Big data-based software development system management method

Abstract

The application discloses a software development system management method based on big data, relates to the field of software development management, and comprises six modules of project data full-dimension collection, demand change-influence link dynamic quantification, multi-dimensional data driven resource-task dynamic adaptation and the like. The core innovation is as follows: the former constructs a demand-task-resource association graph, combines a gradient boosting regression mapping model, quantifies a demand change influence range and a work load increment; and the latter realizes resource-task accurate matching through a personnel capacity vector, a task demand vector and a dynamic adaptation algorithm, contains secondary distribution, full-amount adaptation optimization fault-tolerant mechanism. The method solves the problems of traditional methods, such as fuzzy demand change influence and unreasonable resource allocation, archives achievements to form knowledge reuse, and is suitable for large software development projects in adaptation.

Description

Technical Field

[0001] This invention relates to the field of software development management, specifically a software development system management method based on big data. Background Technology

[0002] Existing software development system management methods suffer from two core problems: First, requirement change management is rudimentary, only recording the content of changes without tracing their full-chain impact on downstream tasks, resource allocation, and project progress. This leads to vague impact scope, inaccurate workload assessments, and frequent rework. Second, resource allocation relies on management experience, failing to dynamically adjust based on personnel skills, task complexity, and real-time project progress. This results in low resource-task matching, overloaded core resources, idle general resources, and low development efficiency. Furthermore, existing methods lack in-depth fusion and analysis of multi-dimensional data, making it difficult to predict schedule delays and quality risks, further impacting project progress efficiency. Summary of the Invention

[0003] The purpose of this invention is to provide a software development system management method based on big data to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a software development system management method based on big data, comprising the following steps: Step S1: Comprehensive Project Data Collection: Collect requirement data, task data, personnel data, progress data, and quality data during the software development process to obtain the collected data, and then transmit the collected data to the central control module. Step S2, Dynamic Quantification of Impact Links of Requirement Changes: Based on the collected data, construct a requirement-task-resource relationship graph, quantify the impact scope and workload increment of requirement changes, and output an impact data assessment report to the central control module; Step S3, Multi-dimensional data-driven dynamic resource-task adaptation: Combining personnel capability profiles, task complexity coefficients, and real-time project progress from the collected data, resources are allocated and task priorities are adjusted through dynamic adaptation algorithms, and the resource adaptation results are output to the central control module. Step S4, Progress and Quality Early Warning: Based on the quantitative impact data in Step S2 and the resource adaptation results in Step S3, the central control module monitors project progress and quality risks in real time and triggers tiered early warnings. Step S5, Results Archiving: Classify and archive the data development results and management data within the central control module to form a reusable knowledge base.

[0005] Preferably, the specific content of the demand data, task data, personnel data, progress data, and quality data collected by the data acquisition software in step S1 includes: Requirements data includes requirement ID, requirement name, function description, priority P0-P3, business module affiliation, proposer, proposal time, and planned delivery node, ensuring coverage of key information throughout the entire requirement lifecycle; Task data includes task ID, task type (development / testing / design / documentation), functional module, complexity level (1-5), estimated workload (person-days), responsible person ID, start time, planned completion time, dependent task IDs, and current status (not started / in progress / completed / blocked).

[0006] Personnel data includes personnel ID, name, position (front-end developer / back-end developer / test engineer / architect), skill tags (such as Java, Python, UI design, automated testing), skill proficiency (0-10 points, based on a combination of skill assessment and project performance), historical project experience (project ID, type of task), maximum workload (person-days / week), current assigned workload, and workload rate (calculated in real time). Progress data includes task completion rate (completed workload / estimated workload), milestone progress (actual completion time of key nodes vs. planned time), task delay duration, and cross-module collaboration progress synchronization data, updated every 5 minutes. Quality data includes code submission frequency, code defect rate (number of defects / number of lines of code), test pass rate, defect fix time, code review pass rate, and documentation completeness score (1-10 points). The data acquisition software employs a combination of real-time acquisition and scheduled synchronization. It connects to development tools (code repositories, project management tools) and personnel terminals via API interfaces, with a data update cycle of ≤5 minutes. After data acquisition, it undergoes cleaning and deduplication before being transmitted to the central control module. The central control module categorizes and stores the data in a distributed database, establishing a data index (associated by requirement ID, task ID, and personnel ID) to provide rapid query support for association graph construction and matching algorithm calculations.

[0007] Preferably, the specific implementation steps of step S2, "Requirement Change - Impact Link Dynamic Quantification," are as follows: Step S2.1, Construction of the association graph: Define the core parameter: Requirement ID is Task ID is Resource ID is The direct relationship between requirements and tasks is as follows: The allocation relationship between tasks and resources is as follows: Based on the collected dependency data, a three-layer relationship graph of requirements, tasks, and resources is constructed, and nodes are stored using a graph database. Assign edges (associations / assignments) and label the relationship strength weights. ,in express and The relationship strength weights, ranging from 0 to 1, are obtained through training on historical data. Step S2.2, Impact on Link Tracing: When demand When changes occur, a depth-first search algorithm is used to traverse the association graph and trace directly related tasks. And indirectly related tasks (downstream dependent tasks derived from directly related tasks). To form a complete influence chain ; Synchronous traceability of resources involved in the link Record the current workload of resources and the status of task associations; Step S2.3, Calculation of Influence Coefficient: Define the influence coefficient. For demand Changes to the task The degree of influence, calculated using the following formula: ,in Adjustment factor for task complexity For the task The complexity level is 1-5. big The larger, Define the total influence range coefficient. ,in For the task Project weight (set based on task priority and milestone correlation, with a value of 0-1). The closer it is to 1, the wider the scope of influence; Step S2.4, Workload Increment Assessment: Construct an impact coefficient-workload increment mapping model based on historical data, and input... With the task Original workload Calculate the workload increment for a single task ,in This is a historical correction factor (obtained from historical data of similar projects). Total workload increment It also outputs impact reports and workload adjustment suggestions, which are then transmitted to the central control module.

