A big data-based software development system management method and system

Abstract

The application provides a software development system management method and system based on big data. The code review data set in remote collaborative software development is collected, and a review knowledge graph is constructed based on the space-time correlation characteristics of the data set. The graph nodes include review behavior patterns associated with geographical and time characteristics. The syntax structure features of the current code changes are extracted synchronously, and they are jointly encoded with the historical review tracks in the knowledge graph to generate processing instructions containing version difference analysis priority. By dynamically matching the processing instructions with the domain knowledge coverage range, the allocation strategy of cross-regional review tasks is determined. Finally, according to the mapping relationship of the allocation strategy, the version priority and the quality evaluation node, the review management control strategy suitable for the software development system is generated. The technical scheme provided by the application can improve the efficiency and accuracy of software development system management.

Description

Technical Field

[0001] This application relates to the field of software development system management technology, and in particular to a software development system management method and system based on big data. Background Technology

[0002] In remote collaborative software development, cross-regional teams need to efficiently handle massive amounts of code review tasks. However, due to differences in time and space and dynamic changes in review resources (such as personnel availability and differences in domain knowledge coverage), traditional centralized manual review methods are no longer sufficient to meet the requirements of efficient, accurate, and scalable review. Therefore, the industry urgently needs a systematic management method that can intelligently analyze and make decisions based on project history data, developer behavior characteristics, and code change patterns to improve the automation level and management efficiency of code reviews.

[0003] Currently, a targeted solution to this technical need is to adopt an automated task distribution system based on modular tags. This system matches tasks with reviewers using predefined rules (such as the functional module to which the code change belongs and the reviewer's historical participation record). For example, the system categorizes code changes by module and assigns tasks based on the module tags that reviewers have previously participated in, while also setting priorities using simple time thresholds. This method relies on a static rule base and tag matching algorithm, partially alleviating the inefficiency of manual assignment.

[0004] While the aforementioned solutions achieve initial automation to some extent, the rule base and tagging system struggle to cover dynamic spatiotemporal correlations (such as sudden defect repair needs and fluctuations in reviewer availability across time zones), leading to rigid task allocation. For example, module tags fail to reflect the cross-module coupling of code changes, and static time thresholds ignore the fluctuating characteristics of defect repair cycles, easily causing high-priority tasks to backlog or inefficient resource consumption. Furthermore, the lack of joint analysis of review behavior patterns and version difference parsing results in a disconnect between allocation strategies and real-time development needs, preventing global optimization. Summary of the Invention

[0005] This application provides a software development system management method and system based on big data, which solves the problems of low efficiency and poor accuracy in the management of existing software development systems.

[0006] Firstly, this application provides a software development system management method based on big data, including:

[0007] Obtain the code review dataset generated during the remote collaborative software development process. The code review dataset includes the geographical distribution density of review comments, the time fluctuation characteristics of defect repair cycles, and the domain knowledge coverage of reviewers.

[0008] Based on the spatiotemporal correlation characteristics of the code review dataset, a review knowledge graph with the software development stage as the evolution axis is constructed. The review knowledge graph contains review behavior patterns associated with the geographical distribution density and temporal fluctuation characteristics.

[0009] Synchronously collect the syntactic structure features of the current software code change record, and jointly encode the syntactic structure features with the historical review trajectory in the review knowledge graph to generate processing instructions that include the priority of version difference parsing;

[0010] The processing instructions are dynamically matched with the domain knowledge coverage to determine the allocation strategy for cross-regional review tasks. The allocation strategy includes quality assessment nodes that are path-related to the review behavior pattern.

[0011] Based on the allocation strategy, version difference resolution priority, and the mapping relationship between quality assessment nodes, a review management and control strategy adapted to the software development system is generated.

[0012] Optionally, the step of generating a review management and control strategy adapted to the software development system based on the allocation strategy, version difference resolution priority, and the mapping relationship between quality assessment nodes includes:

[0013] Based on the geographical association markers of the cross-regional review tasks in the allocation strategy, the path weights of the quality assessment nodes are extracted.

[0014] The version difference parsing priority is divided into continuous intervals according to the matching depth of the syntax structure features, and the intervals are cross-combined with the path weights to generate a sequence of control parameters indexed by the level boundary value.

[0015] Based on the distribution characteristics of the hierarchical boundary values, the geographic association markers are dynamically corrected. The corrected geographic association markers form a spatial constraint relationship with the temporal fluctuation characteristics of the defect repair cycle. The spatial constraint relationship is then mapped to the hierarchical interval of the control parameter sequence to obtain the control rule set.

