Method, system and storage medium for generating web page configuration information via low-code platform

Abstract

The present invention provides a method, system and storage medium for generating web page configuration information on a low-code platform, the method comprising: obtaining an application configuration data set of the low-code platform; cleaning the data set and completing missing data; dividing the cleaned data set into a training set and a test set; encoding the training set and the test set into word vectors; inputting the training set converted into word vectors into a variety of network models respectively, and comparing the loss function scores of each network model, selecting the model with the least score as the model to be optimized; evaluating the accuracy of the model to be optimized using the K-folding method, and adjusting the model parameters based on the accuracy to obtain the optimized model; inputting the test set converted into word vectors to evaluate the optimized model; and using the evaluated optimized model to configure the web page. This method can provide convenience for developers when using the low-code platform.

Description

Technical Field

[0001] The present invention relates to the technical field of low-code platform development, and specifically to a method, system and storage medium for generating web page configuration information by a low-code platform. Background Art

[0002] At present, the technology for building low-code platforms is developing rapidly, and the applications built through low-code platforms are booming, which enables various enterprises to quickly build applications that meet their needs with little or no intervention from developers. The core logic of the implementation of low-code platforms is to design business operation rules on low-code platforms based on customized business logic, and then the low-code platform generates corresponding code based on the business rules.

[0003] The process of building your own exclusive application through a low-code platform has been basically fixed after years of development. However, there are still many problems in the implementation of the low-code platform itself that have not yet formed a mature solution. For the code generation part, the current main process is that users add configurations to their own systems, such as adding page configurations for their own web applications, users add certain buttons to the page, create certain forms, and append certain fields to the form. Then the low-code platform generates the code corresponding to the user configuration based on the user's configuration. In the actual use process, if the low-code platform is used by developers, most of them will still feel that the operation method brought by the low-code platform framework is a restriction rather than a help to their development process; and for non-developers, the configuration of some functions is difficult, and even because the configuration difficulty is too high, they still cannot operate at all even if there are operation documents. Summary of the invention

[0004] One of the main problems that the present invention solves is that the current low-code platform is difficult to develop and apply, some functions are difficult to configure, and ordinary users are difficult to operate.

[0005] According to one aspect of the present invention, the present invention provides a method for generating web page configuration information on a low-code platform, the method comprising:

[0006] Obtain application configuration data sets for low-code platforms;

[0007] Clean the data set and fill in missing data;

[0008] Dividing the cleaned data set into a training set and a test set;

[0009] Convert the training set and the test set encoding into word vectors;

[0010] The training set converted into word vectors is input into various network models respectively, and the loss function scores of each network model are compared, and the model with the least score is selected as the model to be optimized;

[0011] Using the K-fold method to verify the accuracy of the model to be optimized, and adjust the parameters of the model to be optimized;

[0012] The input is converted into a test set of word vectors to evaluate the model to be optimized after adjusting the parameters, and the optimized model is obtained;

[0013] The optimization model is used to configure the web page.

[0014] Furthermore, after the webpage configuration information is generated by using the optimization model, the following steps are also included:

[0015] When the application collects new user configuration information, the new user configuration information is stored in the application configuration data set to obtain a new application configuration data set;

[0016] The optimization model is iteratively trained using the new application configuration dataset.

[0017] Furthermore, the multiple network models include: a decision tree model, a random forest model, a maximum margin support vector machine model and a neural network model.

[0018] Furthermore, obtaining the application configuration data set of the low-code platform includes:

[0019] The data pairs consisting of industry labels and application structure nodes are used as sample data;

[0020] Using the specific configuration value of the application project in the structure node as the label of the sample data;

[0021] constructing the sample data and the label of the sample data into a sample data pair;

[0022] The set of sample data pairs is the application configuration data set.

[0023] Furthermore, cleaning the data set and completing missing data includes: assigning default labels to sample data with missing labels in the data set in batches.

[0024] Further, dividing the cleaned data set into a training set and a test set includes:

[0025] Each time the optimization model is run, the same sample data is divided into the sample set divided in the previous model run.