[0008] Preferably, the specific implementation steps for workload increment assessment in step S2.4 are as follows: Step S2.4.1, Construction and Validation of the Influence Coefficient-Workload Increment Mapping Model: Model data source: Historical data from several similar software development projects were collected, and valid samples were selected. Valid samples include, but are not limited to, complete requirement change records and task impact coefficients. Original workload Actual workload increase Task types and project scales are considered to ensure that effective samples cover different business scenarios (e-commerce, government affairs, industrial software, etc.) and task types (development, testing, design, documentation). Feature variable definition: based on influence coefficient The core independent variable is the original workload of the task. Task type is an auxiliary independent variable, and the actual workload increment is... A multivariate mapping model is constructed with the task type as the dependent variable. This indicates that Development = 1, Testing = 2, Design = 3, and Documentation = 4. Model Fitting and Calibration: The gradient boosting regression algorithm is used for model training, and the model parameters are optimized through 5-fold cross-validation to ensure that the goodness of fit of the multivariate mapping model is R²≥0.95 and the prediction error is ≤7%. After training, the multivariate mapping model is fixed locally in the module, supporting real-time access and subsequent iterative updates (model optimization is automatically triggered every 20 new project samples). Step S2.4.2: Preprocessing and validity verification of input parameters: Influence coefficient Verification: Extract the individual task influence coefficients calculated in step S2.3. Verify whether it is within the valid range. Inside( The impact exceeded a reasonable range and was therefore identified as an outlier; a similar task was then performed. (median substitution) to ensure the reliability of input coefficients; Original workload Confirmed: Task obtained from the project data full-dimensional collection module. Original estimated workload (Unit: man-days), synchronous verification Reasonableness (e.g., development tasks) Human-based, testing-type tasks (If the number of people exceeds the limit, it will be corrected after confirmation by the task creator) to avoid deviations in incremental calculations due to errors in the original data. Historical correction coefficient Matching: Based on task type Match the corresponding value, Based on historical data from similar projects, the value range is clearly defined: Development tasks (Highly variable technical implementation complexity), testing tasks (Verification workload is relatively controllable), design tasks (The impact of the plan adjustment is moderate) Document-related tasks (The workload increase for content updates is relatively small), ensuring the coefficient. Adapted to task characteristics; Step S3.4.3, Increment of workload for a single task Precise calculations: Formula substitution calculation: Substitute the preprocessed... Substitute into the core calculation formula The calculation is performed by the hardware-accelerated computing unit built into the demand change-impact link dynamic quantification module, with a single calculation time of ≤5ms, ensuring real-time performance; Calculation result verification: Perform a reasonableness check and set threshold rules: If If the incremental workload is close to the original workload, a secondary verification will be automatically triggered to recheck the data. The calculation process (such as association weight) Complexity correction factor (Is it accurate?) For man-days (with minimal increments), a uniform calculation of 0.1 man-days is used (to avoid impacting management efficiency with minute increments), ensuring results accurately reflect the actual execution scenario; Result labeling: for each task... Add tags, including the type of impact (feature addition / parameter adjustment / requirement cancellation) and the calculation basis (). value, (Value), which facilitates subsequent traceability and review; Step S2.4.4, Total Workload Increment Summary and dimensional breakdown: Total increment calculation: effective for all affected tasks Summing gives the total increase in workload due to requirement changes. (j=1 to the total number of affected tasks), synchronously calculate the proportion of the total increment to the remaining workload of the project. Used for subsequent progress risk prediction; Multi-dimensional breakdown: Breaking down tasks by type, priority, and business module. Perform categorization and summarization to generate dimensional incremental data (such as the total incremental amount of development tasks). Total increment of P0 level tasks ); Step S2.4.5: Generation of Impact Link Report and Workload Adjustment Suggestions: Core content of the report: Visualized impact chain diagram: Marking nodes of requirement changes and nodes of affected tasks (by...) (Annotate node dimensions) and associate resource nodes to intuitively present the impact propagation path; List of increased workload for affected tasks: including but not limited to task ID, task name, value, , value, Incremental percentage Task priority field; Schedule risk prediction: based on and Predict the overall project delay time (e.g.) Human and Heaven The delay is predicted to be 3-7 days. Humanity and Heaven, the prediction was not significantly delayed); Workload adjustment suggestions: Combined with Size provides differentiated suggestions: It is recommended to extend the task duration by 3-5 days and prioritize the allocation of core resources; It is recommended to extend the construction period by 1-3 days; It is recommended to maintain the original construction period and absorb the incremental workload by optimizing the workflow.

[0009] Data transmission: the calculated results Dimensional incremental data and complete impact chain reports are transmitted in real time to the central control module via an encrypted protocol, and simultaneously synchronized to the project management terminal, triggering subsequent resource-task dynamic adaptation processes and schedule adjustment decisions.

[0010] Preferably, the specific implementation logic of the multi-dimensional data-driven resource-task dynamic adaptation in step S3 is as follows: Step S3.1, Core Parameter Quantization and Dynamic Update: Personnel Capability Profile Construction: Personnel Capability Vector , For the number of skill types, The quantification method is as follows: skills assessment score (40%) + historical similar task completion quality score (30%) + collaboration evaluation score (20%) + task delivery efficiency score (10%), which is recalculated every two weeks. To ensure that capability profiles are accurate in real time; Task requirement vector construction: Task requirement vector , The quantification method is based on task type (e.g., Java development tasks and Java skills). Complexity level () Level Increase by 20%), business priority (P0 level tasks) (Improvement by 10%) Comprehensive calculation, automatically generated when task is created, and updated synchronously when requirements change.

[0011] Auxiliary parameter calculation: personnel load rate (Maximum workload capacity is set based on historical performance, such as core development personnel) (Man-days / week); Task urgency , For an urgent mission, For low-urgency tasks; task complexity The assessment is based on a combination of factors including code size, number of dependencies, and technical difficulty (Level 1 is the simplest, Level 5 is the most complex). Core parameter definition and quantization: Personnel Capability Vector: Defining Personnel The capability vector is ,in For the number of skill types, for In skills Proficiency level (value 0-10, obtained by combining historical performance and skills assessment).

[0012] Task requirement vector: Defines the task The demand vector is ,in For the task skills The demand level (value 0-10, derived from task type and complexity).

[0013] Auxiliary parameter: Current staff load rate , Values ​​range from 0 to 1; Task urgency (Based on project schedule deviation settings, value ranges from 0 to 1), Task complexity (Values ​​range from 1 to 5); Step S3.2, Calculation of capability matching degree: Using cosine similarity algorithm to calculate personnel With the task Basic match degree: ;in For vector dot product, Let the vector magnitude be , The closer to 1, the higher the ability matching degree; Step S3.3, Dynamic Adaptation Algorithm: Construct the adaptation objective function: The constraints are: ,in For personnel Maximum workload capacity (based on historical data).

[0014] Calculate the overall fit score ,in These are the weighting coefficients, and The initial values ​​are 0.5, 0.3, and 0.2, respectively, and can be dynamically adjusted according to the project stage; according to Sort in descending order for each task Assign the person with the highest overall score If the workload exceeds the constraints, a secondary allocation (adjusting the resource allocation for low-priority tasks) will be triggered. Step S3.4, Dynamic Adjustment Mechanism: Recalculate every 30 minutes based on the latest progress data and personnel load data. Fine-tuning of resource allocation; when changes in demand lead to When needed, a full-scale adaptation and optimization will be triggered immediately to ensure that resources and tasks are always optimally matched.

[0015] Preferably, the specific implementation steps of step S4 are as follows: Step S4.1, Definition and Calculation of Early Warning Indicators: Extract input data from the central control module, including the following: Schedule early warning indicator: Schedule deviation rate The planned schedule is calculated based on the quantified workload and resource matching results, i.e., the total workload. , Milestone Node Delay Risk Value , High risk; Quality warning indicator: Code defect rate (Statistics based on the number of defects per thousand lines of code) Quality anomalies; timely defect repair rate , To fix the lag; test case pass rate , The test failed to meet the standards; Step S4.2, Tiered Early Warning Thresholds and Triggering Conditions: Progress warning levels: or Level 1 warning (severe delay). or Level 2 warning (general delay). or Level 3 warning (slight delay); Quality early warning classification: or or Level 1 warning (serious quality problem). or or Level 2 warning (general quality issue); Triggering mechanism: Real-time monitoring of early warning indicators. An early warning is triggered immediately when an indicator reaches a threshold. At the same time, the early warning priority is adjusted based on the impact level of the requirement change (major / moderate / minor) (e.g., major impact + level 1 progress warning, with the highest priority). Early warning response and coordination: Real-time monitoring of early warning indicators; triggering an early warning immediately when an indicator reaches a threshold; and adjusting the early warning priority based on the impact level of changes in demand. Early warning response and coordination: Early warning information push: including early warning level, involved task ID, specific indicator value, and risk cause analysis (e.g., progress warning reason: caused by requirement change). If the resource allocation is not adjusted in a timely manner, suggestions for handling the situation (such as a level 1 progress warning suggestion: add 1-2 skilled personnel to prioritize core tasks) should be pushed to the project manager, resource administrator, and relevant task managers.

[0016] System linkage: After the early warning is triggered, the central control module automatically calls the resource-task dynamic adaptation module to generate a resource adjustment plan based on the handling suggestions (such as temporarily allocating idle resources and adjusting the progress of low-priority tasks), and at the same time updates the progress plan, forming a closed loop of early warning-analysis-handling-optimization.

[0017] Step S4.3, Early Warning Information Push: Includes early warning level, involved task ID, specific indicator value, and risk cause analysis (e.g., progress warning reason: caused by requirement change). If the resource matching is not adjusted in time, suggestions for handling (such as the first-level progress warning suggestion: add 1-2 skilled personnel to prioritize core tasks) are pushed to the project leader, resource administrator and relevant task leaders.