[0016] The set of control rules is embedded into the collaboration interface of the software development system. The domain knowledge coverage of the reviewers is traversed through the collaboration interface. Based on the traversal results, the path weights, hierarchical boundary values ​​and spatial constraint relationships are merged into multi-dimensional constraint boundaries, and the review management control strategy is output.

[0017] Optionally, mapping the spatial constraint relationship to the hierarchical interval of the control parameter sequence to obtain the control rule set includes:

[0018] Extract the starting and ending boundary values ​​of each level interval of the control parameter sequence, determine the level span value by measuring the distance between the starting and ending boundary values, and superimpose the level span value with the time fluctuation characteristics of the defect repair cycle in the spatial constraint relationship according to a preset ratio to generate the resource allocation base of each level interval.

[0019] The path weights of the quality assessment nodes are associated with the starting boundary values ​​of the corresponding hierarchical intervals. Hierarchical intervals whose association results exceed the preset activation values ​​are filtered out and summarized to obtain the task triggering condition set.

[0020] The resource allocation base is adjusted according to the preset resource allocation ratio to determine the resource allocation threshold of the hierarchical interval. The hierarchical interval of the task triggering condition set is bound to the resource allocation threshold to obtain the control rule set.

[0021] Optionally, the step of dividing the version difference parsing priority hierarchy into continuous intervals according to the matching depth of syntactic structural features, and cross-combining the intervals with path weights to generate a sequence of control parameters indexed by hierarchical boundary values, includes:

[0022] Based on the numerical range of the matching depth of the syntax structure features of the version difference parsing priority, the levels of the version difference parsing priority are arranged in ascending order, and the numerical range is divided into continuous intervals according to a preset segmentation density. The starting and ending values ​​of the intervals are defined as level boundary values.

[0023] Traverse the code change records within the interval, calculate the correlation value between the code change record and the path weight, combine the correlation value with the hierarchical boundary value according to a preset ratio, encode the combination result in ascending hierarchical order, and generate a sequence of control parameters indexed by the hierarchical boundary value.

[0024] Optionally, the synchronous collection of syntactic structure features of the current software code change record, and the joint encoding of the syntactic structure features with the historical review trajectory in the review knowledge graph, generates processing instructions containing version difference parsing priorities, including:

[0025] The syntax structure features of the current software code change record are captured in real time, the syntax structure features are decomposed into feature fragments, and the feature fragments are converted into a current feature vector set of uniform dimension.

[0026] Extract the grammatical structure feature vectors and review behavior patterns associated with the nodes of the historical review trajectory in the review knowledge graph to form a historical feature vector set;

[0027] The matching degree of the feature fragments in the current feature vector set and the corresponding grammatical structure feature vectors in the historical feature vector set is calculated. Feature fragments with matching degrees exceeding a preset threshold are filtered out to generate a feature matching mapping table.

[0028] Based on the matching feature fragments in the feature matching mapping table, the spatiotemporal correlation strength of the review behavior pattern is extracted, and the spatiotemporal correlation strength is superimposed with the syntactic structure complexity of the matching feature fragments to generate priority weight parameters for each matching feature fragment.

[0029] The priority weight parameter is proportionally fused with the version change coverage of the corresponding feature segment to generate a processing instruction that includes the priority of version difference parsing.

[0030] Optionally, the step of calculating the matching degree between the feature fragments of the current feature vector set and the corresponding grammatical structure feature vectors in the historical feature vector set, filtering the feature fragments with matching degrees exceeding a preset threshold, and generating a feature matching mapping table includes:

[0031] The feature segments in the current feature vector set are decomposed into independent dimension values ​​according to the dimension, and the corresponding grammatical structure feature vectors in the historical feature vector set are decomposed into historical dimension values ​​according to the same dimension.

[0032] Calculate the difference between the independent dimension value and the historical dimension value, generate a difference score for each dimension, and filter the number of dimensions with a difference score lower than a preset threshold in the feature fragments. When the number of dimensions exceeds the preset number, mark the corresponding feature fragment as a valid matching fragment.

[0033] Based on the identifier of the valid matching fragment, the identifier of the corresponding grammatical structure feature vector is matched, and a feature matching mapping table reflecting the mapping relationship between the feature fragment and the grammatical structure feature vector is generated.

