A User Intent-Based Intelligent Application Development Method

Abstract

This invention discloses an application intelligent development method based on user intent, relating to the field of application intelligent development. The specific method comprises: Step 1, establishing a data management platform for managing and storing application software data; Step 2, analyzing various types of application software data within the data management platform; Step 3, developers implementing basic application software development based on a visual application software editor; Step 4, calculating the performance evaluation indicators of the initial application software model and identifying functional components that can cause the performance evaluation indicators to rise or fall; Step 5, refining the application software development based on existing LLM model technology; and Step 6, developers previewing and debugging the application software to confirm whether to release it. The overall architecture design of this invention greatly simplifies the application software development process, allowing developers to quickly build application frameworks using low-code methods while utilizing the LLM model provided by the platform for deep customization and optimization.

Description

Technical Field

[0001] This invention belongs to the field of application intelligence development technology, specifically relating to an application intelligence development method based on user intent. Background Technology

[0002] With the accelerating pace of digital transformation, enterprises are increasingly demanding the rapid construction and iteration of application systems. Low-Code Development Platforms (LCDPs) have emerged to address this need, aiming to lower the technical barriers to software development and accelerate application delivery cycles through graphical interfaces and minimal code writing. However, despite the significant potential of LCDPs in improving development efficiency, existing platforms still face numerous challenges in practical applications, limiting their widespread adoption and in-depth optimization in complex business scenarios.

[0003] Due to limitations in platform architecture and component design, developers often find it difficult to deeply customize or extend components. This results in applications that struggle to fully adapt to complex and ever-changing business logic and processes, limiting their flexibility and adaptability. As application scale and user volume increase, the demands on system performance also rise. However, existing low-code application development platforms suffer from insufficient flexibility, difficulty in performance optimization, and complex module coupling, making it difficult to improve application deployment for real-world business scenarios. Most platforms offer components that are difficult to deeply customize and optimize for specific application types, resulting in applications that fail to meet the needs of complex and ever-changing application scenarios. Furthermore, existing platforms generally lack performance evaluation for applications handling varying numbers of users and large-scale request volumes, often leading to performance bottlenecks and system response delays when dealing with massive user and request data in the application domain. Summary of the Invention

[0004] The purpose of this invention is to provide an application intelligent development method based on user intent, which solves the technical problems in the prior art of difficulty in deeply extending components and lack of performance evaluation of application software.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] Step 1: Establish a data management platform for managing and storing application software data. This involves creating a database containing various types of application software data. This is achieved by collecting publicly available data from different application software types, providing APIs or web interfaces for developers to upload private data related to their applications, verifying the collected data to ensure its integrity and accuracy, and obtaining various types of application software data. Based on standard regulations for application software data (standard regulations refer to a series of specifications and guidelines for different application software types formulated by authoritative industry organizations), some application software data that does not meet the standards is eliminated. The remaining application software data is then categorized by type, and each type of application software data is sorted according to the number of users of the corresponding application software. Based on the content of each type of application software data, the corresponding data table structure is configured in the database. The data table content is then uploaded to the data management platform. The data management platform parses and stores the data table content and manages it through a visual interface. Application software developers can enable users to access and view relevant data authorized by the developers.

[0007] Step 2: Analyze application software data of various types within the data management platform; based on the application software data, analyze the application software itself, extract the front-end and back-end functional components of each type of application software, and divide them into main functional components and secondary functional components. Secondary functional components refer to components whose assigned functions are extensions of the main functional components. For example, in e-commerce application software, product search is a main functional component, while product photo search and product name search are secondary functional components of product search. Analyze and compare the data of multiple application software of the same type, comparing the frequency indicators of the main functional components of the multiple application software of the same type. Determine the main functional components necessary for the development of this type of application software as the primary functional components, the main functional components with higher frequency indicators as regular main functional components, and the main functional components with lower frequency indicators as special main functional components. For n application software belonging to type 'a', ensure that these n application software meet the standards, are representative, and are updated in real time. Count the number of main functional components of all application software as x, and label the corresponding number of each main functional component, obtaining the set of main functional components as {b1, b2, ..., b...}. x The specific formula for comparing the frequency of occurrence of functional components is as follows:

[0008]

[0009] Where x represents the main functional component number x, F(x) represents the functional component frequency, n(x) represents that there are n(x) application software programs with the main functional component numbered x, i represents the i-th application software, and m i This represents the number of registered users for the i-th application. xAn application software representing a main functional component with number x.