[0018] System linkage: After the early warning is triggered, the central control module automatically calls the resource-task dynamic adaptation module to generate a resource adjustment plan based on the handling suggestions (such as temporarily allocating idle resources and adjusting the progress of low-priority tasks), and at the same time updates the progress plan, forming a closed loop of early warning-analysis-handling-optimization.

[0019] Preferably, the software development system management platform based on big data includes a project data full-dimensional collection module, a requirement change-impact dynamic quantification module, a multi-dimensional data-driven resource-task dynamic adaptation module, a central control module, a progress and quality early warning module, and a deliverable archiving module. The project data full-dimensional collection module collects multi-dimensional data from the entire software development process, covering requirement data, task data, personnel data, progress data, and quality data. It connects to development management tools and personnel terminals through interfaces, and sends the data to the central control module after cleaning and standardization. The dynamic quantification module for the demand change-impact link receives the collected data after processing by the central control module, and then traces the full-link impact of the demand change by constructing a demand-task-resource association graph, quantifies the impact coefficient and workload increment, outputs an accurate impact assessment report, and inputs the impact data assessment report into the central control module. The multi-dimensional data-driven resource-task dynamic adaptation module receives resource adaptation results and task data from the central control module. By constructing personnel capability vectors and task requirement vectors, it calculates a comprehensive adaptation score to achieve optimal allocation of resources and tasks. It also has a secondary allocation fault tolerance mechanism to solve resource overload problems and ensure balanced resource load. Its allocation results are synchronized to the central control module and the progress quality early warning module. The central control module receives data collected by the project data full-dimensional acquisition module and forwards it to the requirement change-impact link dynamic quantification module and the multi-dimensional data-driven resource-task dynamic adaptation module. Receive the impact assessment report from the dynamic quantification module of the impact of requirement changes and the resource allocation scheme of the resource-task dynamic adaptation module driven by multi-dimensional data, integrate and generate execution instructions, link the progress and quality early warning module to push early warning information, and synchronously update the data of the entire system to ensure that all modules work together. The progress and quality early warning module calculates core indicators such as progress deviation rate and defect rate based on the quantitative impact data, resource adaptation results and real-time progress and quality data forwarded by the central control module. It triggers graded early warnings according to thresholds and links the central control module and the adaptation module to push disposal suggestions and trigger dynamic resource adjustments in advance, so as to avoid project delays and quality risks in advance and ensure that the project progresses as planned. The results archiving module receives full-process data (development results and management data) transmitted from the central control module. After classification, screening, standardization, and extraction of reusable experience, a tagged knowledge base is constructed. This not only completes the systematic archiving of project data, but also provides reusable experience such as requirement change handling cases and resource allocation schemes for subsequent similar projects, thereby achieving iterative optimization of system management capabilities.

[0020] Compared with the prior art, the beneficial effects of the present invention are: The impact of demand changes is accurately quantified, and the rework rate is significantly reduced: This invention constructs a demand-task-resource association map through a dynamic quantification module of the demand change-impact link, and combines it with a mapping model constructed by the gradient boosting regression algorithm (goodness of fit R²≥0.95, prediction error≤7%) to achieve accurate calculation of the impact range, impact coefficient and workload increment, thus solving the problem of fuzzy impact assessment in traditional methods. Optimal resource-task matching significantly improves utilization efficiency: The multi-dimensional data-driven resource-task dynamic adaptation module calculates the cosine similarity between personnel capability vectors and task requirement vectors, combined with dynamic adaptation algorithms and secondary allocation and full-scale adaptation optimization mechanisms, thereby improving the matching degree between resources and tasks and the balance rate of personnel workload; and reducing the overload rate of core resources and the idle rate of ordinary resources. Precise and proactive risk management ensures more stable project progress: This invention's schedule and quality early warning module, based on quantified impact data and resource adaptation results, constructs a multi-dimensional, hierarchical early warning system. It monitors key indicators such as schedule deviation rate and code defect rate in real time, predicting project delay risks 3-7 days in advance. Level 1 early warning response efficiency is improved, enabling rapid adjustment of resource allocation through system linkage, increasing on-time project delivery rates, and consistently controlling the defect rate per thousand lines of code below 0.3%. Knowledge reuse forms a closed loop, and management capabilities are continuously optimized: The invention's results archiving module tagged and archived development results and management data, extracted high-value reuse experiences (such as highly matched resource-task combinations and solutions for handling requirement changes), and built a searchable knowledge base. This improved efficiency in subsequent project requirement analysis, shortened resource allocation decision-making time, and enabled the system to continuously optimize management capabilities through data iteration, adapting to the needs of software development projects of different sizes and types. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the overall method flow of the present invention; Figure 2 This is a schematic diagram illustrating the workflow of dynamic quantification of the impact of requirement changes on the link in this invention. Figure 3 This is a schematic diagram of the multi-dimensional data-driven resource-task dynamic adaptation workflow of the present invention. Figure 4 This is a schematic diagram of the progress and quality early warning workflow of the present invention; Figure 5 This is a schematic diagram of the structure of the software development system management platform based on big data according to the present invention. Detailed Implementation

[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0023] Example 1 Please see Figure 1-4 This invention provides a technical solution: a software development system management method based on big data, comprising the following steps: Step S1: Comprehensive Project Data Collection: Collect requirement data, task data, personnel data, progress data, and quality data during the software development process to obtain the collected data, and then transmit the collected data to the central control module. The specific contents of the requirement data, task data, personnel data, progress data, and quality data collected by the data acquisition software in step S1 include: Requirements data includes requirement ID, requirement name, function description, priority P0-P3, business module affiliation, proposer, proposal time, and planned delivery node, ensuring coverage of key information throughout the entire requirement lifecycle; Task data includes task ID, task type (development / testing / design / documentation), functional module, complexity level (1-5), estimated workload (person-days), responsible person ID, start time, planned completion time, dependent task IDs, and current status (not started / in progress / completed / blocked).

[0024] Personnel data includes personnel ID, name, position (front-end developer / back-end developer / test engineer / architect), skill tags (such as Java, Python, UI design, automated testing), skill proficiency (0-10 points, based on a combination of skill assessment and project performance), historical project experience (project ID, type of task), maximum workload (person-days / week), current assigned workload, and workload rate (calculated in real time). Progress data includes task completion rate (completed workload / estimated workload), milestone progress (actual completion time of key nodes vs. planned time), task delay duration, and cross-module collaboration progress synchronization data, updated every 5 minutes. Quality data includes code submission frequency, code defect rate (number of defects / number of lines of code), test pass rate, defect fix time, code review pass rate, and documentation completeness score (1-10 points). The data acquisition software employs a combination of real-time acquisition and scheduled synchronization. It connects to development tools (code repositories, project management tools) and user terminals via API interfaces, with a data update cycle of ≤5 minutes. After acquisition, the data undergoes cleaning and deduplication before being transmitted to the central control module. This module categorizes and stores the data in a distributed database, establishing a data index (associated by requirement ID, task ID, and personnel ID) to provide rapid query support for association graph construction and matching algorithm calculations. Step S2, Dynamic Quantification of Impact Links in Requirement Changes: Based on the collected data, construct a requirement-task-resource relationship graph, quantify the impact scope and workload increment of requirement changes, and output an impact data assessment report to the central control module; the specific implementation steps are as follows: Step S2.1, Construction of the association graph: Define the core parameter: Requirement ID is Task ID is Resource ID is The direct relationship between requirements and tasks is as follows: The allocation relationship between tasks and resources is as follows: Based on the collected dependency data, a three-layer relationship graph of requirements, tasks, and resources is constructed, and nodes are stored using a graph database. Assign edges (associations / assignments) and label the relationship strength weights. ,in express and The relationship strength weights, ranging from 0 to 1, are obtained through training on historical data. Step S2.2, Impact on Link Tracing: When demand When changes occur, a depth-first search algorithm is used to traverse the association graph and trace directly related tasks. And indirectly related tasks (downstream dependent tasks derived from directly related tasks). To form a complete influence chain ; Synchronous traceability of resources involved in the link Record the current workload of resources and the status of task associations; Step S2.3, Calculation of Influence Coefficient: Define the influence coefficient. For demand Changes to the task The degree of influence, calculated using the following formula: ,in Adjustment factor for task complexity For the task The complexity level is 1-5. big The larger, Define the total influence range coefficient. ,in For the task Project weight (set based on task priority and milestone correlation, with a value of 0-1). The closer it is to 1, the wider the scope of influence; Step S2.4, Workload Increment Assessment: Construct an impact coefficient-workload increment mapping model based on historical data, and input... With the task Original workload Calculate the workload increment for a single task ,in This is a historical correction factor (obtained from historical data of similar projects). Total workload increment It also outputs impact reports and workload adjustment suggestions, which are then transmitted to the central control module.

