A software development efficiency management system based on deep reinforcement learning

Abstract

The present invention discloses a software development efficiency management system based on deep reinforcement learning, which relates to the field of software engineering technology, and comprises a data acquisition module, which uses a project management tool to collect software development data and performs preprocessing; a model building module, which uses a deep neural network to build a demand prediction model based on the preprocessed software development data to predict future demand changes; an optimization suggestion module, which generates optimization suggestions according to future demand changes; a progress monitoring module, which monitors project progress in real time through natural language processing (NLP) according to the optimization suggestions, identifies and extracts demand change information, and classifies demand changes; an impact assessment module, which assesses the impact of demand changes according to the classified demand changes combined with the progress and resource status of the current project; and a change plan module, which generates the best change plan according to the impact of the demand changes.

Description

Technical Field

[0001] The present invention relates to the technical field of software engineering, and in particular to a software development efficiency management system based on deep reinforcement learning. Background Art

[0002] With the continuous expansion of software development projects and the increasing complexity, traditional project management methods have been unable to cope with the rapid changes in demand and dynamic adjustment of resources in modern software development. In the past few decades, project management tools such as Jira and Trello have been widely used to help development teams manage task allocation, track project progress and record user needs. However, these tools rely more on manual intervention and static analysis, lack the ability to predict future demand changes and optimize resource allocation in real time. At the same time, with the rapid development of deep learning and natural language processing technologies, more and more research has begun to explore how to use these technologies to improve the level of automated management of software development. For example, deep reinforcement learning has made significant progress in solving complex decision-making problems in recent years. DRL can learn the optimal strategy to cope with complex dynamic environments by building reward mechanisms and state spaces, which provides new ideas for software development efficiency management systems.

[0003] Although the existing software development management system can help the team improve development efficiency to a certain extent, it still has many shortcomings. Most traditional project management tools are based on simple statistical analysis of historical data and cannot dynamically predict future changes in demand, resulting in delayed or excessive waste of development resource allocation. When responding to changes in demand, the existing system relies more on the experience and subjective judgment of developers and lacks systematic decision-making support. This not only easily leads to project delays, but may also increase costs due to poor resource utilization. When dealing with changes in demand, the existing system lacks comprehensive consideration of multiple dimensions such as project progress, resource utilization, and demand satisfaction. This makes the optimization plan often not accurate enough when facing complex changes in demand and cannot respond to dynamic changes in the project in a timely manner. Therefore, how to effectively use deep reinforcement learning and natural language processing technology to build an intelligent software development management system that can predict demand changes in real time and dynamically optimize resource allocation has become a key issue in improving development efficiency. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a software development efficiency management system based on deep reinforcement learning to solve the problem that the existing system lacks comprehensive consideration of multiple dimensions such as project progress, resource utilization and demand satisfaction when dealing with demand changes. This makes the optimization plan often not accurate enough when facing complex demand changes and cannot respond to the dynamic changes of the project in a timely manner.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] The present invention provides a software development efficiency management system based on deep reinforcement learning, which comprises:

[0008] Data collection module, model building module, optimization suggestion module, progress monitoring module, impact assessment module, change plan module,

[0009] The data acquisition module uses a project management tool to collect software development data and perform preprocessing;

[0010] The model building module uses a deep neural network to build a demand forecasting model based on the preprocessed software development data to predict future demand changes;

[0011] The optimization suggestion module generates optimization suggestions according to future demand changes;

[0012] The progress monitoring module monitors the project progress in real time through natural language processing (NLP) according to the optimization suggestions, identifies and extracts demand change information, and classifies demand changes;

[0013] The impact assessment module assesses the impact of the demand changes based on the classified demand changes combined with the progress and resource status of the current project;

[0014] The change plan module generates the best change plan according to the impact of the demand change.

[0015] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, the specific steps of using a project management tool to collect software development data and preprocessing it are as follows:

[0016] Use the project management tool Jira to collect task assignment data, completion status data, time recording data, user story data, and defect report data;

[0017] Clean the acquired data, remove irrelevant fields, use the average value to handle missing values, and perform format conversion;

[0018] The processed data are merged into a complete software development dataset, and the software development dataset is set as ,in Represents task allocation data, Indicates completion status data, Indicates time recording data, Represents user story data, Represents defect report data.