[0010] Determine the functional component frequency thresholds as A and B, calculate the functional component frequency of each main functional component. When F(x) < A, it represents that the main functional component numbered x belongs to a special main functional component. When A ≤ F(x) < B, it represents that the main functional component numbered x belongs to a regular main functional component. When B ≤ F(x), it represents that the main functional component numbered x belongs to a main body functional component. The functional component frequency thresholds A and B are determined according to the functional component frequency set. By determining the true values of A and B, obtain the number N of main body functional components, the number M of regular functional components, and the number L of special components, ensuring that L < N < M and L + N > M.

[0011] Based on the data of the same type of application software, record all the extendable sub-functional components possessed by each main functional component. By calculating the functional component frequency of the sub-functional components belonging to each main functional component, divide the sub-functional components within the extendable range of the main functional component. Use the division method for the frequency index of the main functional component appearance to divide the sub-functional components into main body sub-functional components, regular sub-functional components, and special sub-functional components in the same way, and record the associated logic configuration methods of all front-end functional components to establish different types of visual application software editors.

[0012] Step 3: Developers implement the basic development of the application software based on the visual application software editor; determine the type of application software that the developer needs to develop, select the visual editor corresponding to the application software type. Based on the main body functional components and the main body sub-functional components, determine the application template for this type of application software. This application template is preset in the visual editor and includes all the main body functional components, main body sub-functional components, and the configuration methods of the associated logic between components in this type. Each type of visual editor provides an application template and a blank template for developers to choose and edit. The template page provides functions for adding, deleting, and layout design of functional components. Each functional component can only exist one in the template. Developers drag and drop components for editing, place them in the layout, and configure the data source of the data management platform for the components to correspond to the data tables of the management platform, so as to display or input data. Developers configure the styles of each page according to personalized needs, such as the page theme color, the color matching of each component, the font size and color, etc. Developers configure the logic between pages and the associated logic between the parent page and the child page. Based on the visual application software editor, help developers complete the limitation of the application software function and the design of the framework layout to obtain the initial application software model.

[0013] Step 4: Calculate the performance evaluation metrics of the initial application software model and identify the functional components that can cause the performance evaluation metrics to increase or decrease. Based on the application type of the initial application software model, determine the simulation experiment objectives. Based on the simulation experiment objectives, determine which functional components and configuration logic of the initial application software model will participate in the simulation. For example, for e-commerce application software, the simulation experiment objective is to evaluate the maximum average user request frequency or number of users that the application software model can withstand without affecting the response time, throughput, and concurrent users. Based on the simulation experiment objectives, the product search function component will participate in the simulation. Use relevant simulation tools and platforms to simulate user behavior and system load. For example, use JMeter or LoadRunner simulation tools to simulate user behavior or system load. Let H represent the number of users in the initial application software model, and the average request frequency per user be O. Set the average response time, throughput, and error rate as performance evaluation metrics to evaluate the performance of the initial application software model. The specific performance evaluation metrics set within the simulation duration are shown below:

[0014]

[0015] Where T represents the average response time for a user request. R represents the simulation duration. The throughput within the system is given by w, representing the error rate. Based on the simulation experiment objectives, relevant variables in the performance evaluation metrics are adjusted to obtain changes in the performance evaluation metrics related to the simulation experiment objectives. For example, by gradually increasing the number of users or request frequency, the initial application software is simulated under high load. The changes in system performance are observed and recorded to obtain the critical point where performance begins to decline, thus determining the performance limit of the initial application software model. Different functional components of this type are added or removed, one at a time. This functional component is either a slave component or a master component without slave components. Simulations are then performed. Through simulation experiments, functional components in the initial application software model that cause a decline in performance evaluation metrics are identified and designated as replaceable components. The performance evaluation metrics of the initial application software model are calculated and compared with the performance metrics of the application software model after removing replaceable components. The substitutable components are sorted according to the magnitude of their performance improvement, resulting in a sequence of substitutable components. Through simulation experiments, functional components not present in the initial application software model but whose performance improvement increases after addition are selected and designated as optimized components. The performance evaluation indicators of the initial application software model and the optimized components are calculated. These optimized components are then sorted again according to the magnitude of their performance improvement, resulting in a sequence of optimized components. Based on the substitutable and optimized component sequences, random optimization schemes are generated by randomly removing substitutable components and adding optimized components. The performance evaluation indicators of these random optimization schemes are calculated. Considering the functional consistency between the random optimization scheme and the initial application software model, the optimal random optimization scheme is selected. The specific formula for calculating the functional consistency between the random optimization scheme and the initial application software model is shown below:

[0016]

[0017] Where J represents the number of main functional components in the initial application software model. Let represent the number of components in the stochastic optimization scheme that have the same or similar functions as the main functional components of the initial application software model, denoted as retained components. Let z represent the z-th retained component. In the initial application software model, main functional components that are the same or similar to z are denoted as corresponding components. The z-th retained component corresponds to the z-th corresponding component. Let j(z) represent the number of subordinate functional components shared by the z-th corresponding component. This represents the number of sub-functional components that have the same or similar functions as the z-th retained component and its corresponding component.

[0018] Using the formula YX(r)=I(r)[Y(r)-Y(0)], the performance evaluation index of the stochastic optimization scheme is calculated. Combining the functional consistency between the stochastic optimization scheme and the initial application software model, the best stochastic optimization scheme is selected, where r represents the r-th stochastic optimization scheme, Y(0) represents the performance evaluation index of the initial application software model, and Y(r) represents the performance evaluation index of the r-th stochastic optimization scheme. The stochastic optimization scheme with the largest YX(r) value is selected as the best stochastic optimization scheme.

[0019] Step 5: Develop and improve application software based on existing LLM model technology; Developers design and modify the improved application software model data, which is then uploaded to the big data management platform. The platform allows software users to access and view relevant data authorized by the developers. The big data management platform embeds advanced LLM models, such as GPT and Codex, and uses APIs provided by cloud service providers to access the models. A front-end interface is designed for developers, allowing them to input requirement descriptions and view generated code suggestions. On the back-end, logic is written to process developer input, call the LLM model API, receive responses, and perform probabilistic post-processing, such as code formatting and output optimization. By using asynchronous calls to the LLM model, the LLM model's performance is improved. The LLM model's response speed is assessed by collecting sample code and use cases for commonly used functional components and component sets across various application software types. This serves as a reference to help developers understand how to use the big data management platform. For code written by the developer or obtained through other channels, the big data management platform provides real-time syntax checking, error messages, and code preview after the developer uploads the code. It also integrates breakpoint debugging functionality, allowing developers to step through the code, observe variable states, locate and resolve issues. Developers can rate and provide feedback on the code suggestions generated by the LLM. The platform collects user evaluations of various aspects of the application software before its release and regularly assesses and optimizes the LLM model's response speed and code generation quality to ensure the efficient operation of the LLM model.

[0020] Step Six: Developers preview and debug the application software to confirm whether to release it. The code obtained from or improved using the LLM model is input into the designated free code component area in the visual editor. The visual editor parses the input code, recognizing its format, such as HTML structure, CSS styles, and JavaScript logic. Based on the parsing results, the editor automatically generates corresponding visual components on the visual editing interface. These components will display their appearance and functions according to the definitions in the code. Developers preview the component display in the editor to ensure that each component meets the design requirements. The visual editor provides flexible layout tools, allowing developers to adjust the component layout by dragging and scaling. Various operations are reflected in real time on the visual interface and the code. Based on the synchronization between the code and the visual interface, the visual editor will mark components that are inconsistent between the code and the visual interface to alert developers of any discrepancies. When developers are satisfied with the application software's display in the visual editor, they can choose to release the application software to the corresponding platform. For developers who are not satisfied with the display in the visual editor, additional configuration and optimization work is required until the developers are satisfied with the application software's display.

[0021] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are:

[0022] 1. The present invention proposes an intelligent application development method based on user intent. By introducing a customizable LLM model and deployment framework, developers can flexibly build and optimize relevant modules according to specific application business needs. The model generates code blocks and components based on specific business application scenarios, enabling developers to quickly iterate and optimize the LLM model for specific business scenarios and improve the flexibility and targeting of application development.

[0023] 2. This invention classifies the functional component types of different types of application software based on the frequency of occurrence index, and presets different application software templates according to the application type. The application software templates serve as reference templates for developers to customize application software. The application software templates determine the most basic functions and operations for implementing the application type, helping developers with a low level of understanding of application types to make innovative modifications based on the application software templates.