[0026] Furthermore, before storing the new user configuration information into the application configuration data set, the method further includes:

[0027] Manually screening the new user configuration information to obtain screening configuration information;

[0028] When the data volume of the screening configuration information is sufficient, the screening configuration information is mixed with the application configuration data set.

[0029] According to another aspect of the present invention, a system for generating web page configuration information via a low-code platform is also disclosed. When the system is running, the system is used to implement the steps of any of the above methods for generating web page configuration information via a low-code platform. The system includes:

[0030] The data collection module is used to obtain the application configuration data set of the low-code platform;

[0031] A data cleaning module, used to clean the application configuration data set and complete missing data;

[0032] A data partitioning module, used to partition the cleaned application configuration data set into a training set and a test set;

[0033] An encoding module, used for encoding the training set and the test set into word vectors;

[0034] A training module, used to input the training set converted into word vectors into multiple network models respectively, and compare the loss function scores of each network model, select the model with the least score as the model to be optimized, and use the K-fold method to evaluate the accuracy of the model to be optimized, and adjust the parameters of the model to be optimized based on the accuracy to obtain an optimized model;

[0035] The testing module is used to input the test set converted into word vectors to evaluate the optimization model.

[0036] According to another aspect of the present invention, a storage medium is also disclosed, which is a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is run, it is used to implement the steps of a method for generating web page configuration information by a low-code platform as described in any of the above.

[0037] The present invention introduces a machine learning program into the development process of the low-code platform, so that the program can generate subsequent configurations inferentially through past data after the user performs a small amount of configuration, and try to avoid the parts with high configuration difficulty. This makes it more convenient for developers to use the low-code platform, while lowering the threshold for non-developers to use it; secondly, industry tags are introduced to the low-code platform, so that users can get smarter prompt configurations when they select the industry and then configure the application. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The accompanying drawings, which constitute a part of the specification, illustrate embodiments of the invention and, together with the description, serve to explain the principles of the invention.

[0039] Figure 1 A flowchart of a method for generating web page configuration information on a low-code platform in an embodiment of the present invention.

[0040] Figure 2 This is a schematic diagram of the structure of a web application built through a low-code platform in an embodiment of the present invention. DETAILED DESCRIPTION

[0041] Various exemplary embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings. It should be noted that unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions and numerical values ​​set forth in these embodiments do not limit the scope of the present invention.

[0042] At the same time, it should be understood that for the convenience of description, the sizes of the various parts shown in the drawings are not drawn according to the actual proportional relationship.

[0043] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the invention, its application, or uses.

[0044] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with specific embodiments and with reference to the accompanying drawings.

[0045] Technologies, methods, and apparatus known to ordinary technicians in the relevant field may not be discussed in detail, but where appropriate, such technologies, methods, and apparatus should be considered part of the authorization specification.

[0046] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0047] It should be noted that like reference numerals and letters refer to similar items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0048] Embodiment 1, as Figure 1 The figure shows a flow chart of a method for generating web page configuration information on a low-code platform. The specific steps are as follows:

[0049] Step 1: Create or obtain an application configuration data set. Generally speaking, this type of data is rare and is created manually. To facilitate subsequent calculations and processing, the data can be stored in a relational database as an application configuration data set. As for the data model of the data set, since the low-code platform itself is also a computer program, the applications built on it must conform to the corresponding program model.

[0050] For web applications built through the platform, the structure is roughly as follows Figure 2 As shown, it includes projects, modules, functions and pages. The page includes function point configuration and list configuration. The function table configuration includes form design. The form design includes prompt information configuration and data source field configuration.

[0051] In some embodiments, in other web applications, other web configuration function nodes besides the aforementioned functions may also be included.

[0052] Based on the above structure, (industry label + application structure node) can be used as sample data, where the i-th sample data can be recorded as x i ; The specific configuration value of the project in the structure node is used as the label of the sample data, where the label of the i-th sample data is recorded as y i Based on this, we can construct the sample data pair (x i ,y i ), thus obtaining a data set (X, Y) consisting of sample data pairs, where X is x i The set of y i A collection of .