[0025] The specific implementation steps for workload increment assessment in step S2.4 are as follows: Step S2.4.1, Construction and Validation of the Influence Coefficient-Workload Increment Mapping Model: Model data source: Historical data from several similar software development projects were collected, and valid samples were selected. Valid samples include, but are not limited to, complete requirement change records and task impact coefficients. Original workload Actual workload increase Task types and project scales are considered to ensure that effective samples cover different business scenarios (e-commerce, government affairs, industrial software, etc.) and task types (development, testing, design, documentation). Feature variable definition: based on influence coefficient The core independent variable is the original workload of the task. Task type is an auxiliary independent variable, and the actual workload increment is... A multivariate mapping model is constructed with the task type as the dependent variable. This indicates that Development = 1, Testing = 2, Design = 3, and Documentation = 4. Model Fitting and Calibration: The gradient boosting regression algorithm is used for model training, and the model parameters are optimized through 5-fold cross-validation to ensure that the goodness of fit of the multivariate mapping model is R²≥0.95 and the prediction error is ≤7%. After training, the multivariate mapping model is fixed locally in the module, supporting real-time access and subsequent iterative updates (model optimization is automatically triggered every 20 new project samples). Step S2.4.2: Preprocessing and validity verification of input parameters: Influence coefficient Verification: Extract the individual task influence coefficients calculated in step S2.3. Verify whether it is within the valid range. Inside( The impact exceeded a reasonable range and was therefore identified as an outlier; a similar task was then performed. (median substitution) to ensure the reliability of input coefficients; Original workload Confirmed: Task obtained from the project data full-dimensional collection module. Original estimated workload (Unit: man-days), synchronous verification Reasonableness (e.g., development tasks) Human-based, testing-type tasks (If the number of people exceeds the limit, it will be corrected after confirmation by the task creator) to avoid deviations in incremental calculations due to errors in the original data. Historical correction coefficient Matching: Based on task type Match the corresponding value, Based on historical data from similar projects, the value range is clearly defined: Development tasks (Highly variable technical implementation complexity), testing tasks (Verification workload is relatively controllable), design tasks (The impact of the plan adjustment is moderate) Document-related tasks (The workload increase for content updates is relatively small), ensuring the coefficient. Adapted to task characteristics; Step S3.4.3, Increment of workload for a single task Precise calculations: Formula substitution calculation: Substitute the preprocessed... Substitute into the core calculation formula The calculation is performed by the hardware-accelerated computing unit built into the demand change-impact link dynamic quantification module, with a single calculation time of ≤5ms, ensuring real-time performance; Calculation result verification: Perform a reasonableness check and set threshold rules: If If the incremental workload is close to the original workload, a secondary verification will be automatically triggered to recheck the data. The calculation process (such as association weight) Complexity correction factor (Is it accurate?) For man-days (with minimal increments), a uniform calculation of 0.1 man-days is used (to avoid impacting management efficiency with minute increments), ensuring results accurately reflect the actual execution scenario; Result labeling: for each task... Add tags, including the type of impact (feature addition / parameter adjustment / requirement cancellation) and the calculation basis (). value, (Value), which facilitates subsequent tracing and review. Step S2.4.4, Total Workload Increment Summary and dimensional breakdown: Total increment calculation: effective for all affected tasks Summing gives the total increase in workload due to requirement changes. (j=1 to the total number of affected tasks), synchronously calculate the proportion of the total increment to the remaining workload of the project. Used for subsequent progress risk prediction; Multi-dimensional breakdown: Breaking down tasks by type, priority, and business module. Perform categorization and summarization to generate dimensional incremental data (such as the total incremental amount of development tasks). Total increment of P0 level tasks ); Step S2.4.5: Generation of Impact Link Report and Workload Adjustment Suggestions: Core content of the report: Visualized impact chain diagram: Marking nodes of requirement changes and nodes of affected tasks (by...) (Annotate node dimensions) and associate resource nodes to intuitively present the impact propagation path; List of increased workload for affected tasks: including but not limited to task ID, task name, value, , value, Incremental percentage Task priority field; Schedule risk prediction: based on and Predict the overall project delay time (e.g.) Human and Heaven The delay is predicted to be 3-7 days. Humanity and Heaven, the prediction was not significantly delayed); Workload adjustment suggestions: Combined with Size provides differentiated suggestions: It is recommended to extend the task duration by 3-5 days and prioritize the allocation of core resources; It is recommended to extend the construction period by 1-3 days; It is recommended to maintain the original construction period and absorb the incremental workload by optimizing the workflow.

[0026] Data transmission: the calculated results Dimensional incremental data and complete impact chain reports are transmitted in real time to the central control module via an encrypted protocol, and simultaneously synchronized to the project management terminal, triggering subsequent resource-task dynamic adaptation processes and schedule adjustment decisions. Step S3, Multi-dimensional Data-Driven Dynamic Resource-Task Adaptation: Combining personnel capability profiles, task complexity coefficients, and real-time project progress from the collected data, resources are allocated and task priorities are adjusted through a dynamic adaptation algorithm, and the resource adaptation results are output to the central control module; the specific implementation logic is as follows: Step S3.1, Core Parameter Quantization and Dynamic Update: Personnel Capability Profile Construction: Personnel Capability Vector , For the number of skill types, The quantification method is as follows: skills assessment score (40%) + historical similar task completion quality score (30%) + collaboration evaluation score (20%) + task delivery efficiency score (10%), which is recalculated every two weeks. To ensure that capability profiles are accurate in real time; Task requirement vector construction: Task requirement vector , The quantification method is based on task type (e.g., Java development tasks and Java skills). Complexity level () Level Increase by 20%), business priority (P0 level tasks) (Improvement by 10%) Comprehensive calculation, automatically generated when task is created, and updated synchronously when requirements change.

[0027] Auxiliary parameter calculation: personnel load rate (Maximum workload capacity is set based on historical performance, such as core development personnel) (Man-days / week); Task urgency , For an urgent mission, For low-urgency tasks; task complexity The assessment is based on a combination of factors including code size, number of dependencies, and technical difficulty (Level 1 is the simplest, Level 5 is the most complex). Core parameter definition and quantization: Personnel Capability Vector: Defining Personnel The capability vector is ,in For the number of skill types, for In skills The proficiency level (valued from 0-10, derived from a combination of historical performance and skills assessment) is considered a skill level. The personnel capability vector is a multi-dimensional digital description of a developer's skill level. This refers to the total number of skill types involved in the project (such as Java, Python, UI design, automated testing, etc., which is set according to business requirements during project initialization). The quantification of (higher scores indicate greater proficiency) needs to combine objective data with comprehensive evaluation to ensure a complete and accurate assessment; a brief explanation is provided below, for example: Quantitative Dimensions and Weighting: Basic Skills Assessment (40%): Based on the personnel skills assessment data collected from the project data full-dimensional collection module, the technical leader organizes regular (quarterly) skills tests, covering theoretical knowledge, practical cases, etc. The test scores are converted from 0 to 10 points (e.g., a test score of 85 points corresponds to 7.5 points).

[0028] Historical performance contribution (30%): Extract data on similar skill tasks completed by the person in the past 3 months from the data collection module, and calculate the scores of the core indicators: task completion quality (defect rate reverse calculation), delivery efficiency (actual duration / planned duration reverse calculation), and demand satisfaction (product side evaluation). The three indicators are weighted and summed and then normalized to a score of 0-10.

