AI-based software demand intelligent analysis and prediction method and system

Abstract

The invention relates to the technical field of AI demand intelligent analysis, and discloses an AI-based software demand intelligent analysis and prediction method and system, and the method comprises the steps: collecting the use data of software when a software product is used; demand feedback information is generated by acquiring user feedback; carrying out demand analysis on the use data and the demand feedback information by utilizing a transfer learning strategy; and gaming a value analysis result and resource occupation to generate a final software demand prediction result. While the demand satisfaction degree is ensured, the resource consumption is minimized, and the development efficiency is improved. By constructing a function value evaluation system, the method can balance the function income and implementation cost, provides scientific and automatic decision basis for enterprises, greatly reduces the software iteration period, and improves the user experience.

Description

Technical Field

[0001] The present invention relates to the technical field of AI demand intelligent analysis, and specifically to an AI-based software demand intelligent analysis and prediction method and system. Background Art

[0002] In the modern software development process, accurately identifying and meeting user needs has become the key to improving product competitiveness. However, traditional software requirements analysis methods usually rely on user interviews, questionnaires, market analysis, etc. These methods are not only costly and time-consuming, but also have problems such as strong subjectivity and untimely data updates. In addition, the diversity and uncertainty of user needs make it difficult for the development team to accurately determine which functions have higher value and which functions may be ignored or waste development resources.

[0003] In recent years, with the development of big data, artificial intelligence (AI) and machine learning (ML) technologies, software requirements analysis has gradually evolved towards data-driven and intelligent directions. Through user behavior data analysis, we can more deeply explore users' operating habits, functional preferences and potential unmet needs during software use. For example, based on user operation logs, clickstream data, search query records and other information, we can build an intelligent requirements analysis system to more accurately understand the changing trends of user requirements.

[0004] In addition, cross-domain data migration has also attracted widespread attention in software requirements analysis. Different software products may have similarities in interface structure, interaction logic, and functional modules. Therefore, through data migration and learning, the new system's dependence on large-scale user feedback data can be reduced, and the efficiency of demand forecasting can be improved. At the same time, how to balance functional value and resource consumption in the process of demand analysis, optimize resource allocation, and enable software development to take into account user experience, development costs, and system performance has become an important research direction in the industry. Summary of the invention

[0005] In view of the above-mentioned problems, the present invention is proposed.

[0006] Therefore, the technical problem solved by the present invention is that the existing intelligent analysis method for software requirements has problems such as the analysis results cannot be adapted to the deployment of software and how to select high-quality requirements.

[0007] In order to solve the above technical problems, the present invention provides the following technical solutions: an AI-based software demand intelligent analysis and prediction method, comprising: When the software product is in use, the software usage data is collected; and demand feedback information is generated by obtaining user feedback; Using a transfer learning strategy, performing demand analysis on the usage data and demand feedback information; The results of value analysis are compared with resource usage to generate the final software demand forecast results.

[0008] As a preferred solution of the AI-based software demand intelligent analysis and prediction method described in the present invention, the usage data includes the user's operation records in the software, the query records of the search box, and the records of the user's simultaneous use of multiple software functions of the software; Each record includes the time when the record occurred.

[0009] As a preferred solution of the AI-based software demand intelligent analysis and prediction method described in the present invention, the demand feedback information includes text information on software requirements provided by users through feedback channels.

[0010] As a preferred solution of the AI-based software demand intelligent analysis and prediction method described in the present invention, the transfer learning strategy includes obtaining different software products, assuming that the i-th software product is ; Take the main menu as the starting node, each interactive submenu as a node, and each software function as the end point. The connecting lines represent the interactive relationship; construct this software respectively and The path for each function to be implemented, as a function map; During the user's interactive behavior, when the user exits the operation at a node other than the end point and the starting point, it is recorded as a hidden event; the frequency of hidden events is counted: ;in, represents the frequency of hidden events occurring at node b; Indicates the number of times the hidden event occurs at node b; represents the total number of times node b is reached through interactions; Through pre-trained deep learning algorithms, this software Forecast the demand; input: and Functional map, frequency of each latent event; output: There are and The endpoints that are not in , and the probability that each endpoint represents a software requirement are output, and the endpoint set Z is obtained.