[0053] Step 2: Create a custom data cleaning program to clean the data set obtained in step 1. Due to the differences in business between applications, there may be cases where sample data does not contain industry labels. In this case, you can develop a custom program to batch assign default labels to sample data with missing labels to maintain the consistency of the sample data structure. You can also clean the data using conventional methods, such as setting the default label directly to the string "None", or directly removing samples without labels, or using conventional zero-padding operations to set the label to 0.

[0054] Step 3: Divide the data set. Randomly divide the sample data into training set and test set in a ratio of 8:2.

[0055] In some embodiments, avoid dividing different data into two types of sample sets each time the optimization model is applied, or divide the same data into the same sample set divided before the previous model training. For example, there are 10 samples, numbered 1 to 10, and it should be avoided that when the program is run for the first time, samples numbered 1 to 7 are divided into the training set, and when the program is run for the second time, samples divided into the training set are not numbered 1 to 7. Instead, it should be ensured that each time the optimization model is run, samples numbered 1 to 7 are divided into the training set.

[0056] Step 4: Encode the divided training set and test set data. The core of data encoding is to regard the sample data as word vectors in natural language processing (NLP) tasks, that is, to perform one-hot encoding on individual characters and convert the entire sample data into word vectors.

[0057] Step 5: Select some candidate models. The candidate models include decision tree model, random forest model, SVM (maximum margin support vector machine) and neural network model. After inputting the training set into the aforementioned models and running the training, the model with the smallest loss function value after training is selected for subsequent model tuning.

[0058] In some embodiments, after selecting the model with the smallest loss function value, the accuracy of the selected model is further evaluated by the K-fold method, and the model parameters are fine-tuned based on the evaluated accuracy. The K-fold method is to divide the training set into K sample sets, randomly select (K-1) sample sets as training samples each time, and select the remaining one as the verification sample set, so that the model can be trained and evaluated multiple times. The K-fold method will eventually output the scores of the K loss functions for subsequent analysis.

[0059] Step 6: After the model is fine-tuned, it is evaluated on the test set converted to word vectors in step 4 to obtain the optimized model. At this point, even if the model does not perform well on the final test set, it should not be adjusted further, because even if the model is adjusted to perform well on the final test set, it may not generalize to real data.

[0060] Step 7: After the model is finally evaluated to obtain the optimized model, the optimized model can be published and deployed for application.

[0061] In some embodiments, if the user agrees to the development program to collect user configuration data and send the user configuration data back to the server, the user configuration data is stored in the manner of step 1 of this solution. After the user configuration data is sent back to the server, it is necessary to manually screen the data, that is, to manually evaluate and eliminate data with obvious errors. When the amount of data is valid and sufficient for the next round of training, the data is mixed with the original data for secondary training or multiple trainings to further iterate and upgrade the model.

[0062] Embodiment 2, a system for generating web page configuration information on a low-code platform, the system comprising: a data collection module, used to obtain an application configuration data set of the low-code platform; a data cleaning module, used to clean the application configuration data set and complete missing data; a data partitioning module, used to divide the cleaned application configuration data set into a training set and a test set; an encoding module, used to encode the training set and the test set into word vectors; a training module, used to input the training set converted into word vectors into multiple network models respectively, and compare the loss function scores of each network model, select the model with the least score as the model to be optimized, and use the K-fold method to evaluate the accuracy of the model to be optimized, fine-tune the model parameters based on the accuracy, so as to obtain an optimized model; a testing module, used to input the test set converted into word vectors to evaluate the optimized model.

[0063] Embodiment three also discloses a storage medium, which is a computer-readable storage medium. A computer program is stored on the computer-readable storage medium. When the computer program is running, it is used to implement the steps of the method for generating web page configuration information by any low-code platform as described in the previous embodiments.

[0064] Compared with the existing technology, the advantages of the method for generating web page configuration information by a low-code platform in the present invention are:

[0065] 1. Machine learning tasks are introduced to help solve the problem that users need to perform a lot of configuration before code generation. For developers, the inferred generation of configuration information is very similar to the code completion function in various IDEs (integrated development environments) that developers use daily. Therefore, this method can provide convenience for developers when using low-code platforms.

[0066] 2. For non-developers, the existence of inferred configuration information can lower their threshold for using low-code platforms.