[0034] Optionally, the step of constructing a review knowledge graph with software development stages as the evolution axis based on the spatiotemporal correlation characteristics of the code review dataset includes:

[0035] The geographical distribution density of review comments in the code review dataset is converted into a geographical density parameter, the time fluctuation characteristics of the defect repair cycle are converted into a time fluctuation parameter, and the domain knowledge coverage of reviewers is decomposed into a knowledge weight vector.

[0036] Using the time sequence of the software development phase as the evolution axis, the geographic density parameter and the time fluctuation parameter are combined according to a preset weight ratio to generate the review behavior pattern parameter;

[0037] The review behavior pattern parameters are mapped to independent nodes, and a spatiotemporal association path between nodes is established based on the difference in geographic density parameters between adjacent nodes and the correlation of time fluctuation parameters.

[0038] The knowledge weight vector is dynamically bound to the review behavior mode parameters to obtain the knowledge transfer links between nodes. By integrating the nodes, spatiotemporal association paths and knowledge transfer links, a review knowledge graph is generated.

[0039] Secondly, this application provides a software development system management system based on big data, including:

[0040] The acquisition module acquires the code review dataset generated during the remote collaborative software development process. The code review dataset includes the geographical distribution density of review comments, the time fluctuation characteristics of defect repair cycles, and the domain knowledge coverage of reviewers.

[0041] The construction module, based on the spatiotemporal correlation characteristics of the code review dataset, constructs a review knowledge graph with the software development stage as the evolution axis. The review knowledge graph contains review behavior patterns associated with the geographical distribution density and temporal fluctuation characteristics.

[0042] The encoding module synchronously collects the syntactic structure features of the current software code change record, and jointly encodes the syntactic structure features with the historical review trajectory in the review knowledge graph to generate processing instructions that include the priority of version difference parsing.

[0043] The matching module dynamically matches the processing instructions with the domain knowledge coverage to determine the allocation strategy for cross-regional review tasks. The allocation strategy includes quality assessment nodes that are path-related to the review behavior pattern.

[0044] The generation module generates a review management and control strategy adapted to the software development system based on the allocation strategy, version difference parsing priority, and the mapping relationship between quality assessment nodes.

[0045] Thirdly, embodiments of this application provide a computing device, including a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement a software development system management method based on big data as described in the first aspect above.

[0046] Fourthly, embodiments of this application provide a computer storage medium storing a computer program, which, when executed by a computer, implements a software development system management method based on big data as described in the first aspect.

[0047] In this embodiment, a code review dataset generated during remote collaborative software development is acquired. This dataset includes the geographical distribution density of review comments, the temporal fluctuation characteristics of defect repair cycles, and the domain knowledge coverage of reviewers. Based on the spatiotemporal correlation characteristics of the code review dataset, a review knowledge graph is constructed with the software development stage as the evolution axis. This review knowledge graph includes review behavior patterns associated with the geographical distribution density and temporal fluctuation characteristics. Syntactic structural features of current software code change records are simultaneously collected, and these features are jointly encoded with historical review trajectories in the review knowledge graph to generate processing instructions containing version difference parsing priorities. These processing instructions are dynamically matched with the domain knowledge coverage to determine a cross-regional review task allocation strategy. This allocation strategy includes quality assessment nodes that are path-related to the review behavior patterns. Based on the mapping relationship between the allocation strategy, version difference parsing priorities, and quality assessment nodes, a review management and control strategy adapted to the software development system is generated.

[0048] The technical solution of this application has the following beneficial effects:

[0049] This application integrates geographical distribution density, temporal fluctuations in defect repair cycles, and the domain knowledge coverage of reviewers to provide multi-dimensional data support for cross-regional task allocation, addressing the lack of spatiotemporal correlation features in traditional solutions. Based on spatiotemporal correlation characteristics, a knowledge graph with development stages as the evolutionary axis is established, mapping geographical distribution and temporal fluctuations into quantifiable and analyzable review behavior patterns, achieving global visualization and dynamic correlation of review logic. Through joint encoding of syntactic structure features and historical review trajectories, the priority of version difference parsing for code changes is transformed into executable instructions, enhancing the real-time performance and contextual relevance of version difference processing. Combined with dynamic matching of processing instructions and domain knowledge coverage, a cross-regional task allocation strategy including quality assessment nodes is generated, ensuring precise matching of task allocation with reviewer capabilities and version requirements. Based on the mapping relationship between allocation strategies, version priorities, and quality assessment nodes, a review management and control strategy adapted to the development system is formed, achieving synergistic optimization of task allocation and review priorities.