[0011] As a preferred solution of the AI-based software demand intelligent analysis and prediction method of the present invention, wherein: according to the demand feedback information, The software functions in the deep learning algorithm are matched twice, and according to the result of the second matching, each endpoint output by the deep learning algorithm is analyzed for value; the second matching includes extracting keywords from the demand feedback information and calculating the distance with each endpoint in the endpoint set Z; for each endpoint, the similarity is multiplied by the number of each keyword, and then the sum is processed to obtain the value of each endpoint in the endpoint set Z, where each endpoint represents a demand selection; Assume that the user uses m The function is expressed as ; For m software combinations randomly, calculate the value of each functional combination: ;in, represents the value of the yth combination; It represents the probability that the functions represented by the elements in the yth combination are used at the same time when the function represented by the nth element in the combination is used; n represents the number of elements in the yth combination; represents the expansion coefficient, ; Each feature combination represents a feature selection.

[0012] As a preferred solution of the AI-based software demand intelligent analysis and prediction method described in the present invention, the resource occupation includes: for each function selection, in this software It is implemented on the basis of existing functions and requires resources for calculation; The resource usage of each feature selection is calculated using a pre-trained machine learning algorithm; input: feature selection and Existing functions; Output: resources required for function selection.

[0013] As a preferred solution of the AI-based software requirement intelligent analysis and prediction method of the present invention, the game between the result of value analysis and the occupation of resources includes obtaining all function options, which are set to G; randomly extracting the G function options, the number of each extraction is at most G, and the combination of function options extracted each time is different; Get Q function selection combinations, calculate the total value and total occupied resources of each function selection combination; remove the function selection combinations whose total occupied resources exceed λ, and get the alternative function selection combinations; where λ represents the preset total occupied resource threshold; The alternative function selection combinations are used as prediction results, and high-quality combinations are screened and marked to generate analysis reports; The high-quality combination includes, for each prediction result, playing a game with the goal of maximizing the total value and minimizing the total occupied resources; establishing a two-dimensional coordinate system with the horizontal axis representing the total occupied resources and the vertical axis representing the total value; Multiply the total value of each prediction result by After that, it is used as the vertical axis; the total occupied resources of each prediction result is used as the horizontal axis; each prediction result is deployed in a two-dimensional coordinate system; in, =The maximum value of the total value in the prediction results / , so that the maximum values ​​of the horizontal and vertical coordinates are , if the prediction result falls into , , If the three points form a triangle, it is considered a high-quality combination; The analysis report records each prediction result, the value analysis result of the prediction result and the occupied resources.

[0014] An AI-based software demand intelligent analysis and prediction system using any method described in the present invention, wherein: The collection unit collects the usage data of the software when the software product is used; and generates demand feedback information by obtaining user feedback; An analysis unit, using a transfer learning strategy to perform demand analysis on the usage data and demand feedback information; The game unit plays a game between the result of value analysis and the occupation of resources to generate the final software demand forecast result.

[0015] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein: the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0016] A computer-readable storage medium stores a computer program, wherein: when the computer program is executed by a processor, the steps of any one of the methods of the present invention are implemented.

[0017] Beneficial effects of the present invention: The AI-based intelligent analysis and prediction method for software requirements provided by the present invention effectively improves the accuracy and optimization capabilities of software requirement prediction through intelligent demand analysis, transfer learning, resource optimization and game theory decision-making. Compared with traditional methods, the present invention can automatically extract user behavior data, combine deep learning models to mine potential needs, and use transfer learning strategies to predict possible functional missing points across software. In addition, the present invention combines game theory optimization algorithms to minimize resource consumption and improve development efficiency while ensuring demand satisfaction. By constructing a functional value evaluation system, the present invention can balance functional benefits and implementation costs, provide enterprises with scientific and automated decision-making basis, significantly reduce software iteration cycles, and improve user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0019] Figure 1 An overall flow chart of an AI-based software requirements intelligent analysis and prediction method provided for the first embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0021] Example 1, reference Figure 1 , as an embodiment of the present invention, provides a software requirement intelligent analysis and prediction method based on AI, comprising: S1: When the software product is in use, the software usage data is collected; and demand feedback information is generated by obtaining user feedback.