[0024] 3. This invention uses relevant simulation tools and platforms to simulate user behavior and system load, and comprehensively evaluates the performance of application software under different user behaviors and system loads by considering average response time, throughput, and error rate. By adding or removing different functional components on the basis of application software, the impact of each functional component on performance evaluation indicators is determined, helping developers adjust the functional component structure of application software. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 A flowchart illustrating the steps of an application intelligence development method based on user intent is shown.

[0027] Figure 2 The functional flowchart of the low-code visual editor of the present invention is shown;

[0028] Figure 3 The functional flowchart of the proposed LLM model is shown. Detailed Implementation

[0029] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0030] Example 1: Figure 1 , Figure 2 , Figure 3 As shown, this embodiment provides an application intelligence development method based on user intent, and the specific method steps include:

[0031] Step 1: Establish a data management platform for managing and storing application software data. This involves creating a database containing various types of application software data. This is achieved by collecting publicly available data from different application software types, providing APIs or web interfaces for developers to upload private data related to their applications, verifying the collected data to ensure its integrity and accuracy, and obtaining various types of application software data. Based on standard regulations for application software data (standard regulations refer to a series of specifications and guidelines for different application software types formulated by authoritative industry organizations), some application software data that does not meet the standards is eliminated. The remaining application software data is then categorized by type, and each type of application software data is sorted according to the number of users of the corresponding application software. Based on the content of each type of application software data, the corresponding data table structure is configured in the database. The data table content is then uploaded to the data management platform. The data management platform parses and stores the data table content and manages it through a visual interface. Application software developers can enable users to access and view relevant data authorized by the developers.

[0032] Step 2: Analyze application software data of various types within the data management platform; based on the application software data, analyze the application software itself, extract the front-end and back-end functional components of each type of application software, and divide them into main functional components and secondary functional components. Secondary functional components refer to components whose assigned functions are extensions of the main functional components. For example, in e-commerce application software, product search is a main functional component, while product photo search and product name search are secondary functional components of product search. Analyze and compare the data of multiple application software of the same type, comparing the frequency indicators of the main functional components of the multiple application software of the same type. Determine the main functional components necessary for the development of this type of application software as the primary functional components, the main functional components with higher frequency indicators as regular main functional components, and the main functional components with lower frequency indicators as special main functional components. For n application software belonging to type 'a', ensure that these n application software meet the standards, are representative, and are updated in real time. Count the number of main functional components of all application software as x, and label the corresponding number of each main functional component, obtaining the set of main functional components as {b1, b2, ..., b...}. x The specific formula for comparing the frequency of occurrence of functional components is as follows:

[0033]

[0034] Where x represents the main functional component number x, F(x) represents the functional component frequency, n(x) represents that there are n(x) application software programs with the main functional component numbered x, i represents the i-th application software, and m i This represents the number of registered users for the i-th application. xAn application software representing a main functional component with number x.

[0035] Determine the functional component frequency thresholds as A and B, calculate the functional component frequency of each main functional component. When F(x) < A, it means the main functional component numbered x belongs to a special main functional component. When A ≤ F(x) < B, it means the main functional component numbered x belongs to a regular main functional component. When B ≤ F(x), it means the main functional component numbered x belongs to a main body functional component. Here, the functional component frequency thresholds A and B are determined according to the functional component frequency set. By determining the true values of A and B, obtain the number N of main body functional components, the number M of regular functional components, and the number L of special components, ensuring that L < N < M and L + N > M.

[0036] Based on the data of the same type of application software, record all the extendable subordinate functional components possessed by each main functional component. By calculating the functional component frequency of the subordinate functional components belonging to each main functional component, divide the subordinate functional components within the extendable range of the main functional component. Use the division method for the frequency index of the main functional component appearance to divide the subordinate functional components into main body subordinate functional components, regular subordinate functional components, and special subordinate functional components in the same way, and record the associated logic configuration methods of all front - end functional components to establish different types of visual application software editors.