[0050] Furthermore, based on the geographic association markers in the allocation strategy, path weights of quality assessment nodes are extracted. Version difference parsing priorities are divided into continuous intervals according to syntactic structure matching depth, and these are cross-combined with path weights to generate a sequence of control parameters indexed by hierarchical boundary values. Geographic association markers are dynamically corrected through the distribution characteristics of hierarchical boundary values, establishing a spatial constraint relationship with the temporal fluctuations of the defect repair cycle. This relationship is then mapped to the hierarchical intervals of the control parameter sequence, generating a set of control rules. Finally, the rule set is embedded into the system collaboration interface, traversing the knowledge coverage of reviewers, merging path weights, hierarchical boundary values, and spatial constraint relationships into multidimensional constraint boundaries, and outputting the control strategy. Through the dynamic correction of geographic association markers and the cross-mapping of hierarchical boundary values, multidimensional collaborative control of version priority, spatiotemporal constraints, and review resources is achieved, resolving spatiotemporal conflicts and priority fragmentation issues in cross-regional task allocation, and improving the dynamic adaptability and global optimization capability of the review strategy.

[0051] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart of a software development system management method based on big data provided in this application is shown;

[0054] Figure 2 A schematic diagram of the structure of a software development system management system based on big data provided in this application is shown;

[0055] Figure 3 A schematic diagram of the structure of a computing device provided in this application is shown. Detailed Implementation

[0056] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0057] In some of the processes described in the specification, claims, and accompanying drawings of this application, multiple operations appearing in a specific order are included. However, it should be clearly understood that these operations may not be executed in the order they appear herein, or may be executed in parallel. The operation numbers, such as 101, 102, etc., are merely used to distinguish different operations and do not themselves represent any execution order. Furthermore, these processes may include more or fewer operations, and these operations may be executed sequentially or in parallel. It should be noted that the descriptions such as "first," "second," etc., in this document are used to distinguish different messages, devices, modules, etc., and do not represent a chronological order, nor do they limit "first" and "second" to different types.

[0058] Researchers have found that in existing remote collaborative software development, the allocation of cross-regional code review tasks often suffers from rigid task allocation and delayed version priority adjustments due to differences in geographical distribution, fluctuations in defect repair cycles, and mismatches in the knowledge coverage of reviewers. Based on this, a big data-driven software development system management method is proposed. This method can construct a spatiotemporally correlated review knowledge graph by collecting geographical distribution density, temporal fluctuation characteristics, and domain knowledge coverage from code review datasets. It then co-encodes the syntactic structure features of real-time code changes with historical review trajectories to generate version difference parsing priorities. By dynamically matching processing instructions with domain knowledge coverage, a cross-regional task allocation strategy and quality assessment node mapping relationship are formed. Finally, a review management and control strategy adapted to the development system is generated, achieving precise allocation of cross-regional tasks and dynamic collaborative adjustment of version priorities, thereby improving collaboration efficiency and review quality.

[0059] The entire research process optimizes code review management methods in remote collaborative software development, integrates multi-source heterogeneous data, and constructs a dynamically evolving knowledge graph to capture the spatiotemporal correlation between development phases and review behaviors. Based on this, it combines the syntactic features of real-time code changes with historical review patterns, identifies high-risk change points through priority coding, and achieves precise allocation of review tasks based on domain knowledge matching. Finally, through closed-loop feedback of strategy mapping and quality assessment nodes, an adaptive review control mechanism is formed, thereby solving the inefficiency problems caused by geographical dispersion, knowledge barriers, and resource mismatch in remote teams, and improving the intelligence level and collaborative efficiency of code review.

[0060] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0061] Figure 1 A flowchart illustrating a software development system management method based on big data is provided in this application embodiment, such as... Figure 1 As shown, the method includes:

[0062] 101. Obtain the code review dataset generated during the remote collaborative software development process. The code review dataset includes the geographical distribution density of review comments, the time fluctuation characteristics of defect repair cycles, and the domain knowledge coverage of reviewers.

[0063] In this step, remote collaboration software is a collection of tools that supports real-time communication, collaborative work, and project management for teams across regions and time zones, aiming to improve the work efficiency and collaboration quality of distributed teams through digital means.

[0064] Code review datasets are multi-dimensional data collections gathered during the software development process through version control systems and project management tools, used to quantitatively analyze code review behavior and its quality characteristics.

[0065] Geographic distribution density indicates the density of code review comments across different geographical regions, such as the percentage of review comments triggered by code changes submitted in a particular region.

[0066] The time fluctuation characteristics of defect repair cycle refer to the dynamic changes in the time required for a defect to be repaired from discovery, such as delayed repair during peak periods and rapid response during trough periods.