[0019] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, wherein: based on the preprocessed software development data set, a demand prediction model is constructed using a deep neural network, and the specific steps are:

[0020] Define the state space and action space:

[0021] ;

[0022] in, Indicates the time of the project The state space at time, Represents a time index;

[0023] ;

[0024] in, Indicates the time of the project The action space at the moment, Indicates demand changes and development resource adjustment operations. Index variables representing demand changes and development resource adjustment operations;

[0025] Defining the reward function To optimize demand forecasting, consider the following factors:

[0026] ;

[0027] in, Indicates changes in project progress, Indicates time The project progress at all times, Indicates time Project progress at all times;

[0028] ;

[0029] in, represents an improvement in resource utilization, Indicates time Resource utilization at any time, Indicates time Resource utilization at each moment;

[0030] ;

[0031] in, represents the change in demand satisfaction. Indicates time The satisfaction of needs at all times, Indicates time The satisfaction of needs at all times;

[0032] Reward Function The expression is:

[0033] ;

[0034] in, express The weight parameter, express The weight parameter, express The weight parameter of

[0035] Calculate the action-value function based on the reward function.

[0036] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, wherein: the action value function is calculated according to the reward function, and the expression is:

[0037] ;

[0038] in, represents the action-value function, represents the expected value, represents the discount factor, Indicates the maximum value, Indicates time The state space at time, Indicates time Action space at any moment.

[0039] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, the specific steps of predicting future demand changes are as follows:

[0040] The demand forecasting model expression is:

[0041] ;

[0042] in, Indicates time point The predicted value of demand change, express The index variable, Indicates the software development dataset Class data at time point The value of represents the regression coefficient, express The number of

[0043] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, wherein: generating optimization suggestions according to future demand changes, the specific steps are:

[0044] when When the demand increases, it indicates that there will be an increase in future demand, and it is necessary to increase development resources or improve production efficiency to meet the demand;

[0045] when When the demand decreases, it indicates that the development resources will be appropriately reduced;

[0046] when , indicating that demand remains stable and the current resource allocation is maintained.

[0047] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, wherein: the project progress is monitored in real time through natural language processing (NLP) according to the optimization suggestions, and the requirement change information is identified and extracted. The specific steps are:

[0048] Through natural language processing (NLP), task description data, defect report data, and requirement change data of the software development team are collected in real time from the project management tool Jira. After removing stop words, part-of-speech tagging, named entity recognition, and word embedding processing, text vectors are obtained. ;

[0049] A text classification model based on convolutional neural network (CNN) is used to classify the demand change information. Multiple convolution kernels are used to extract the local features of the demand change information text. The convolution operation expression is:

[0050] ;

[0051] in, Represents text vector The index variable, Indicates The output features of the convolutional layers are Indicates The convolution kernel weights of the convolutional layers, Indicates input text vector, Indicates The bias term of the convolutional layer, Represents the activation function.

[0052] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, the specific steps of classifying the demand changes are as follows:

[0053] Perform maximum pooling on the convolution output to reduce the feature dimension. The expression is:

[0054] ;

[0055] in, Indicates Features after pooling of convolutional layers;

[0056] The pooled features are passed through the fully connected layer to output the classification results, and the expression is:

[0057] ;

[0058] in, Represents the classification result, Softmax represents the activation function, represents the weight of the fully connected layer, represents the bias term of the fully connected layer, Indicates the dimension index of the output.

[0059] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, wherein: the impact of the demand change is evaluated based on the classified demand change combined with the progress and resource status of the current project, and the specific steps are:

[0060] The impact of demand changes on project progress is:

[0061] ;

[0062] in, An index variable representing a requirement change, Indicates a change in requirements The impact of changes on the project schedule, Indicates a change in requirements The complexity of Indicates a change in requirements The type of corresponding classification results, Indicates the time remaining in the project;

[0063] The impact of demand changes on resource utilization improvements is:

[0064] ;

[0065] in, Indicates a change in requirements The impact of improvements in resource utilization, Indicates the total number of resource types. Represents the index of the resource. Indicates Resources in time availability, represents the smoothing term;

[0066] The impact of demand changes on demand satisfaction is:

[0067] ;

[0068] in, Indicates a change in requirements The impact of changes in demand satisfaction, Indicates a change in requirements The progress goals to be achieved, Indicates a change in requirements expected completion time;

[0069] According to the classification results and the current project status, the impact of the demand change on the progress, resources and demand satisfaction is evaluated. The expression is:

[0070] ;

[0071] in, Represents the comprehensive impact index of demand changes.