[0029] Collaboration adaptability (20%): Collect collaboration evaluation data within the project team (such as cross-task communication efficiency, problem response speed), team management scores, and quantify them into excellent (9-10 points), good (7-8 points), average (5-6 points), fair (3-4 points), and poor (0-2 points).

[0030] Learning iteration ability (10%): Based on the skill improvement records of the data collection module (such as participating in training, obtaining certificates, and independently completing skill expansion tasks), the improvement frequency and effect are quantified (e.g., 1 point is added for completing 1 skill training and passing the assessment within the quarter, up to 10 points). Quantitative calculation steps: Step 1: Obtain the original scores for each dimension, which are denoted as V1 (basic skills), V2 (historical performance), V3 (collaboration adaptation), and V4 (learning iteration).

[0031] Step 2: Calculate the weighted score according to the weights: .

[0032] Step 3: Score verification and correction, if If the score exceeds the range of 0-10, it will be automatically truncated to the nearest boundary value (e.g., if the calculated score is 11, it will be corrected to 10). If data for a certain dimension is missing (e.g., new employees have no historical performance), the weight allocation will be adjusted (basic skills account for 60%, collaboration and adaptation account for 30%, and learning and iteration account for 10%). Dynamic update mechanism: Scheduled updates: Recalculation is triggered every two weeks to synchronize the latest task completion data and collaboration evaluation data from the data acquisition module, ensuring that the capability vector aligns with the real-time status of personnel. Triggered updates: When personnel complete significant skills training, obtain key certificates, or experience major task quality issues, a single-dimensional score update is immediately triggered, and the calculation is recalculated. .

[0033] Task requirement vector: Defines the task The demand vector is ,in For the task skills The demand level (value 0-10, derived from task type and complexity) is illustrated below with a brief example: Quantification benchmarks and correction rules: Benchmark Demand Setting: Based on task type data collected from the project data full-dimensional collection module, preset benchmark demand levels for each skill for different task types. For example: Java backend development task: Java skills required: 8-10 points; database skills: 7-9 points; interface design skills: 6-8 points; other irrelevant skills (such as UI design): 0 points. Automated testing task: Python skills 7-9 points, testing tool skills 8-10 points, business understanding skills 6-8 points, other irrelevant skills 0 points.

[0034] Complexity Correction: Combining the task complexity level of the data acquisition module (Levels 1-5), corrected according to the formula: (like At level 1, the correction factor is 1.2, and the demand increases by 20%.

[0035] Priority adjustment: Based on the task priority (P0-P3) of the data acquisition module, the P0 level task has an additional adjustment coefficient of 1.1 (increases demand by 10%), the P1-P2 level coefficient is 1.0, and the P3 level coefficient is 0.9 (decreases demand by 10%).

[0036] Quantization calculation steps: Step 1: Based on the task type, extract the baseline demand for each skill from the preset baseline library. .

[0037] Step 2: Demand level after computational complexity correction: .

[0038] Step 3: Calculate the final demand level: The result is rounded to one decimal place to ensure that the value is within the range of 0-10 (if it exceeds the range, it will be truncated to the boundary).

[0039] Dynamic update mechanism: Initial generation: When a task is created, the system automatically calculates the initial generation based on the task type, complexity, and priority. Synchronize to the module parameter library. Triggered update: When a requirement change (triggered by the requirement change-affecting dynamic quantization module) causes an adjustment to the task type, complexity, or priority, recalculate immediately. Update the task requirement vector.

[0040] Auxiliary parameter: Current staff load rate , Values ​​range from 0 to 1; Task urgency (Based on project schedule deviation settings, value ranges from 0 to 1), Task complexity (Values ​​range from 1 to 5); The following are brief examples illustrating the auxiliary parameters: Auxiliary parameters are used to adjust the priority and rationality of resource-task matching, covering personnel status (load rate). ) and task attributes (urgency) Complexity All data is derived from real-time project data to ensure an objective reflection of the actual situation. Current staff load rate ( (Values ​​range from 0 to 1); Core definition: Reflects the current workload of personnel, avoiding resource overload or idleness.

[0041] Quantification steps: Step 1: Extract personnel from the project data full-dimensional collection module Assigned but unfinished workload (Unit: man-days, including estimated workload for all ongoing and pending tasks).

[0042] Step 2: Extract personnel Maximum workload (Unit: man-days / week, set based on historical performance data, such as core development personnel) Man-days / week, Junior Developer (People / days / week).

[0043] Step 3: Calculate the load factor: If the calculation result is >1, then take 1 (complete overload); if it is <0, then take 0 (complete idle).

[0044] Dynamic updates: The task status data of the data acquisition module (such as task completion, newly assigned tasks) is synchronized every 5 minutes, and updated in real time. .

[0045] Task urgency ( (Values ​​0-1) Core definition: Reflects the urgency of completing a task on schedule; higher priority urgent tasks should be allocated resources first.

[0046] Quantification steps: Step 1: Extract tasks from the project data full-dimensional collection module Planned construction period Current remaining time ( ).

[0047] Step 2: Calculate the schedule deviation coefficient: If If so, the progress will be without deviation. (Low urgency); if ,but (like ).

[0048] Step 3: Boundary Correction The time is calculated at 0.9 (maximum urgency). The minimum urgency level is calculated as 0.1.

[0049] Dynamic updates: Progress data is synchronized and recalculated every 10 minutes. If the task priority is adjusted, additional corrections will be made (P0 level tasks). P3 level mission ).

[0050] Task complexity ( (Values ​​1-5); Core definition: Reflects the difficulty of task execution. High-complexity tasks require highly capable personnel with low workload.

[0051] Quantification steps: Step 1: Determine the evaluation dimensions: code volume (or task output volume), number of dependencies (number of tasks / modules requiring collaboration), technical difficulty (whether it involves new technologies / complex logic), and business complexity (whether it is a core business / new business).

[0052] Step 2: Score each dimension on a scale of 1-5 (level 1 being the simplest and level 5 the most complex), and then sum the scores using a weighted average (code volume 30%, dependencies 25%, technical difficulty 30%, business complexity 15%) to obtain the total score. .

[0053] Step 3: Map complexity levels by total score: .

[0054] Dynamic updates: When a task is created, the technical lead conducts an initial evaluation, and the process is automatically calibrated based on detailed task data from the data collection module (such as code volume statistics and dependency graphs); when changes in requirements lead to adjustments in the task scope, the task is re-evaluated. .

[0055] Step S3.2, Calculation of capability matching degree: Using cosine similarity algorithm to calculate personnel With the task Basic match degree: ;in For vector dot product, Let the vector magnitude be , The closer to 1, the higher the ability matching degree; Step S3.3, Dynamic Adaptation Algorithm: Construct the adaptation objective function: The constraints are: ,in For personnel Maximum workload capacity (based on historical data).

[0056] Calculate the overall fit score ,in These are the weighting coefficients, and The initial values ​​are 0.5, 0.3, and 0.2, respectively, and can be dynamically adjusted according to the project stage; according to Sort in descending order for each task Assign the person with the highest overall score If the workload exceeds the constraints, a secondary allocation (adjusting the resource allocation for low-priority tasks) will be triggered. The secondary allocation mechanism is a fault-tolerant adjustment mechanism for the multi-dimensional data-driven resource-task dynamic adaptation module. Its core purpose is to address the issue of some personnel exceeding their workload constraints after the initial allocation. This is achieved by optimizing the resource allocation of low-priority tasks to ensure a balanced overall resource load. The specific steps are as follows: Overload detection: After the initial allocation, a certain person is detected. The load satisfies (Exceeding the maximum capacity) Immediately lock the low-priority tasks under that person's name (P2-P3 level, divided according to task priority).

[0057] Task filtering: From the locked low-priority tasks, prioritize those with high overall suitability scores. The lowest-cost tasks (tasks with poor adaptability have lower migration costs) are excluded, while tasks that are in critical dependency links and cannot be interrupted are also excluded.