[0067] Domain knowledge coverage refers to the range of reviewers' review experience and knowledge accumulation in a specific technical field.

[0068] In this embodiment, the original code review dataset is first collected through the log system of a remote collaborative software development platform. Then, the submission time of review comments, the geographical location of reviewers, the timestamp of defect repair, and historical project tags in the code review dataset are analyzed. Finally, the geographical distribution density is extracted by statistically analyzing the proportion of review comments by region; the mean and variance of repair cycles are calculated by time window, and the time fluctuation characteristics of defect repair timestamps are calculated based on the mean and variance; a knowledge tag vector is constructed based on historical review records, and the domain knowledge coverage is determined based on the knowledge tag vector.

[0069] In a cross-regional collaborative e-commerce platform development project, the system automatically collected review data from multiple development nodes. For example, a code submission for a promotional function module triggered a cross-regional review process. The system records revealed that reviewers in one geographical region, due to differences in their working hours, concentrated their review comments within a specific timeframe, while reviewers in another region, involved in a wider range of module types, provided comments covering multiple technical areas. Defect repair cycle analysis showed that repair time significantly increased as the promotional activity approached launch, with delays in defect repairs related to high-concurrency modules being particularly pronounced. Historical data from reviewers indicated that some personnel had high engagement in reviewing the payment system module, while others focused on the user behavior analysis module, creating differentiated knowledge coverage tags.

[0070] 102. Based on the spatiotemporal correlation characteristics of the code review dataset, construct a review knowledge graph with the software development stage as the evolution axis. The review knowledge graph contains review behavior patterns associated with the geographical distribution density and temporal fluctuation characteristics.

[0071] In this step, spatiotemporal correlation characteristics refer to the dynamic relationship between geographical location and time dimension implicit in the data. For example, a delay in the review of a certain area may cause a blockage in subsequent development stages.

[0072] The evolution axis refers to the dynamic change path of review behavior patterns, with the software development stage as the main timeline.

[0073] Censorship knowledge graph is a semantic network model based on graph structure, specifically designed to organize, associate, and analyze multi-source heterogeneous data generated during the censorship process. It assists in intelligent decision-making by revealing the correlation patterns between censorship behavior, resource allocation, and quality assessment.

[0074] Review behavior patterns refer to typical review operation rules formed by the combination of geographical distribution density and time fluctuation characteristics, such as concentrated review during peak periods and automatic delay in low-density areas.

[0075] In this embodiment, firstly, based on the code submission timestamps and version evolution, the entire software development process is divided into multiple logical stages, such as requirements analysis, design implementation, testing and verification, and maintenance iteration. Next, combining the geographical distribution density and temporal fluctuation characteristics obtained in step 101, the review behavior patterns within each stage are analyzed, such as review intensity and fix response speed. Then, using graph database technology, each review behavior is modeled as a node in a graph, and the relationships between nodes are established based on their temporal order and spatial distribution, forming a review knowledge graph with spatiotemporal characteristics.

[0076] Based on e-commerce platform development data, the system identifies review behavior characteristics at different stages. In the early stages of feature development, the density of review comments in a certain area is low, and the fix cycle fluctuates less. However, during stress testing, the review density in this area increases sharply due to the need to handle defects in high-concurrency scenarios, and the fix cycle fluctuates more significantly. Through spatiotemporal clustering, the system constructs "low-density stable fix" and "high-density fluctuating fix" patterns as knowledge graph nodes, connecting them along the development stage axis. In the evolutionary path from "requirement design" to "stress testing," the pattern switching weights indicate that the testing phase is more likely to trigger high-density reviews. The review knowledge graph supports real-time updates; when new promotional features cause sudden defects, node attributes are dynamically adjusted to reflect the latest trends.

[0077] 103. Synchronously collect the syntactic structure features of the current software code change record, and jointly encode the syntactic structure features with the historical review trajectory in the review knowledge graph to generate processing instructions that include the priority of version difference parsing;

[0078] In this step, syntactic structure features refer to the distribution of syntactic elements in code changes, such as function call depth and dependency complexity.

[0079] Historical censorship trajectories refer to the sequence of past censorship behavior patterns recorded in the knowledge graph.

[0080] Version difference analysis priority refers to the review urgency indicator determined by the degree of difference between the current code changes and the historical trajectory.

[0081] Joint coding refers to a communication technology that integrates source coding and channel coding.