[0022] The usage data includes the user's operation records in the software, the query records of the search box, and the records of the user's simultaneous use of multiple software functions of the software; wherein each record includes the time when the record occurred. The operation record includes the sequence of human-computer interactions generated during a single continuous operation. The query record includes the text content to be queried entered in the search box. When the user uses multiple software functions of the software at the same time, not only the time is collected, but also the software functions used at the same time. The demand feedback information includes text information on software requirements provided by users through feedback channels.

[0023] In other words, there are actually two types of feedback. One is the implicit feedback of "collecting software usage data", through which the implicit expression of demand can be predicted. The second is the explicit feedback directly provided by the user. Through the joint action of explicit feedback and implicit feedback, the subsequent analysis and prediction process is realized. Explicit feedback can directly guide the direction of software optimization, while implicit feedback can supplement the needs that users have not clearly expressed. The combination of the two can build a more comprehensive user demand model and improve the accuracy of subsequent demand prediction.

[0024] S2: Utilize the transfer learning strategy to perform demand analysis on the usage data and demand feedback information.

[0025] The transfer learning strategy includes obtaining different software products, assuming that: the i-th software product is ; Take the main menu as the starting node, each interactive submenu as a node, and each software function as the end point. The connecting lines represent the interactive relationship; construct this software respectively and The path for each function to be implemented is used as a function map. By constructing the path, there is no need to vectorize text data or perform complex relationship analysis. The analysis can be completed through the simplest comparison relationship, which can greatly reduce the computing resource usage.

[0026] During the user's interactive behavior, when the user exits the operation at a node other than the end point and the starting point, it is recorded as a hidden event; the frequency of hidden events is counted: ;in, represents the frequency of hidden events occurring at node b; Indicates the number of times the hidden event occurs at node b; Indicates the total number of times node b is reached through interaction. This software is trained using a deep learning algorithm with a pre-training ratio of 8:2. Forecast the demand; input: and Functional map, frequency of occurrence of each hidden event (the input of this element is mainly to increase the attention of hidden events, because when selecting functions, users will mostly search in similar sub-menus, and the occurrence of hidden events can generally be understood as "not finding a certain function" and therefore may be needed); output: There are and The endpoints that are not in the set are output, and the probability that each endpoint represents a software requirement is output (that is, this deep learning algorithm judges each endpoint and determines its probability of being a "requirement") to obtain the endpoint set Z.

[0027] It should be said that during the operation of the software, users usually enter the function page from the main menu, but if they frequently exit in the lower menu, it may mean that the user has not found the expected function, or the current menu organization does not meet the user's expectations. By counting the frequency of hidden events (that is, the number of times the user exits at a certain menu point), it is possible to quantify whether there is a high demand but unmet function missing problem in a certain interface. By constructing a functional map and analyzing the hierarchical relationship between different functions, it is ensured that the demand analysis not only considers a single interface operation, but also considers the correlation between different functions. For example, if the user exits the "Privacy Options" in the "Settings" menu many times, it is likely that the menu lacks specific functions (such as "Advanced Privacy Settings"). Since users will look for functions in similar lower menus in most cases, deep learning algorithms should focus on those high-frequency hidden events and increase the weight of these data in demand forecasting. The model not only analyzes the usage of existing functions, but also infers the functions that are not available in existing software but may be needed, and calculates the probability of each function being a demand, thereby generating the optimal demand forecast.