[0072] As a preferred solution of the software development efficiency management system based on deep reinforcement learning described in the present invention, wherein: the optimal change plan is generated according to the impact of the demand change, and the specific steps are:

[0073] Introducing optimization functions , combined with the impact index of all demand changes and the current status of the project to determine the optimal solution, the expression is:

[0074] ;

[0075] in, Indicates the priority coefficient of each requirement change, represents the penalty factor, Indicates the total number of requirement changes;

[0076] Generate the best change plan by solving the optimization function;

[0077] According to the classification results and comprehensive impact index , priority processing The biggest change in requirements;

[0078] when When insufficient resource utilization is displayed, corresponding resources are added to meet the requirements of demand changes;

[0079] when When it is shown that the project progress is affected, adjust the project plan, extend the construction period and increase the input of personnel;

[0080] when When the demand satisfaction level is decreasing, tasks with high demand satisfaction are prioritized.

[0081] The beneficial effects of the present invention are: a complete intelligent project management system based on deep reinforcement learning is formed, the efficient collection and preprocessing of data provide strong data support for subsequent prediction models, and the application of deep neural networks enables the system to accurately predict future changes in demand and then generate optimization suggestions. Through NLP and CNN technology, the system can automatically identify and classify demand changes, and generate the best change plan to ensure the rational allocation of resources and optimization of development progress. Each step is closely related to form a closed-loop system of dynamic adjustment, real-time optimization, and intelligent decision-making, which effectively improves the overall efficiency and management level of software development. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] 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.

[0083] Figure 1 This is a schematic diagram of a software development efficiency management system based on deep reinforcement learning in Example 1.

[0084] Figure 2 This is a schematic diagram of the demand forecasting model in Example 1. DETAILED DESCRIPTION

[0085] In order to make the above-mentioned objects, 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 accompanying drawings.

[0086] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0087] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0088] Example 1, reference Figure 1 and Figure 2, which is the first embodiment of the present invention, and provides a software development efficiency management system based on deep reinforcement learning, comprising the following steps:

[0089] Data collection module, model building module, optimization suggestion module, progress monitoring module, impact assessment module, change plan module,

[0090] Data collection module, which uses project management tools to collect software development data and perform preprocessing;

[0091] The model building module uses a deep neural network to build a demand forecasting model based on preprocessed software development data to predict future demand changes;

[0092] The optimization suggestion module generates optimization suggestions based on future demand changes;

[0093] The progress monitoring module monitors project progress in real time through natural language processing (NLP) based on optimization suggestions, identifies and extracts requirement change information, and classifies requirement changes;

[0094] Impact assessment module, which assesses the impact of demand changes based on the classified demand changes and the current project progress and resource status;

[0095] The change plan module generates the best change plan based on the impact of the demand change.

[0096] Use project management tools to collect and pre-process software development data;

[0097] Use the project management tool Jira to collect task assignment data, completion status data, time recording data, user story data, and defect report data;

[0098] Use outlier detection to clean the acquired data, remove irrelevant fields, use the average value to handle missing values, and perform format conversion;

[0099] It ensures that the input data is complete, accurate and consistent, providing a reliable basic data set for subsequent deep learning models. By eliminating noise and inconsistency in the data, the system can avoid incorrect predictions or decisions due to input data quality issues.

[0100] The processed data are merged into a complete software development dataset, and the software development dataset is set as ,in Represents task allocation data, Indicates completion status data, Indicates time recording data, Represents user story data, Represents defect report data.

[0101] Based on the preprocessed software development data, a demand forecasting model is built using a deep neural network to predict future demand changes and generate optimization suggestions based on future demand changes;

[0102] Through the powerful learning ability of deep neural networks, the system can accurately predict future demand changes, thereby providing forward-looking optimization suggestions to the development team;

[0103] Define the state space and action space:

[0104] ;

[0105] in, Indicates the time of the project The state space at time, Represents a time index;

[0106] ;

[0107] in, Indicates the time of the project The action space at the moment, Indicates demand changes and development resource adjustment operations. Index variables representing demand changes and development resource adjustment operations;

[0108] Defining the reward function To optimize demand forecasting, consider the following factors:

[0109] ;

[0110] in, Indicates changes in project progress, Indicates time The project progress at all times, Indicates time Project progress at all times;

[0111] ;

[0112] in, represents an improvement in resource utilization, Indicates time Resource utilization at any time, Indicates time Resource utilization at each moment;