[0082] Software code change logs are the code modification paths recorded in a version control system, including the changes made in each commit, the author, the timestamp, and the associated review information.

[0083] Processing instructions are a set of commands in a computer system that instruct hardware or software to perform specific operations, covering a wide range from low-level hardware instructions to high-level document processing instructions.

[0084] In this embodiment, firstly, the change record of the current software code is captured in real time, including newly added, modified, or deleted code snippets. Then, a syntax parser is used to extract the syntactic structure features of this code, such as function call chains, control flow structures, and variable scope. Next, the current syntactic structure features are jointly encoded with the historical review trajectories in the review knowledge graph constructed in step 102. Natural language processing and graph neural network techniques are used to identify historical defect repair cases similar to the current change. Based on this, and combined with the complexity evaluation results of version differences, processing instructions containing priority ranking are generated.

[0085] During a code change to an e-commerce platform's user points system, developers refactored the points calculation logic. The system analyzed its syntactic structure features, such as multi-layered conditional statements and external interface dependencies, generating feature vectors. When matching this with the knowledge graph, it was found that the change closely resembled the trajectory of the "high-density fluctuation repair" pattern during the stress testing phase. Given that the system was currently in stress testing, it determined that this change required priority review, generating an "emergency response priority" instruction that required cross-regional collaborative handling of potential high-concurrency compatibility issues.

[0086] 104. Dynamically match the processing instructions with the domain knowledge coverage to determine the allocation strategy for cross-regional review tasks. The allocation strategy includes quality assessment nodes that are path-related to the review behavior pattern.

[0087] In this step, dynamic matching refers to adjusting task allocation based on the fit between real-time processing instructions and the knowledge coverage of reviewers.

[0088] The strategy for allocating cross-regional review tasks refers to developing a reasonable task scheduling mechanism based on the availability, professional competence, and geographical location of reviewers.

[0089] Path association refers to identifying a series of continuous paths with specific meanings by analyzing the connection relationships between nodes in graph structure data.

[0090] Quality assessment nodes refer to virtual nodes in a knowledge graph used to measure the rationality of task allocation, and their path association rules constrain the allocation logic.

[0091] In this embodiment, firstly, the priority in the processing instructions is dynamically matched with the domain knowledge coverage of the reviewers, and a collaborative filtering algorithm is used to screen candidate reviewers. Secondly, the path weights of the quality assessment nodes associated with the task in the review knowledge graph are analyzed to determine the importance of the task in the entire review process. Furthermore, geographical association markers are introduced into the matching results to consider factors such as network latency and time zone differences in the reviewers' locations, optimizing the allocation path for cross-regional review tasks. Finally, a review task allocation strategy that comprehensively considers technical capabilities, task priority, and geographical factors is generated.

[0092] In a cross-regional collaborative e-commerce platform development project, the system prioritizes reviewers with experience in handling high concurrency based on "emergency response priority." For example, a reviewer might have participated in high-concurrency module reviews multiple times during historical stress tests, but their assigned region is currently at full capacity. The system triggers the "load splitting rule" of the quality assessment node, dividing the task into core verification and auxiliary verification. Core verification is assigned to a high-experience-weighted team for real-time stress testing review, while auxiliary verification is assigned to an asynchronous review queue in a low-load region. The final strategy is "core team leads critical path review, auxiliary team verifies edge scenarios in parallel," and the task progress and defect fix rate are monitored in real-time through the quality assessment node.

[0093] 105. Based on the allocation strategy, version difference parsing priority, and quality assessment node mapping relationship, generate a review management and control strategy adapted to the software development system.

[0094] In this step, the review management and control strategy is a set of rules generated based on the mapping relationship between allocation strategy, version difference priority and quality assessment nodes, which is adapted to the development system. These rules include task scheduling order, review feedback incentive mechanism, etc.

[0095] Mapping relationships are a fundamental concept describing the correspondence rules between elements in the allocation strategy, version difference resolution priority, and quality assessment nodes.

[0096] In this embodiment, firstly, the allocation strategy obtained in step 104 is mapped and analyzed with the version difference parsing priority in step 103 to clarify the execution order and resource configuration requirements of different tasks. Then, combining the quality assessment nodes in the review knowledge graph in step 102, a correlation between task execution performance and quality indicators is established. Based on this, a rule-based reasoning mechanism and machine learning model are used to generate a set of review management and control strategies suitable for the current software development system, covering task scheduling rules, resource allocation schemes, and quality assurance measures. Finally, this control strategy is embedded into the automated process of the development system to achieve intelligent management of the entire process from code submission to review closure, improving overall collaboration efficiency and code quality.