[0037] Step 3: Developers implement the basic development of the application software based on the visual application software editor; determine the type of application software that the developers need to develop, select the corresponding visual editor for the application software type. Based on the main body functional components and the main body subordinate functional components, determine the application template for this type of application software. This application template is preset in the visual editor and includes all the main body functional components, main body subordinate functional components, and the configuration method of the associated logic between components in this type. Each type of visual editor provides an application template and a blank template for developers to choose and edit. The template page provides functions for adding, deleting, and layout design of functional components. Each functional component can only exist once in the template. Developers drag and drop components for editing, place them in the layout, and configure the data source of the data management platform for the components to correspond to the data tables in the management platform, so as to display or input data. Developers configure the styles of each page according to personalized needs, such as the page theme color, the color matching of each component, the font size and color, etc. Developers configure the logic between pages and the associated logic between the parent page and the child page. Based on the visual application software editor, help developers complete the limitation of the application software functions and the design of the framework layout, and obtain the initial application software model.

[0038] Step 4: Calculate the performance evaluation metrics of the initial application software model and identify the functional components that can cause the performance evaluation metrics to increase or decrease. Based on the application type of the initial application software model, determine the simulation experiment objectives. Based on the simulation experiment objectives, determine which functional components and configuration logic of the initial application software model will participate in the simulation. For example, for e-commerce application software, the simulation experiment objective is to evaluate the maximum average user request frequency or number of users that the application software model can withstand without affecting the response time, throughput, and concurrent users. Based on the simulation experiment objectives, the product search function component will participate in the simulation. Use relevant simulation tools and platforms to simulate user behavior and system load. For example, use JMeter or LoadRunner simulation tools to simulate user behavior or system load. Let H represent the number of users in the initial application software model, and the average request frequency per user be O. Set the average response time, throughput, and error rate as performance evaluation metrics to evaluate the performance of the initial application software model. The specific performance evaluation metrics set within the simulation duration are shown below:

[0039]

[0040] Where T represents the average response time for a user request. R represents the simulation duration. The throughput within the system is given by w, representing the error rate. Based on the simulation experiment objectives, relevant variables in the performance evaluation metrics are adjusted to obtain changes in the performance evaluation metrics related to the simulation experiment objectives. For example, by gradually increasing the number of users or request frequency, the initial application software is simulated under high load. The changes in system performance are observed and recorded to obtain the critical point where performance begins to decline, thus determining the performance limit of the initial application software model. Different functional components of this type are added or removed, one at a time. This functional component is either a slave component or a master component without slave components. Simulations are then performed. Through simulation experiments, functional components in the initial application software model that cause a decline in performance evaluation metrics are identified and designated as replaceable components. The performance evaluation metrics of the initial application software model are calculated and compared with the performance metrics of the application software model after removing replaceable components. The substitutable components are sorted according to the magnitude of their performance improvement, resulting in a sequence of substitutable components. Through simulation experiments, functional components not present in the initial application software model but whose performance improvement increases after addition are selected and designated as optimized components. The performance evaluation indicators of the initial application software model and the optimized components are calculated. These optimized components are then sorted again according to the magnitude of their performance improvement, resulting in a sequence of optimized components. Based on the substitutable and optimized component sequences, random optimization schemes are generated by randomly removing substitutable components and adding optimized components. The performance evaluation indicators of these random optimization schemes are calculated. Considering the functional consistency between the random optimization scheme and the initial application software model, the optimal random optimization scheme is selected. The specific formula for calculating the functional consistency between the random optimization scheme and the initial application software model is shown below:

[0041]

[0042] Where J represents the number of main functional components in the initial application software model. Let represent the number of components in the stochastic optimization scheme that have the same or similar functions as the main functional components of the initial application software model, denoted as retained components. Let z represent the z-th retained component. In the initial application software model, main functional components that are the same or similar to z are denoted as corresponding components. The z-th retained component corresponds to the z-th corresponding component. Let j(z) represent the number of subordinate functional components shared by the z-th corresponding component. This represents the number of sub-functional components that have the same or similar functions as the z-th retained component and its corresponding component.

[0043] Using the formula YX(r)=I(r)[Y(r)-Y(0)], the performance evaluation index of the stochastic optimization scheme is calculated. Combining the functional consistency between the stochastic optimization scheme and the initial application software model, the best stochastic optimization scheme is selected, where r represents the r-th stochastic optimization scheme, Y(0) represents the performance evaluation index of the initial application software model, and Y(r) represents the performance evaluation index of the r-th stochastic optimization scheme. The stochastic optimization scheme with the largest YX(r) value is selected as the best stochastic optimization scheme.