[0058] Resource matching: Extract the current load rate from the central control module. For idle / low-load personnel, calculate their overall fit score with the selected low-priority tasks. The person with the highest score and sufficient workload redundancy will be selected as the new person in charge.

[0059] Assignment Update: Officially transfer the task to the new person in charge and update the load rates of both parties simultaneously. The system associates tasks with data, generates adjustment records (including the reason for the adjustment, the original person in charge, the new person in charge, and the task ID), and feeds them back to the central control module to ensure that the subsequent dynamic adaptation algorithm is based on the latest data.

[0060] Step S3.4, Dynamic Adjustment Mechanism: Recalculate every 30 minutes based on the latest progress data and personnel load data. Fine-tuning of resource allocation; when changes in demand lead to When needed, a full-scale adaptation and optimization will be triggered immediately to ensure that resources and tasks are always matched in the best possible way. The full-scale adaptation optimization is an emergency adjustment mechanism for the resource-task dynamic adaptation module driven by multi-dimensional data. Its core is to re-examine all tasks and resources to ensure optimal matching when requirements change or unexpected situations occur. The specific steps are as follows: Trigger confirmation and data synchronization: Detected In the event of unforeseen circumstances such as staff leave or task blockage, immediately request the latest full data from the central control module, including updated information on all unfinished tasks (workload, requirement vector, urgency), real-time staff status (capability vector, load factor), and quantitative results of requirement changes, to ensure complete and synchronized data.

[0061] Full object review: Filter all incomplete tasks in the system (including newly added, unassigned, and assigned but incomplete tasks), and exclude tasks that have entered the final stage and cannot be adjusted; simultaneously extract all available personnel information and mark their current load redundancy and skill adaptation direction.

[0062] The fit score is recalculated: the comprehensive fit score formula is used. Substitute the latest parameters (such as the changed task requirement vector and the updated personnel load rate) and recalculate the adaptation score for each "personnel-task" pair.

[0063] Full optimal allocation: by The assignment is sorted in descending order, following the principles of prioritizing high-scoring personnel, ensuring that the workload does not exceed limits, and prioritizing urgent tasks. The optimal personnel are assigned to each unfinished task. For combinations with a matching score lower than 0.6 in the original allocation scheme, they are forcibly adjusted to a better match.

[0064] Results synchronization and effectiveness: Generate a full resource allocation update list, synchronize it to the central control module, relevant personnel terminals and progress quality early warning module, update core data such as task leaders and personnel load rates, and ensure that subsequent work is carried out based on the latest and optimal allocation.

[0065] Step S4, Schedule and Quality Early Warning: Based on the quantified impact data in Step S2 and the resource adaptation results in Step S3, the central control module monitors project schedule and quality risks in real time and triggers tiered early warnings; the specific implementation steps are as follows: Step S4.1, Definition and Calculation of Early Warning Indicators: Extract input data from the central control module, including the following: Schedule early warning indicator: Schedule deviation rate The planned schedule is calculated based on the quantified workload and resource matching results, i.e., the total workload. , Milestone Node Delay Risk Value , High risk; Quality warning indicator: Code defect rate (Statistics based on the number of defects per thousand lines of code) Quality anomalies; timely defect repair rate , To fix the lag; test case pass rate , The test failed to meet the standards; Step S4.2, Tiered Early Warning Thresholds and Triggering Conditions: Progress warning levels: or Level 1 warning (severe delay). or Level 2 warning (general delay). or Level 3 warning (slight delay); Quality early warning classification: or or Level 1 warning (serious quality problem). or or Level 2 warning (general quality issue); Triggering mechanism: Real-time monitoring of early warning indicators. An early warning is triggered immediately when an indicator reaches a threshold. At the same time, the early warning priority is adjusted based on the impact level of the requirement change (major / moderate / minor) (e.g., major impact + level 1 progress warning, with the highest priority). Early warning response and coordination: Real-time monitoring of early warning indicators; triggering an early warning immediately when an indicator reaches a threshold; and adjusting the early warning priority based on the impact level of changes in demand. Early warning response and coordination: Early warning information push: including early warning level, involved task ID, specific indicator value, and risk cause analysis (e.g., progress warning reason: caused by requirement change). If the resource allocation is not adjusted in a timely manner, suggestions for handling the situation (such as a level 1 progress warning suggestion: add 1-2 skilled personnel to prioritize core tasks) should be pushed to the project manager, resource administrator, and relevant task managers.

[0066] System linkage: After the early warning is triggered, the central control module automatically calls the resource-task dynamic adaptation module to generate a resource adjustment plan based on the handling suggestions (such as temporarily allocating idle resources and adjusting the progress of low-priority tasks), and at the same time updates the progress plan, forming a closed loop of early warning-analysis-handling-optimization.

[0067] Step S4.3, Early Warning Information Push: Includes early warning level, involved task ID, specific indicator value, and risk cause analysis (e.g., progress warning reason: caused by requirement change). If the resource matching is not adjusted in time, suggestions for handling (such as the first-level progress warning suggestion: add 1-2 skilled personnel to prioritize core tasks) are pushed to the project leader, resource administrator and relevant task leaders.

[0068] System linkage: After the early warning is triggered, the central control module automatically calls the resource-task dynamic adaptation module to generate a resource adjustment plan based on the handling suggestions (such as temporarily allocating idle resources and adjusting the progress of low-priority tasks), and at the same time updates the progress plan, forming a closed loop of early warning-analysis-handling-optimization.

[0069] Step S5, Results Archiving: Classify and archive the data development results and management data within the central control module to form a reusable knowledge base. It should be noted that this step utilizes existing technology; a brief explanation of the implementation steps is provided below: Data classification and filtering: Extract full-process data from the central control module and classify it into "development deliverables" and "management data". Development deliverables include code files, requirement documents, design schemes, test cases, etc., associated with corresponding requirement IDs and task IDs; management data includes requirement change records, resource allocation plans, progress and quality early warning and handling results, personnel performance data, etc., to ensure that the data covers the entire project lifecycle.

[0070] Standardization processing: Standardize the format of categorized data (e.g., name code files by version number and organize documents according to fixed templates), add multi-dimensional tags (e.g., project type, business module, task type, skill tags) to facilitate accurate retrieval later, and remove duplicate and invalid data to ensure the quality of the knowledge base.

[0071] Experience extraction and reuse: Extract high-value content from archived data, such as cases of high matching of resources and tasks, optimal solutions for quantifying the impact of requirement changes, and methods for rapid handling of quality issues, and form standardized experience items, which are then stored in association with the original data.

[0072] Knowledge base construction and updates: Processed results and experience entries are entered into the knowledge base, a search index is established, and queries by tag combination are supported; after the completion of subsequent projects, archived data and experience entries are updated synchronously to realize dynamic iteration of the knowledge base and continuously empower the management of similar projects.