[0097] In a cross-regional collaborative e-commerce platform system, task allocation strategies were linked to real-time load data and integrated with the collaboration interface. The core team's review tasks were set to real-time synchronous mode and required completion within a specified time window. Tasks for support teams were set to flexible mode, allowing for dynamic adjustment of deadlines. When the core team experienced a backlog of tasks due to unforeseen defects, the system automatically escalated some support verification tasks to urgent tasks and reassigned them to the redundant resource pool of high-load teams. Simultaneously, for task delays in low-load regions, the system triggered warnings and suggested temporarily adjusting the reviewers' work schedules. The final control strategy included dynamic priority reordering, flexible resource allocation, and a closed-loop anomaly feedback mechanism.

[0098] In summary, steps 101 to 105 construct a review knowledge graph by integrating multi-dimensional data such as geographical distribution, temporal fluctuations, and domain knowledge. This graph associates the syntactic features of code changes with historical trajectories to generate priority instructions. Combined with dynamic matching rules, it achieves precise allocation and priority control of cross-regional tasks, ultimately solving the problems of rigid task allocation and delayed priority response in remote collaboration, and improving review efficiency and system adaptability.

[0099] To address the insufficient coordination between cross-regional code review task allocation and version priority, a dynamic control strategy is constructed by analyzing the geographical correlation characteristics of cross-regional review tasks and the path weights of quality assessment nodes, combined with the priority hierarchy of code version difference analysis. First, the syntax structure matching depth of version differences is divided into continuous intervals, which are then cross-combined with path weights to generate a sequence of control parameters. Then, based on the distribution characteristics of hierarchical boundary values, geographical correlation markers are adjusted to form a spatial constraint relationship with the temporal fluctuations of the defect repair cycle, and this constraint is mapped to the hierarchical interval of the control parameters to generate a rule set. Finally, the knowledge coverage of reviewers is traversed through the collaborative interface of the software development system to integrate the multi-dimensional constraint boundaries into a review management control strategy. In some embodiments, step 105, which describes generating a review management control strategy adapted to the software development system based on the allocation strategy, version difference analysis priority, and the mapping relationship of quality assessment nodes, includes:

[0100] 201. Based on the geographical association markers of the cross-regional review tasks in the allocation strategy, extract the path weights of the quality assessment nodes;

[0101] In step 201, cross-regional review tasks typically refer to tasks such as work evaluation, project review, or academic review conducted between experts or teams in different geographical locations. Geographic association tags refer to labels in the task allocation strategy that identify the geographical region to which the task belongs, such as "Region A is high load, Region B is low density." Path weights refer to the association strength values ​​between quality assessment nodes and review behavior patterns in the review knowledge graph, reflecting the rationality of the task allocation path.

[0102] In this embodiment, firstly, based on the geographical association markers of cross-regional review tasks included in the allocation strategy, the distribution characteristics and collaborative relationships of review tasks across different geographical regions are identified. These geographical association markers are quantified using dimensions such as spatial distance, network latency, and the time zone of the reviewers, forming multi-dimensional geographical association features. Subsequently, quality assessment nodes associated with the current review task are located in the review knowledge graph. These nodes represent the assessment criteria for key quality risk points during code review. Based on this, combined with graph neural networks or path propagation algorithms, the path influence from the task's starting point to the quality assessment nodes is calculated along the node connection paths in the knowledge graph, thereby extracting the path weights of each quality assessment node.

[0103] 202. Divide the priority level of the version difference parsing into continuous intervals according to the matching depth of the syntax structure features, and cross-combine the intervals with the path weights to generate a sequence of control parameters indexed by the level boundary value.

[0104] In step 202, matching depth refers to the degree of similarity between the syntactic structure features of the code change and the historical review trajectory, such as "shallow matching" or "deep matching." Hierarchical boundary value refers to the quantized threshold of each interval boundary after dividing the version difference parsing priority into intervals according to matching depth. A control parameter sequence refers to a sequence formed by arranging a set of parameters in a certain order to control the behavior of a series of operations or processes. Crossover combination is a concept commonly found in combinatorics, experimental design, and genetic algorithms. An interval refers to the set of all points between two points within a hierarchical set.

[0105] In this embodiment, firstly, a clustering algorithm is used to divide the syntactic structure features of code changes related to version difference resolution priorities and the similarity to historical review trajectories into multiple continuous intervals, such as low matching, medium matching, and high matching. The boundary value of each interval is determined by dynamic threshold calculation. Subsequently, the continuous intervals are cross-combined with the path weights extracted in step 201, for example, high matching intervals are associated with high path weights, generating a sequence of control parameters indexed by boundary values.