[0067] 3. Treat the user's configuration itself as a sample label, and encode the prompt information according to the encoding method of the word vector. With the further use of the optimization model, the model can be further trained and upgraded using the data on the web page configuration voluntarily provided by the user, that is, the configuration information can be inferred, making the optimization model more accurate as it is used.

[0068] 4. Combine machine learning tasks with code configuration generation tasks. When the environment changes, such as when the configurations of most applications in a certain field slowly begin to follow another set of rules, the program can use machine learning to gradually bring the inferred configuration information closer to the configuration information corresponding to the other set of rules without human intervention.

[0069] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

[0070] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity 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, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. A method for generating web page configuration information on a low-code platform, characterized in that: The method comprises: Obtain the application configuration data set of the low-code platform; including: The data pairs consisting of industry labels and application structure nodes are used as sample data; Using the specific configuration value of the application project in the structure node as the label of the sample data; constructing the sample data and the label of the sample data into a sample data pair; The set of sample data pairs is the application configuration data set; Clean the data set and fill in missing data; Dividing the cleaned data set into a training set and a test set; Convert the training set and the test set encoding into word vectors; The training set converted into word vectors is input into various network models respectively, and the loss function scores of each network model are compared, and the model with the least score is selected as the model to be optimized; Using the K-fold method to verify the accuracy of the model to be optimized, and adjust the parameters of the model to be optimized; The input is converted into a test set of word vectors to evaluate the model to be optimized after adjusting the parameters, and the optimized model is obtained; The optimization model is used to configure the webpage, and when the application collects new user configuration information, the new user configuration information is stored in the application configuration data set to obtain a new application configuration data set; The optimization model is iteratively trained using the new application configuration dataset.

2. A method for generating web page configuration information using a low-code platform as claimed in claim 1, characterized in that: The multiple network models include: a decision tree model, a random forest model, a maximum margin support vector machine model and a neural network model.

3. A method for generating web page configuration information using a low-code platform as claimed in claim 1, characterized in that: Cleaning the data set and filling in missing data includes: Assign default labels to sample data with missing labels in the dataset in batches.

4. A method for generating web page configuration information using a low-code platform as claimed in claim 1, characterized in that: The cleaned data set is divided into a training set and a test set including: Each time the optimization model is run, the same sample data is divided into the sample set divided in the previous model run.

5. A method for generating web page configuration information using a low-code platform as claimed in claim 1, characterized in that: Before storing the new user configuration information into the application configuration data set, the method further includes: Manually screening the new user configuration information to obtain screening configuration information; When the data volume of the screening configuration information is sufficient, the screening configuration information is mixed with the application configuration data set.

6. A system for generating web page configuration information on a low-code platform, characterized in that: When the system is running, the system is used to implement the steps of a method for generating web page configuration information by a low-code platform as described in any one of claims 1 to 5, and the system includes: The data collection module is used to obtain the application configuration data set of the low-code platform; including: The data pairs consisting of industry labels and application structure nodes are used as sample data; Using the specific configuration value of the application project in the structure node as the label of the sample data; constructing the sample data and the label of the sample data into a sample data pair; The set of sample data pairs is the application configuration data set; A data cleaning module, used to clean the application configuration data set and complete missing data; A data partitioning module, used to partition the cleaned application configuration data set into a training set and a test set; An encoding module, used for encoding the training set and the test set and converting them into word vectors; A training module, used to input the training set converted into word vectors into multiple network models respectively, and compare the loss function values ​​of each network model, select the model with the smallest loss function value as the model to be optimized, and use the K-fold method to evaluate the accuracy of the model to be optimized, and adjust the parameters of the model to be optimized based on the accuracy; The testing module is used to input the test set converted into word vectors to evaluate the model to be optimized after adjusting the parameters, so as to obtain the optimized model; the optimized model is used to configure the web page, and when the application collects new user configuration information, the new user configuration information is stored in the application configuration data set to obtain a new application configuration data set; The optimization model is iteratively trained using the new application configuration dataset.

7. A storage medium, characterized in that: The storage medium is a computer-readable storage medium, on which a computer program is stored. When the computer program is run, it is used to implement the steps of a method for generating web page configuration information by a low-code platform as described in any one of claims 1-5.

Images