[0044] Step 5: Develop and improve application software based on existing LLM model technology; Developers design and modify the improved application software model data, which is then uploaded to the big data management platform. The platform allows software users to access and view relevant data authorized by the developers. The big data management platform embeds advanced LLM models, such as GPT and Codex, and uses APIs provided by cloud service providers to access the models. A front-end interface is designed for developers, allowing them to input requirement descriptions and view generated code suggestions. On the back-end, logic is written to process developer input, call the LLM model API, receive responses, and perform probabilistic post-processing, such as code formatting and output optimization. By using asynchronous calls to the LLM model, the LLM model's performance is improved. The LLM model's response speed is assessed by collecting sample code and use cases for commonly used functional components and component sets across various application software types. This serves as a reference to help developers understand how to use the big data management platform. For code written by the developer or obtained through other channels, the big data management platform provides real-time syntax checking, error messages, and code preview after the developer uploads the code. It also integrates breakpoint debugging functionality, allowing developers to step through the code, observe variable states, locate and resolve issues. Developers can rate and provide feedback on the code suggestions generated by the LLM. The platform collects user evaluations of various aspects of the application software before its release and regularly assesses and optimizes the LLM model's response speed and code generation quality to ensure the efficient operation of the LLM model.

[0045] Step Six: Developers preview and debug the application software to confirm whether to release it. The code obtained from or improved using the LLM model is input into the designated free code component area in the visual editor. The visual editor parses the input code, recognizing its format, such as HTML structure, CSS styles, and JavaScript logic. Based on the parsing results, the editor automatically generates corresponding visual components on the visual editing interface. These components will display their appearance and functions according to the definitions in the code. Developers preview the component display in the editor to ensure that each component meets the design requirements. The visual editor provides flexible layout tools, allowing developers to adjust the component layout by dragging and scaling. Various operations are reflected in real time on the visual interface and the code. Based on the synchronization between the code and the visual interface, the visual editor will mark components that are inconsistent between the code and the visual interface to alert developers of any discrepancies. When developers are satisfied with the application software's display in the visual editor, they can choose to release the application software to the corresponding platform. For developers who are not satisfied with the display in the visual editor, additional configuration and optimization work is required until the developers are satisfied with the application software's display.

[0046] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

[0047] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to specific implementations. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A user intent-based application intelligent development method, characterized in that, include: Step 1: Establish a data management platform for managing and storing application software data; Application software data is categorized by type, and application software data belonging to each type is sorted by the number of users of the corresponding application software. Step 2: Analyze the data of various types of application software within the data management platform; Extract the front-end and back-end functional components of various types of application software. Based on the application software data, classify the functional components according to the main function and subordinate function. Distinguish functional components by main functional components and subordinate functional components. Subordinate functional components refer to components whose functions are extensions of the functions assigned to the main functional components. Based on the frequency index of different categories of functional components in various types of application software, determine the importance level of different categories of functional components. Step 3: Developers use a visual application software editor to implement the basic development of the application software; Step 4: Calculate the performance evaluation index of the initial application software model and identify the functional components that can cause the performance evaluation index to rise or fall. Based on the application type of the initial application software model, the simulation experiment objectives are determined. Based on the simulation experiment objectives, the functional components of the initial application software model participating in the simulation are determined. User behavior and system load are simulated based on relevant simulation tools and platforms. The performance evaluation indicators of the initial application software model are obtained by comprehensively considering the number of users, the average request frequency, the average response time, the throughput, and the error rate. According to the simulation experiment objectives, the relevant variables in the performance evaluation indicators are adjusted to obtain the changes in the performance evaluation indicators related to the simulation experiment objectives. Through formula Determine the performance evaluation metrics for the initial application software model, where T represents the average response time for user requests, H represents the number of users in the initial application software model, and O represents the average request frequency per user. Let N represent the response time of the k-th user request, N represent the total number of user requests, and R represent the simulation duration. The throughput within the range, w represents the error rate. Based on the simulation experiment target, the relevant variables in the performance evaluation index are adjusted to obtain the changes in the performance evaluation index related to the simulation experiment target. Step 5: Develop application software based on existing LLM model technology; Step Six: Developers preview and debug the application software to confirm whether to release it.