[0113] ;

[0114] in, represents the change in demand satisfaction. Indicates time The satisfaction of needs at all times, Indicates time The satisfaction of needs at all times;

[0115] Reward Function The expression is:

[0116] ;

[0117] in, express The weight parameter, express The weight parameter, express The weight parameter of

[0118] The action value function is calculated based on the reward function, and the expression is:

[0119] ;

[0120] in, represents the action-value function, represents the expected value, represents the discount factor, Indicates the maximum value, Indicates time The state space at time, Indicates time The action space at the moment;

[0121] The demand forecasting model expression is:

[0122] ;

[0123] in, Indicates time point The predicted value of demand change, express The index variable, Indicates the software development dataset Class data at time point The value of represents the regression coefficient, express the number of

[0124] when When the demand increases, it indicates that there will be an increase in future demand, and it is necessary to increase development resources or improve production efficiency to meet the demand;

[0125] when When the demand decreases, it indicates that the development resources will be appropriately reduced;

[0126] when When , it indicates that the demand remains stable and the current resource allocation is maintained;

[0127] It not only improves the flexibility and adaptability of project management, but also significantly improves the efficiency of resource utilization, minimizes uncertainty in the project development process, and ensures that the project can be delivered on time and with quality.

[0128] Monitor project progress in real time through natural language processing (NLP) based on optimization suggestions, identify and extract requirement change information, and classify requirement changes;

[0129] Through natural language processing (NLP), task description data, defect report data, and requirement change data of the software development team are collected in real time from the project management tool Jira. Stop words are removed by using a custom stop word list, part-of-speech tagging is performed by a POS tagger, named entity recognition is performed by a NER model, and word embedding is performed using Word2Vec. ;

[0130] A text classification model based on convolutional neural network (CNN) is used to classify the demand change information. Multiple convolution kernels are used to extract the local features of the demand change information text. The convolution operation expression is:

[0131] ;

[0132] in, Represents text vector The index variable, Indicates The output features of the convolutional layers are Indicates The convolution kernel weights of the convolutional layers, Indicates input text vector, Indicates The bias term of the convolutional layer, represents the activation function, select as an activation function.

[0133] Classify the demand changes. The specific steps are as follows:

[0134] Perform maximum pooling on the convolution output to reduce the feature dimension. The expression is:

[0135] ;

[0136] in, Represents the features after pooling;

[0137] Combined with CNN's classification capabilities, the system can accurately determine the priority of different demand types and provide support for generating the best change plan;

[0138] The pooled features are passed through the fully connected layer to output the classification results: including function enhancements, bug fixes, and document updates. The expression is:

[0139] ;

[0140] in, Represents the classification result, Softmax represents the activation function, represents the weight of the fully connected layer, represents the bias term of the fully connected layer, Indicates the dimension index of the output;

[0141] Through NLP technology, the system can extract text data from project management tools in real time and perform intelligent analysis, automatically identifying project change requirements, reducing dependence on developers and improving the level of automation in project management.

[0142] According to the classified demand changes and the current project progress and resource status, the impact of demand changes is evaluated and the best change plan is generated;

[0143] The impact of demand changes on project progress is:

[0144] ;

[0145] in, An index variable representing a requirement change, Indicates a change in requirements The impact of changes on the project schedule, Indicates a change in requirements The complexity of Represents the classification result The type of demand change. Indicates the time remaining in the project;

[0146] The impact of demand changes on resource utilization improvements is:

[0147] ;

[0148] in, Indicates a change in requirements The impact of improvements in resource utilization, Indicates the total number of resource types. Represents the index of the resource. Indicates Resources in time availability, represents the smoothing term;

[0149] The impact of demand changes on demand satisfaction is:

[0150] ;

[0151] in, Indicates a change in requirements The impact of changes in demand satisfaction, Indicates a change in requirements The progress goals to be achieved, Indicates a change in requirements expected completion time;

[0152] According to the classification results and current project status (project progress time left Resource Utilization , Target progress of requirement changes , Expected completion time of requirement changes , the complexity of demand changes ), evaluate the impact of demand changes on progress, resources and demand satisfaction, the expression is:

[0153] ;

[0154] in, Represents the comprehensive impact index of demand changes;

[0155] Introducing an optimization function , combined with the impact index of all demand changes and the current status of the project to determine the optimal solution, the expression is:

[0156] ;