[0073] Example 2 Please see Figure 5 The software development system management platform based on big data includes a project data full-dimensional collection module, a requirement change-impact dynamic quantification module, a multi-dimensional data-driven resource-task dynamic adaptation module, a central control module, a progress and quality early warning module, and a deliverable archiving module. The project data full-dimensional collection module collects multi-dimensional data from the entire software development process, covering requirement data, task data, personnel data, progress data, and quality data. It connects to development management tools and personnel terminals through interfaces, and sends the data to the central control module after cleaning and standardization. The Requirement Change - Impact Link Dynamic Quantification Module receives the collected data after processing by the central control module, and then traces the full-link impact of the requirement change by constructing a requirement-task-resource association graph, quantifies the impact coefficient and workload increment, outputs an accurate impact assessment report, and inputs the impact data assessment report into the central control module. The multi-dimensional data-driven resource-task dynamic adaptation module receives resource adaptation results and task data from the central control module. By constructing personnel capability vectors and task requirement vectors, it calculates a comprehensive adaptation score to achieve optimal allocation of resources and tasks. It also has a secondary allocation fault tolerance mechanism to solve resource overload problems and ensure balanced resource load. Its allocation results are synchronized to the central control module and the progress quality early warning module. The central control module receives data collected by the project data full-dimensional acquisition module and forwards it to the requirement change-impact link dynamic quantification module and the multi-dimensional data-driven resource-task dynamic adaptation module. Receive the impact assessment report from the dynamic quantification module of the impact of requirement changes and the resource allocation scheme of the resource-task dynamic adaptation module driven by multi-dimensional data, integrate and generate execution instructions, link the progress and quality early warning module to push early warning information, and synchronously update the data of the entire system to ensure that all modules work together. The progress and quality early warning module calculates core indicators such as progress deviation rate and defect rate based on the quantitative impact data, resource adaptation results and real-time progress and quality data forwarded by the central control module. It triggers graded early warnings according to thresholds, and at the same time links the central control module and the adaptation module to push disposal suggestions and trigger dynamic resource adjustments, so as to avoid project delays and quality risks in advance and ensure that the project progresses as planned. The results archiving module receives full-process data (development results and management data) transmitted from the central control module. After classification, screening, standardization, and extraction of reuse experience, a tagged knowledge base is built. This not only completes the systematic archiving of project data, but also provides reusable experience such as requirement change handling cases and resource allocation schemes for subsequent similar projects, thereby achieving iterative optimization of system management capabilities.

[0074] This invention discloses a software development system management method based on big data, belonging to the field of software development management technology. It includes a project data full-dimensional collection module, a requirement change-impact link dynamic quantification module, a multi-dimensional data-driven resource-task dynamic adaptation module, a progress and quality early warning module, a central control module, and a deliverable archiving module. The core innovation lies in: the requirement change-impact link dynamic quantification module quantifies the impact of requirement changes on tasks, resources, and schedule by constructing a full-link correlation graph; the multi-dimensional data-driven resource-task dynamic adaptation module achieves precise dynamic matching of resources and tasks based on personnel capability profiles, task complexity, and real-time project data. This invention solves the problems of low development efficiency and high rework rates caused by the ambiguity of requirement change impacts and unreasonable resource allocation in existing methods. Through the collaboration of the requirement change-impact link dynamic quantification module and the multi-dimensional data-driven resource-task dynamic adaptation module, it improves the intelligence and accuracy of software development management, ensuring efficient project progress.

[0075] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A software development system management method based on big data, characterized in that, Includes the following steps: Step S1: Comprehensive Project Data Collection: Collect requirement data, task data, personnel data, progress data, and quality data during the software development process to obtain the collected data, and then transmit the collected data to the central control module. Step S2, Dynamic Quantification of Impact Links of Requirement Changes: Based on the collected data, construct a requirement-task-resource relationship graph, quantify the impact scope and workload increment of requirement changes, and output an impact data assessment report to the central control module; Step S3, Multi-dimensional data-driven dynamic resource-task adaptation: Combining personnel capability profiles, task complexity coefficients, and real-time project progress from the collected data, resources are allocated and task priorities are adjusted through dynamic adaptation algorithms, and the resource adaptation results are output to the central control module. Step S4, Progress and Quality Early Warning: Based on the quantitative impact data in Step S2 and the resource adaptation results in Step S3, the central control module monitors project progress and quality risks in real time and triggers tiered early warnings. Step S5, Results Archiving: Classify and archive the data development results and management data within the central control module to form a reusable knowledge base.

2. The software development system management method based on big data according to claim 1, characterized in that: The specific contents of the requirement data, task data, personnel data, progress data, and quality data collected by the data acquisition software in step S1 include: Requirements data includes requirement ID, requirement name, function description, priority P0-P3, business module affiliation, proposer, proposal time, and planned delivery node, ensuring coverage of key information throughout the entire requirement lifecycle; Task data includes task ID, task type, functional module, complexity level, estimated workload, responsible person ID, start time, planned completion time, dependent task IDs, and current status. Personnel data includes personnel ID, name, position, skill tags, skill proficiency, historical project experience, maximum workload, current assigned workload, and workload rate. Progress data includes task completion rate, milestone progress, task delay duration, and cross-module collaboration progress synchronization data, updated every 5 minutes. Quality data includes code commit frequency, code defect rate, test pass rate, defect fix time, code review pass rate, and documentation completeness score. The data acquisition software adopts a combination of real-time acquisition and timed synchronization. It connects to development tools (code repositories, project management tools) and personnel terminals through API interfaces. The data update cycle is ≤5 minutes. After data acquisition, it is cleaned and deduplicated before being transmitted to the central control module.

3. The software development system management method based on big data according to claim 2, characterized in that: The specific implementation steps for the requirement change impacting dynamic link quantification in step S2 are as follows: Step S2.1, Construction of the association graph: Define the core parameter: Requirement ID is Task ID is Resource ID is The direct relationship between requirements and tasks is as follows: The allocation relationship between tasks and resources is as follows: Based on the collected dependency data, a three-layer relationship graph of requirements, tasks, and resources is constructed, and nodes are stored using a graph database. ( ) and edges, and label the relationship strength weights. ,in express and The relationship strength weights, ranging from 0 to 1, are obtained through training on historical data. Step S2.2, Impact on Link Tracing: When demand When changes occur, a depth-first search algorithm is used to traverse the association graph and trace directly related tasks. And indirectly related tasks (downstream dependent tasks derived from directly related tasks). To form a complete influence chain ; Synchronous traceability of resources involved in the link Record the current workload of resources and the status of task associations; Step S2.3, Calculation of Influence Coefficient: Define the influence coefficient. For demand Changes to the task The degree of influence, calculated using the following formula: ,in Adjustment factor for task complexity For the task The complexity level is 1-5. big The larger, Define the total influence range coefficient. ,in For the task Project weight, The closer it is to 1, the wider the scope of influence; Step S2.4, Workload Increment Assessment: Construct an impact coefficient-workload increment mapping model based on historical data, and input... With the task Original workload Calculate the workload increment for a single task ,in Historical correction factor; total workload increment It also outputs impact reports and workload adjustment suggestions, which are then transmitted to the central control module.

4. The software development system management method based on big data according to claim 1, characterized in that: The specific implementation steps for workload increment assessment in step S2.4 are as follows: Step S2.4.1, Construction and Validation of the Influence Coefficient-Workload Increment Mapping Model: Model data source: Historical data from several similar software development projects were collected, and valid samples were selected. Valid samples include, but are not limited to, complete requirement change records and task impact coefficients. Original workload Actual workload increase Task types and project scales should be considered to ensure that effective samples cover different business scenarios and task types. Feature variable definition: based on influence coefficient The core independent variable is the original workload of the task. Task type is an auxiliary independent variable, and the actual workload increment is... A multivariate mapping model is constructed with the task type as the dependent variable. This indicates that Development = 1, Testing = 2, Design = 3, and Documentation = 4. Model Fitting and Calibration: The gradient boosting regression algorithm is used for model training, and the model parameters are optimized through 5-fold cross-validation to ensure that the goodness of fit of the multivariate mapping model is R²≥0.95 and the prediction error is ≤7%. After training, the multivariate mapping model is stored locally in the module, supporting real-time access and subsequent iterative updates. Step S2.4.2: Preprocessing and validity verification of input parameters: Influence coefficient Verification: Extract the individual task influence coefficients calculated in step S2.