[0028] In this embodiment, the core task of the deep learning algorithm is to predict the functions that may be missing in the software based on the function map and the user's hidden events, and calculate the probability of their demand. In order to achieve this goal, the deep learning algorithm is implemented by the Graph Neural Network (GNN). The function map is a natural graph structure. There are hierarchical relationships and interaction paths between the menus and functions of the software, which can be modeled by a graph, where each menu item and function is a node, and the click / jump behavior is an edge. GNN is very suitable for processing this kind of graph structure data, and can learn the association between different functions, helping to infer functions that are not provided but may be in demand. Since the functional layout of different software may be similar, GNN can use the functional maps of different software as training data to improve prediction capabilities through transfer learning.

[0029] In other optional embodiments, the Transformer + self-attention mechanism can also be used. The Transformer model can process sequence data and calculate the user's interaction relationship between different interfaces and functions through the self-attention mechanism. Or LSTM (Long Short-Term Memory Network); LSTM can remember long-term historical operations and is suitable for modeling user interaction behavior sequences, especially for scenarios with time-sequential characteristics.

[0030] According to the feedback information of the requirements, A secondary match is performed on the software functions in the deep learning algorithm, and based on the result of the secondary match, a value analysis is performed on each endpoint output by the deep learning algorithm; the secondary matching includes extracting keywords from the demand feedback information and calculating the distance with each endpoint in the endpoint set Z; for each endpoint, the similarity is multiplied by the number of each keyword, and then the sum is performed to obtain the value of each endpoint in the endpoint set Z, where each endpoint represents a demand selection.

[0031]

[0032] Where T represents the number of endpoints in the endpoint set Z; t represents the index, represents the value of the tth endpoint, Indicates the similarity between the tth endpoint and the keyword v; represents the number of keywords v; V represents the total number of keywords.

[0033] It should be noted that user feedback is generally not used directly as a basis for demand generation, but only for reference. Because some user feedback is too far-fetched to be realized. However, the function represented by the "end point" after transfer learning is achievable. Therefore, user feedback is only used as a reference factor in this solution.

[0034] It should be noted that in demand analysis, the text information provided by users is often unstructured. In order to accurately match user needs, secondary matching is performed to ensure that the final demand prediction results are highly consistent with the user's actual needs and reduce irrelevant or low-value functional interference.

[0035] Assume that the user uses m The function is expressed as ; For m software combinations randomly, calculate the value of each functional combination: ;in, represents the value of the yth combination; It represents the probability that the functions represented by the elements in the yth combination are used at the same time when the function represented by the nth element in the combination is used; n represents the number of elements in the yth combination; represents the expansion coefficient, (Make the value of the combination much greater than "the value of each endpoint in endpoint set Z". Because this part of the combination of existing functions takes up almost no resources, or takes up very little resources, it has a high cost-effectiveness); each function combination represents a function selection.

[0036] S3: Compare the results of value analysis with resource usage to generate the final software demand forecast results.

[0037] Furthermore, the resource usage includes, for each function selection, Based on the existing functions, it needs to occupy resources for calculation. Through the pre-trained machine learning algorithm, the occupied resources of each function selection are calculated; input: function selection and Existing functions; Output: resources required for function selection.

[0038] Through pre-trained machine learning models, combined with the existing functions of the software, the storage, computing, bandwidth, development time and other resource usage of each function selection is predicted to ensure that the optimization plan maximizes the value while achieving optimal resource utilization, thereby balancing user needs and development feasibility, and improving software iteration efficiency and cost-effectiveness.

[0039] This embodiment uses a gradient boosting decision tree (GBDT) for resource occupancy prediction for the following reasons: The input features are nonlinear and complex, and GBDT can handle nonlinear mappings: Resource occupancy is affected by function call chains, UI complexity, data processing requirements, etc. GBDT is suitable for such complex relationship modeling. High training efficiency and low computing resource consumption: Compared with deep learning, GBDT is faster to train, occupies fewer resources, and is suitable for prediction tasks of medium-sized data sets. Strong interpretability and support for feature importance analysis: It can analyze which factors have the greatest impact on resource occupancy, helping developers optimize resource allocation strategies.