[0157] in, Indicates the priority coefficient of each requirement change, represents the penalty factor, Indicates the total number of requirement changes;

[0158] Generate the best change plan by solving the optimization function;

[0159] According to the classification results and impact index , priority processing The biggest change in requirements;

[0160] when When insufficient resource utilization is displayed, corresponding resources are added to meet the requirements of demand changes;

[0161] when When it is shown that the project progress is affected, adjust the project plan, extend the construction period and increase the input of personnel;

[0162] when When the demand satisfaction level decreases, tasks with high demand satisfaction are prioritized;

[0163] Generate the best change plan to ensure the rational allocation of resources and optimization of development progress, forming a closed-loop system of dynamic adjustment, real-time optimization, and intelligent decision-making, effectively improving the overall efficiency and management level of software development.

[0164] In summary, the present invention forms a complete intelligent project management system based on deep reinforcement learning. The efficient collection and preprocessing of data provide strong data support for subsequent prediction models. The application of deep neural networks enables the system to accurately predict future changes in demand and generate optimization suggestions. Through NLP and CNN technology, the system can automatically identify and classify demand changes, and generate the best change plan to ensure the rational allocation of resources and optimization of development progress. Each step is closely related to form a closed-loop system with dynamic adjustment, real-time optimization, and intelligent decision-making, which effectively improves the overall efficiency and management level of software development.

[0165] Example 2, referring to Table 1, is the second example of the present invention. In order to further verify the technical solution of the present invention, experimental simulation data of a software development efficiency management system based on deep reinforcement learning is provided.

[0166] In order to verify the advantages of the software development efficiency management system based on deep reinforcement learning, an experiment was conducted to compare the performance of the prior art and the present invention in software development projects. The experiment selected a medium-sized software development project with a project cycle of 6 months and 20 participants. The experimental process is divided into two stages: the first stage uses the traditional project management tool Jira for demand management and resource allocation, and the second stage uses the system of the present invention, combined with deep reinforcement learning and natural language processing technology, to predict demand and optimize resource allocation.

[0167] First, in the two phases of the experiment, Jira was used to collect project development data, including task assignment data, task completion status, time records, user story data, and defect report data. For traditional methods, the data is mainly used for manual analysis and decision-making, while the system of the present invention preprocesses the data, including data cleaning, removal of irrelevant fields, use of average values ​​to handle missing values, format conversion, etc., to ensure the integrity and consistency of the input data.

[0168] Then, based on the preprocessed data, the system of the present invention constructs a demand forecasting model through a deep neural network. The model predicts future demand changes based on the state space and action space of the project, and defines a reward function to comprehensively evaluate changes in project progress, resource utilization, and demand satisfaction. Based on these prediction results, the system automatically generates optimization suggestions and adjusts development resources and schedule plans based on these suggestions.

[0169] In the second phase of the experiment, the system extracted task descriptions, defect reports, and requirement change information from Jira in real time through natural language processing (NLP) technology. The system used convolutional neural networks (CNN) to classify the extracted data and identify the priority and type of requirement changes. Combining this information, the system evaluated the impact of the requirement changes and generated the best change plan to ensure reasonable resource allocation, project schedule optimization, and improved requirement satisfaction.

[0170] The experimental data was recorded and compared using the following parameters: project progress (percentage of tasks completed on time), resource utilization (rate of development resource utilization), requirement satisfaction (degree of requirement satisfaction), as well as defect fixing time and development efficiency (average task completion time per developer).

[0171] The details are shown in Table 1 below:

[0172] Table 1 Software development efficiency system performance comparison table

[0173]

[0174] The on-time completion rate of the project progress of the system of the present invention is 95%, which is significantly improved compared with 82% of the prior art. This is because the present invention uses a deep neural network for demand forecasting, which can effectively predict future changes in demand and adjust resource allocation and development strategies in advance according to the forecast results, thereby reducing delays caused by changes in demand or improper resource allocation during project development. By optimizing resource allocation, the system improves the on-time delivery rate of development tasks and ensures that the project can be completed within the scheduled time.

[0175] The resource utilization rate of the present invention reaches 92%, which is a significant improvement over the 76% of the prior art. Traditional methods rely more on manual decision-making in resource allocation, which easily leads to waste or shortage of resources. The present invention optimizes resource allocation in real time through the reward mechanism of deep reinforcement learning to ensure that development resources are maximized. This not only improves the efficiency of resource utilization, but also reduces the waste of project costs.