3. Verify whether it is within the valid range. Internally, ensure the reliability of the input coefficients; Original workload Confirmed: Task obtained from the project data full-dimensional collection module. Original estimated workload Synchronous verification To ensure the rationality of the calculation and avoid deviations in incremental calculations caused by errors in the original data; Historical correction coefficient Matching: Based on task type Match the corresponding value, Based on historical data from similar projects, the value range is clearly defined: Development tasks Test tasks Design tasks Document-related tasks Ensure coefficient Adapted to task characteristics; Step S3.4.3, Increment of workload for a single task Precise calculations: Formula substitution calculation: Substitute the preprocessed... Substitute into the core calculation formula The calculation is performed by the hardware-accelerated computing unit built into the demand change-impact link dynamic quantification module, with a single calculation time of ≤5ms, ensuring real-time performance; Calculation result verification: Perform a reasonableness check and set threshold rules: If If this happens, a secondary verification will be automatically triggered to re-check the data. The calculation process (such as association weight) Complexity correction factor (Is it accurate?) For man-days, a uniform calculation of 0.1 man-days will be used to ensure that the results match the actual execution scenario; Result labeling: for each task Add tags, including the type of impact and the basis for calculation ( value, (Value), which facilitates subsequent traceability and review; Step S2.4.4, Total Workload Increment Summary and dimensional breakdown: Total increment calculation: effective for all affected tasks Summing gives the total increase in workload due to requirement changes. Simultaneously calculate the proportion of the total increment to the remaining workload of the project. Used for subsequent progress risk prediction; Multi-dimensional breakdown: Breaking down tasks by type, priority, and business module. Perform categorization and summarization to generate dimensional incremental data; Step S2.4.5: Generation of Impact Link Report and Workload Adjustment Suggestions: Key content of the report: Visualized impact chain diagram: Marking nodes of requirement changes, affected task nodes, and related resource nodes, intuitively presenting the impact transmission path; List of affected task workload increments: including but not limited to task ID, task name, value, , value, Incremental percentage Task priority field; Schedule risk prediction: based on and Predict the overall project delay time; Workload adjustment suggestions: Combined with Size provides differentiated suggestions: It is recommended to extend the task duration by 3-5 days and prioritize the allocation of core resources; It is recommended to extend the construction period by 1-3 days; It is recommended to maintain the original construction period and absorb the incremental workload by optimizing the workflow; Data transmission: the calculated results Dimensional incremental data and complete impact chain reports are transmitted in real time to the central control module via an encrypted protocol, and simultaneously synchronized to the project management terminal, triggering subsequent resource-task dynamic adaptation processes and schedule adjustment decisions.

5. The software development system management method based on big data according to claim 1, characterized in that: The specific implementation logic of the multi-dimensional data-driven dynamic resource-task adaptation in step S3 is as follows: Step S3.1, Core Parameter Quantization and Dynamic Update: Personnel Capability Profile Construction: Personnel Capability Vector , For the number of skill types, The quantification method is: skills assessment score + historical similar task completion quality score + collaboration evaluation score, which is recalculated every two weeks. To ensure that capability profiles are accurate in real time; Task requirement vector construction: Task requirement vector , The quantification method is as follows: it is calculated based on a comprehensive consideration of task type, complexity level, and business priority. It is automatically generated when the task is created and updated synchronously when requirements change. Auxiliary parameter calculation: personnel load rate ; Task urgency , For an urgent mission, For low-urgency tasks; task complexity The evaluation is based on a comprehensive assessment of code size, number of dependencies, and technical difficulty. Core parameter definition and quantization: Personnel Capability Vector: Defining Personnel The capability vector is ,in For the number of skill types, for In skills Proficiency level (value 0-10, derived from a combination of historical performance and skills assessment); Task requirement vector: Defines the task The demand vector is ,in For the task skills Demand level; Auxiliary parameter: Current staff load rate , Values ​​range from 0 to 1; Task urgency (Based on project schedule deviation settings, value ranges from 0 to 1), Task complexity (Values ​​range from 1 to 5); Step S3.2, Calculation of capability matching degree: Using cosine similarity algorithm to calculate personnel With the task Basic match degree: ;in For vector dot product, Let the vector magnitude be , The closer to 1, the higher the ability matching degree; Step S3.3, Dynamic Adaptation Algorithm: Construct the adaptation objective function: The constraints are: ,in For personnel Maximum workload capacity (based on historical data). Calculate the overall fit score ,in These are the weighting coefficients, and The initial values ​​were 0.5, 0.3, and 0.2 respectively; according to Sort in descending order for each task Assign the person with the highest overall score If the personnel load exceeds the constraints, a secondary allocation will be triggered; Step S3.4, Dynamic Adjustment Mechanism: Recalculate every 30 minutes based on the latest progress data and personnel load data. Fine-tuning of resource allocation; When demand changes lead to When needed, a full-scale adaptation optimization is immediately triggered to ensure that resources and tasks are always optimally matched.

6. The software development system management method based on big data according to claim 1, characterized in that: The specific implementation steps of step S4 are as follows: Step S4.1, Definition and Calculation of Early Warning Indicators: Extract input data from the central control module, including the following: Schedule early warning indicator: Schedule deviation rate The planned schedule is calculated based on the quantified workload and resource matching results, i.e., the total workload. , Milestone Node Delay Risk Value , High risk; Quality warning indicator: Code defect rate , This is a quality anomaly. Defect repair timeliness , To fix the lag; test case pass rate , The test failed to meet the standards; Step S4.2, Tiered Early Warning Thresholds and Triggering Conditions: Progress warning levels: or Level 1 warning or It is a level 2 warning. or It is a Level 3 warning; Quality early warning classification: or or Level 1 warning or or Level II warning; Triggering mechanism: Real-time monitoring of early warning indicators; an early warning is triggered immediately when an indicator reaches a threshold; and the early warning priority is adjusted based on the impact level of changes in demand. Early warning response and coordination: Early warning information push: including early warning level, involved task ID, specific indicator value, risk cause analysis, and handling suggestions, is pushed to project leader, resource administrator and relevant task leader. System linkage: After the early warning is triggered, the central control module automatically calls the resource-task dynamic adaptation module to generate a resource adjustment plan based on the handling suggestions, and at the same time update the progress plan, forming a closed loop of early warning-analysis-handling-optimization; Step S4.3, Early Warning Information Push: This includes the early warning level, the ID of the task involved, the specific value of the indicator, the risk cause analysis, and the handling suggestions, and is pushed to the project leader, resource administrator and relevant task leaders. System linkage: After the early warning is triggered, the central control module automatically calls the resource-task dynamic adaptation module to generate a resource adjustment plan based on the handling suggestions, and at the same time update the progress plan.

7. The software development system management platform based on big data according to any one of claims 1-6, characterized in that: It includes a project data full-dimensional collection module, a requirement change-impact dynamic quantification module, a multi-dimensional data-driven resource-task dynamic adaptation module, a central control module, a progress and quality early warning module, and a deliverable archiving module. The project data full-dimensional collection module collects multi-dimensional data from the entire software development process, covering requirement data, task data, personnel data, progress data, and quality data. It connects to development management tools and personnel terminals through interfaces, and sends the data to the central control module after cleaning and standardization. The dynamic quantification module for the demand change-impact link receives the collected data after processing by the central control module, and then traces the full-link impact of the demand change by constructing a demand-task-resource association graph, quantifies the impact coefficient and workload increment, outputs an accurate impact assessment report, and inputs the impact data assessment report into the central control module. The multi-dimensional data-driven resource-task dynamic adaptation module receives resource adaptation results and task data from the central control module. By constructing personnel capability vectors and task requirement vectors, it calculates a comprehensive adaptation score to achieve optimal allocation of resources and tasks. The allocation results are synchronized to the central control module and the progress and quality early warning module. The central control module receives data collected by the project data full-dimensional collection module and forwards it to the requirement change-impact link dynamic quantification module and the multi-dimensional data-driven resource-task dynamic adaptation module. Receive the impact assessment report from the dynamic quantification module of the impact of requirement changes and the resource allocation scheme of the resource-task dynamic adaptation module driven by multi-dimensional data, integrate and generate execution instructions, link the progress and quality early warning module to push early warning information, and synchronously update the data of the entire system to ensure that all modules work together. The progress and quality early warning module triggers tiered early warnings based on the quantitative impact data, resource adaptation results and core indicators forwarded by the central control module. It also links the central control module and the adaptation module to push disposal suggestions and trigger dynamic resource adjustments, thereby avoiding project delays and quality risks in advance and ensuring that the project progresses as planned. The results archiving module receives full-process data transmitted from the central control module, and constructs a tagged knowledge base through classification, screening, standardization, and experience extraction for reuse.

Images