online computing interface; A determination module, used to determine an operator chain according to the feature processing operator and the trained machine learning model; the operator chain includes the feature processing operator and the trained machine learning model operator; the feature processing operator includes a first operator and a second operator, and the operator of the machine learning algorithm includes a first operator; the first operator includes a first online computing interface and a first offline computing interface, and the second operator includes a second online computing interface and a second offline computing interface; the first operator is an Estimator type operator, and the second operator is a Transformer type operator; An online service generation module is used to generate a model online service based on the first online computing interface and the second online computing interface in the operator link; the model online service is used to process real-time requested data and generate model calculation results.

7. An electronic device comprising a memory and a processor, wherein the memory stores a computer program that can be run on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 5 are implemented.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores machine executable instructions. When the computer-executable instructions are called and executed by a processor, the computer-executable instructions prompt the processor to execute the method according to any one of claims 1 to 5.

Images