[0176] The demand satisfaction rate has been improved from 70% in the prior art to 93% in the present invention. Through the application of natural language processing and convolutional neural networks, the system of the present invention can monitor and classify demand changes in real time, automatically generate the best change plan, and ensure that the demand can be responded to and processed in a timely manner. This reduces customer dissatisfaction caused by untimely response to demand changes and significantly improves demand satisfaction.

[0177] The present invention shortens the defect repair time from 5.2 days of the traditional method to 3.1 days. This improvement is due to the system's automatic extraction and classification of defect reports, which can quickly identify high-priority defects and prioritize resource allocation for processing, thereby significantly shortening the defect repair cycle.

[0178] The present invention also significantly improves development efficiency. The average number of tasks completed by each developer per month has increased from 15 in the prior art to 21. Thanks to the deep learning model's accurate prediction of demand and reasonable allocation of resources, developers can complete tasks more efficiently, reducing waiting time or task backlogs caused by insufficient or over-allocation of resources.

[0179] The response time for demand changes in the present invention is significantly shortened from the traditional 4.5 days to 2.3 days. Through real-time monitoring and automatic classification of demand changes by NLP, the system can quickly identify demand changes and generate processing solutions, reducing the delay of manual processing of demand changes in traditional methods.

[0180] The system of the present invention also successfully reduced the number of project delay days from 12 days to 3 days. This result once again proves that the system can effectively reduce the risk of project delays and ensure that the project is completed on time by optimizing resources and progress management.

[0181] 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 development efficiency management system based on deep reinforcement learning, characterized by: Including data collection module, model building module, optimization suggestion module, progress monitoring module, impact assessment module, and change plan module; The data acquisition module uses a project management tool to collect software development data and perform preprocessing; The model building module uses a deep neural network to build a demand forecasting model based on the preprocessed software development data to predict future demand changes; The optimization suggestion module generates optimization suggestions according to future demand changes; The progress monitoring module monitors the project progress in real time through natural language processing (NLP) according to the optimization suggestions, identifies and extracts requirement change information, and classifies the requirement changes; The impact assessment module assesses the impact of the demand changes based on the classified demand changes combined with the progress and resource status of the current project; The change plan module generates the best change plan according to the impact of the demand change; The project management tool is used to collect software development data and perform preprocessing. The specific steps are as follows: Use the project management tool Jira to collect task assignment data, completion status data, time recording data, user story data, and defect report data; Clean the acquired task assignment data, completion status data, time record data, user story data, and defect report data, use the average value to handle missing values, and perform format conversion; The processed data are merged into a complete software development data set, and the software development data set is set to x = {x1, x2, x3, x4, x5}, where x1 represents task assignment data, x2 represents completion status data, x3 represents time record data, x4 represents user story data, and x5 represents defect report data; The demand prediction model is constructed based on the preprocessed software development data set using a deep neural network. The specific steps are as follows: Define the state space and action space: S t ={x1(t),x2(t),x3(t),x4(t),x5(t)}; Among them, S t Represents the state space of the project at time t, where t represents the time index; <h2 style=";text-align:left;direction:ltr">A<h2 style=";text-align:left;direction:ltr"> t <h2 style=";text-align:left;direction:ltr"> (a1,a2,...,a)<h2 style=";text-align:left;direction:ltr"> m <h2 style=";text-align:left;direction:ltr">}; Among them, A t represents the action space of the project at time t, a represents the demand change and development resource adjustment operation, and m represents the index variable of the demand change and development resource adjustment operation; Define the reward function R(S t , A t ) to optimize demand forecasting, taking into account the following factors: Among them, ΔP represents the change in project progress, P t represents the project progress at time t, P t+1 represents the project progress at time t+1, ΔR represents the improvement in resource utilization, and R t represents the resource utilization at time t, R t+1 represents the resource utilization at time t+1, ΔU represents the change in demand satisfaction, and U t represents the demand satisfaction at time t, U t+1 Indicates the satisfaction of demand at time t+1; Reward function R(S t , A t ) is expressed as: R(S t ,A t )=α1·ΔP+α2·ΔR+α3·ΔU; Among them, α1 represents the weight parameter of ΔP, α2 represents the weight parameter of ΔR, and α3 represents the weight parameter of ΔU; Calculate the action value function based on the reward function; The action value function is calculated according to the reward function, and the expression is: Among them, Q represents the action value function, E represents the expected value, γ represents the discount factor, max represents the maximum value, S t+1 represents the state space at time t+1, A t+1 represents the action space at time t+1; The specific steps of predicting future demand changes are as follows: The demand forecasting model expression is: Among them, Y t represents the predicted value of demand change at time point t, i represents x i (t) index variable, x i (t) represents the value of the i-th type of data in the software development dataset at time point t, β i represents the regression coefficient, n represents x i The number of 2. The software development efficiency management system based on deep reinforcement learning according to claim 1, characterized in that: The specific steps of generating optimization suggestions according to future demand changes are as follows: When Y t When it is >0, it indicates that future demand will increase, and it is necessary to increase development resources or improve production efficiency to meet the demand; When Y t When <0, it indicates that future demand will decrease and development resources should be appropriately reduced; When Y t When =0, it indicates that the demand remains stable and the current resource allocation is maintained.