[0040] Furthermore, all function options are obtained, set to G; the G function options are randomly selected, the maximum number of each selection is G, and the function selection combination selected each time is different. Q function selection combinations are obtained, and the total value and total occupied resources of each function selection combination are calculated; after eliminating the function selection combination with a total occupied resource exceeding λ, the alternative function selection combination is obtained; where λ represents the preset total occupied resource threshold.

[0041] The alternative function selection combinations are used as prediction results, and the high-quality combinations are screened and marked to generate analysis reports.

[0042] The high-quality combination includes, for each prediction result, playing a game with the goal of maximizing the total value and minimizing the total occupied resources; establishing a two-dimensional coordinate system with the horizontal axis representing the total occupied resources and the vertical axis representing the total value. After that, it is used as the vertical axis; the total occupied resources of each prediction result is used as the horizontal axis; each prediction result is deployed in a two-dimensional coordinate system.

[0043] in, =The maximum value of the total value in the prediction results / , so that the maximum values ​​of the horizontal and vertical coordinates are , if the prediction result falls into , , The triangle formed by the three points is judged to be a high-quality combination.

[0044] In the analysis report, each prediction result, as well as the value analysis results and occupied resources of the prediction result are recorded. It should be noted that the goal of the above scheme is to optimize the selection of function combinations among multiple function selection schemes to ensure that the results of software demand prediction are both of high value and meet resource constraints. Specific purposes include: balancing function value and resource occupation to improve software optimization efficiency: by calculating the total value and resource occupation of the function combination, high-yield and low-cost function sets are screened to avoid waste of development resources. Setting the resource threshold λ, eliminating function combinations that occupy too many resources, and ensuring that the optimization scheme meets the actual development feasibility. Adopting game theory optimization to find the optimal function combination: adopting the dual-objective optimization strategy of maximizing total value & minimizing resources to ensure that the final solution not only meets user needs but also does not cause system burden. Through the two-dimensional coordinate game strategy, screening is performed between different function combinations, so that the final recommended solution is optimal in value-resource ratio. Using geometric methods to screen high-quality combinations and improve decision-making efficiency: in the two-dimensional coordinate system, the horizontal axis represents resource occupation, the vertical axis represents value, and a high-quality combination area (a triangle composed of three key points) is constructed. If the functional combination falls into this area, it is considered that the combination is optimal in terms of resource and value trade-off, ensuring that the selected solution is the most cost-effective for software optimization. Automatically generate analysis reports to assist software functional decision-making: visualize the optimization results to help developers more intuitively understand the pros and cons of different solutions and reduce subjective errors in manual judgment. Automatically mark "high-quality combinations" to improve the scientificity, intelligence and development efficiency of software product optimization.

[0045] On the other hand, this embodiment also provides an AI-based software demand intelligent analysis and prediction system, which includes: The collection unit collects the usage data of the software when the software product is used, and generates demand feedback information by obtaining user feedback.

[0046] The analysis unit uses a transfer learning strategy to perform demand analysis on the usage data and demand feedback information.

[0047] The game unit plays a game between the result of value analysis and the occupation of resources to generate the final software demand forecast result.

[0048] If the above 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, 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. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc., which can store program code.

[0049] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0050] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0051] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0052] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A software requirement intelligent analysis and prediction method based on AI, characterized in that: include: When the software product is in use, collect the usage data of the software; and generate demand feedback information by obtaining user feedback; Using the transfer learning strategy, the usage data is analyzed for demand; based on the demand feedback information, the software functions obtained from the demand analysis are matched twice, and based on the results of the secondary matching, each end point output by the deep learning algorithm is analyzed for value; at the same time, based on the usage data, the combined value analysis of the existing functions of the software is directly performed; Compare the results of value analysis with resource usage to generate the final software demand forecast results: select the alternative function combinations as the forecast results, select high-quality combinations and mark them, and generate an analysis report; The analysis report records each prediction result, the value analysis result of the prediction result and the occupied resources.