2. The application intelligent development method based on user intent according to claim 1, characterized in that, Based on the frequency of occurrence of different categories of functional components in various types of application software, the importance level of different categories of functional components is determined. The specific method is as follows: Analyze and compare data from multiple application software programs of the same type, and compare the frequency of occurrence of functional components in these programs. For n application software belonging to type 'a', ensuring that these n application software meet the standards, are representative, and are updated in real time, count the number of main functional components of all application software or the number of subordinate functional components of a certain main functional component, let x, and label the corresponding number of each functional component to obtain the set of functional components. Through formula Determine the frequency index of the functional component, where x represents the functional component number x, F(x) represents the frequency index of the functional component in n application software, n(x) represents that there are n(x) application software with functional component number x, and i represents the i-th application software. This represents the number of registered users for the i-th application. This refers to application software with functional components numbered x. The frequency index of different functional components of the same type is calculated using a formula to determine the importance level of different categories of functional components.

3. The application intelligent development method based on user intent according to claim 2, characterized in that, The method for determining the importance level of different categories of functional components is as follows: Calculate the occurrence frequency index of all functional components in multiple application software of the same type. Compare functional components of the same category with those of the same category, and determine that the thresholds for the functional component occurrence frequency index are A and B. Calculate the functional component occurrence frequency index for each functional component. When F(x) < A, it means that the functional component numbered x belongs to a special functional component. When it means that the functional component numbered x belongs to a regular functional component. When it means that the functional component numbered x belongs to a main functional component or a main sub-functional component. The thresholds A and B for the functional component occurrence frequency index are determined according to the set of functional component occurrence frequency indices. By determining the true values of A and B, the number N of functional components belonging to the main functional component or the main sub-functional component, the number M of regular functional components, and the number L of special components are obtained, ensuring and .

4. The application intelligent development method based on user intent according to claim 1, characterized in that, The specific method for identifying functional components that can cause performance evaluation metrics to increase or decrease is as follows: Add or remove different functional components of this type, one at a time. This functional component can be a secondary component or a primary component without secondary components. Simulations are then performed. Through simulation experiments, functional components that cause a decrease in performance evaluation indicators in the initial application software model are identified and designated as replaceable components. The performance evaluation indicators of the initial application software model and the performance evaluation indicators of the application software model after removing replaceable components are calculated. Replaceable components are sorted in descending order of performance evaluation indicator improvement, resulting in a replaceable component sequence. Through simulation experiments, functional components that are not present in the initial application software model but whose performance evaluation indicators increase after addition are identified and designated as optimized components. The performance evaluation indicators of the initial application software model and the performance evaluation indicators of the application software model after adding optimized components are calculated. Optimized components are sorted in descending order of performance evaluation indicator improvement, resulting in an optimized component sequence. Based on the replaceable component sequence and the optimized component sequence, random optimization schemes are generated by randomly removing replaceable components and adding optimized components. The performance evaluation indicators of the random optimization schemes are calculated. Considering the functional consistency between the random optimization schemes and the initial application software model, the best random optimization scheme is selected.

5. The application intelligent development method based on user intent according to claim 4, characterized in that, Calculate the performance evaluation index of the stochastic optimization scheme, and select the best stochastic optimization scheme by considering the functional consistency between the stochastic optimization scheme and the initial application software model. The specific method is as follows: Using formula Determine functional consistency, where J represents the number of main functional components in the initial application software model. Let represent the number of components in the stochastic optimization scheme that have the same or similar functions as the main functional components of the initial application software model, denoted as retained components, and z represent the z-th retained component. In the initial application software model, main functional components that are the same or similar to z are denoted as corresponding components. The z-th retained component corresponds to the z-th corresponding component, and j(z) represents the number of subordinate functional components shared by the z-th corresponding component. This represents the number of functional components that have the same or similar functions as the z-th retained component and its corresponding component. Using formula Calculate the performance evaluation index of the random optimization scheme, and select the best random optimization scheme by combining the functional consistency between the random optimization scheme and the initial application software model. Here, r represents the r-th random optimization scheme, Y(0) represents the performance evaluation index of the initial application software model, and Y(r) represents the performance evaluation index of the r-th random optimization scheme. Select the random optimization scheme with the largest YX(r) value as the best random optimization scheme.

Images