3. The software development efficiency management system based on deep reinforcement learning as claimed in claim 2, characterized in that: The specific steps of monitoring the project progress in real time and identifying and extracting demand change information through natural language processing (NLP) according to the optimization suggestions are as follows: Through natural language processing (NLP), task description data, defect report data, and requirement change data of the software development team are collected in real time from the project management tool Jira. After removing stop words, part-of-speech tagging, named entity recognition, and word embedding processing, the text vector z is obtained. A text classification model based on convolutional neural network (CNN) is used to classify the demand change information. Multiple convolution kernels are used to extract the local features of the demand change information text. The convolution operation expression is: h g =ReLU(W g ·With j +b g ); Among them, j represents the index variable of the text vector, h g represents the output feature of the g-th convolutional layer, W g represents the convolution kernel weight of the g-th convolution layer, z j represents the jth input text vector, b g represents the bias term of the g-th convolutional layer, and ReLU represents the activation function.

4. The software development efficiency management system based on deep reinforcement learning as claimed in claim 3, characterized in that: The specific steps for classifying demand changes are as follows: Perform maximum pooling on the convolution output to reduce the feature dimension. The expression is: i g =max(h g ); Among them, θ g Represents the features after pooling of the g-th convolutional layer; The pooled features are passed through the fully connected layer to output the classification results, and the expression is: y=Softmax(W f ·θ g +b f ): Among them, y represents the classification result, Softmax represents the activation function, and W f represents the weight of the fully connected layer, b f represents the bias term of the fully connected layer, and f represents the dimension index of the output.

5. The software development efficiency management system based on deep reinforcement learning as claimed in claim 4, characterized in that: The impact of the demand changes is evaluated based on the classified demand changes combined with the current project progress and resource status. The specific steps are: The impact of demand changes on project progress is: Where v represents the index variable of demand change, ΔP v represents the impact of requirement change v on project schedule, C v represents the complexity of the requirement change v, y v Indicates the type of classification result corresponding to the requirement change v, T r Indicates the time remaining in the project; The impact of demand changes on resource utilization improvements is: Among them, ΔR v represents the impact of the requirement change v on the improvement of resource utilization, d represents the total number of resource types, c represents the index of the resource, and R c，t represents the availability of the cth resource at time t, ∈ represents the smoothing term; The impact of demand changes on demand satisfaction is: Among them, ΔU v represents the impact of the demand change v on the change of demand satisfaction, P v Indicates the progress target required to complete the requirement change v, T v It represents the expected completion time of the requirement change v; According to the classification results and the current project status, the impact of the demand change on the progress, resources and demand satisfaction is evaluated. The expression is: Among them, I v Represents the comprehensive impact index of demand changes.

6. The software development efficiency management system based on deep reinforcement learning as claimed in claim 5, characterized in that: Generate the best change plan based on the impact of the demand change. The specific steps are as follows: The optimization function J is introduced to determine the optimal solution by combining the impact index of all requirement changes and the current status of the project. The expression is: Among them, λ v represents the priority coefficient of each requirement change, τ represents the penalty factor, and k represents the total number of requirement changes; Generate the best change plan by solving the optimization function; According to the classification result y and the comprehensive impact index I v , priority processing I v The biggest change in requirements; When ΔR v When insufficient resource utilization is displayed, corresponding resources are added to meet the requirements of demand changes; When ΔP v When it is shown that the project progress is affected, adjust the project plan, extend the construction period and increase the input of personnel; When ΔU v When the demand satisfaction level is decreasing, tasks with high demand satisfaction are prioritized.

Images