2. The AI-based software requirements intelligent analysis and prediction method according to claim 1, characterized in that: The usage data includes the user's operation records in the software, the query records of the search box, and the records of the user's simultaneous use of multiple software functions of the software; Each record includes the time when the record occurred.

3. The AI-based software requirements intelligent analysis and prediction method according to claim 2, characterized in that: The demand feedback information includes text information on software requirements provided by users through feedback channels.

4. The AI-based software requirements intelligent analysis and prediction method according to claim 3, characterized in that: The transfer learning strategy includes obtaining different software products, setting :The i-th software product is ; The main menu is the starting node, each interactive sub-menu is a node, each software function is the end point, and the connecting line represents the interactive relationship; construct this software respectively and The path for each function to be implemented, as a function map; When a user exits an interactive behavior at a node other than the end point or the start point, it is recorded as an implicit event. Count the frequency of hidden events: ;in, represents the frequency of hidden events occurring at node b; Indicates the number of times the hidden event occurs at node b; represents the total number of times node b is reached through interactions; Through pre-trained deep learning algorithms, this software To forecast the demand; input: and The functional map of the event, the frequency of each latent event; Output: There are and The endpoints that are not in , and the probability that each endpoint represents a software requirement are output, and the endpoint set Z is obtained.

5. The AI-based software requirements intelligent analysis and prediction method according to claim 4, characterized in that: The secondary matching includes, after extracting keywords from the demand feedback information, calculating the distance with each endpoint in the endpoint set Z; for each endpoint, multiplying the similarity by the number of keywords, and then summing the similarities to obtain the value of each endpoint in the endpoint set Z, where each endpoint represents a demand selection; The combination value analysis includes: assuming that the user uses m The function is expressed as ; For m software combinations randomly, calculate the value of each functional combination: ;in, represents the value of the yth combination; It represents the probability that the functions represented by the elements in the yth combination are used at the same time when the function represented by the nth element in the combination is used; n represents the number of elements in the yth combination; represents the expansion coefficient, ; Each feature combination represents a feature selection.

6. The AI-based software requirements intelligent analysis and prediction method according to claim 5, characterized in that: The resource usage includes, for each function selection, It is implemented on the basis of existing functions and requires resources for calculation; Use pre-trained machine learning algorithms to calculate the resources occupied by each function selection; Input: Feature selection and Existing features of Output: The resources required for function selection.

7. The AI-based software requirements intelligent analysis and prediction method according to claim 6, characterized in that: The game between the result of value analysis and the occupation of resources includes obtaining all function options, which are set to G; randomly extracting the G function options, with the maximum number of each extraction being G, and the combination of function options extracted each time being different; Get Q function selection combinations, and calculate the total value and total occupied resources of each function selection combination; After eliminating the function selection combinations whose total occupied resources exceed λ, an alternative function selection combination is obtained; wherein λ represents a preset total occupied resource threshold; The high-quality combination includes, for each prediction result, playing a game with the goal of maximizing the total value and minimizing the total occupied resources; establishing a two-dimensional coordinate system with the horizontal axis representing the total occupied resources and the vertical axis representing the total value; Multiply the total value of each prediction result by After that, it is used as the vertical axis; the total occupied resources of each prediction result is used as the horizontal axis; each prediction result is deployed in a two-dimensional coordinate system; in, =The maximum value of the total value in the prediction results / , so that the maximum values ​​of the horizontal and vertical coordinates are , if the prediction result falls into , , The triangle formed by the three points is judged to be a high-quality combination.

8. An AI-based software requirements intelligent analysis and prediction system using the method according to any one of claims 1 to 7, characterized in that: The collection unit collects the usage data of the software when the software product is used; and generates demand feedback information by obtaining user feedback; An analysis unit, using a transfer learning strategy to perform demand analysis on the usage data and demand feedback information; The game unit plays a game between the result of value analysis and the occupation of resources to generate the final software demand forecast result.

9. A computer device comprising: A memory and a processor; the memory stores a computer program, wherein the processor implements the steps of any method as claimed in claim 1 when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images