[0106] 203. Based on the distribution characteristics of the hierarchical boundary values, the geographic association markers are dynamically corrected. The corrected geographic association markers form a spatial constraint relationship with the time fluctuation characteristics of the defect repair cycle, and the spatial constraint relationship is mapped to the hierarchical interval of the control parameter sequence to obtain the control rule set.

[0107] In step 203, spatial constraint relationships refer to the dynamic binding rules between the corrected geographic association markers and the fluctuation characteristics of the defect repair cycle. Dynamic correction typically refers to the process of adjusting a system, model, or parameters based on real-time feedback or changing conditions. A set of control rules is a set of rules and standards used to guide or constrain a system, process, or behavior.

[0108] In this embodiment, firstly, the distribution characteristics of the hierarchical boundary values ​​obtained in the analysis step are analyzed, and the density and trend of these values ​​in different regions are extracted using a sliding window statistical method. These distribution characteristics reflect the potential correlation between review activities and code defect repair in geospatial space. Next, the original geographic association markers are dynamically corrected using these distribution characteristics. Spatial interpolation algorithms are used to enhance the spatial continuity and rationality of the markers, enabling the corrected geographic association markers to more accurately reflect the distribution of actual review resources. Subsequently, the corrected geographic association markers are fused with the temporal fluctuation characteristics of the defect repair cycle to construct a spatial constraint relationship between the two, ensuring that the change in repair efficiency over time is constrained by the influence of geographic distribution. Finally, this spatial constraint relationship is mapped to the hierarchical intervals of the control parameter sequence. Based on the mapping results, a set of logically related control rules is generated to guide the optimized allocation and quality control of subsequent cross-regional review tasks. The entire process realizes a closed-loop derivation flow from spatial distribution characteristic analysis to control rule generation.

[0109] 204. Embed the set of control rules into the collaboration interface of the software development system, traverse the domain knowledge coverage of the reviewers through the collaboration interface, and merge the path weights, hierarchical boundary values ​​and spatial constraint relationships into multidimensional constraint boundaries based on the traversal results, and output the review management control strategy.

[0110] In step 204, the multidimensional constraint boundary refers to the joint constraint conditions for task allocation and priority control formed by the combination of path weights, hierarchical boundary values, and spatial constraint relationships. The collaboration interface refers to the interface through which different systems, modules, or organizations interact and communicate to achieve specific goals. Domain knowledge coverage refers to the total amount of knowledge, breadth, and depth possessed by an individual, team, or system within their professional domain. Review management and control strategies refer to the methodological system involving the comprehensive inspection, supervision, and adjustment of review processes, projects, or activities.

[0111] In this embodiment, the control rule set is first embedded into the collaborative interface of the development system. The domain knowledge coverage of reviewers is traversed, and their knowledge tag vectors are extracted; for example, a reviewer may be proficient in payment modules or high-concurrency scenarios. Then, based on the traversal results, path weights, hierarchical boundary values, and spatial constraints are superimposed and calculated. A weighted fusion algorithm is used to generate multi-dimensional constraint boundaries, such as "payment module tasks must satisfy a matching depth greater than a threshold and a path weight higher than the historical average." Finally, the task queue is dynamically adjusted according to the multi-dimensional constraint boundaries, and the review management control strategy is output.

[0112] Here is a specific example:

[0113] In the cross-regional development of promotional functions on e-commerce platforms, the system marks promotional tasks as "high-load areas" and extracts the path weights of associated quality assessment nodes from the knowledge graph, identifying them as high-weight values. This indicates that the path has a high success rate and strong quality assurance capabilities in historical reviews. Next, the system analyzes the syntactic structure characteristics of promotional code changes, dividing their matching depth into "high-match intervals." These intervals are then cross-combined with the aforementioned path weights to generate the control parameter item "high matching and high weight," used to identify tasks with high technical stability and review priority. Subsequently, considering that the proportion of high-match tasks exceeds the expected threshold, the system dynamically corrects the original geographical label "high-load area" to "elastic load area," establishing a mapping relationship between this and the fluctuation characteristics of the defect repair cycle during the testing phase. This generates the control rule "elastic areas only process high-match tasks" to achieve reasonable resource allocation. Finally, in the output control strategy stage, the system traverses the domain knowledge coverage of reviewers through the collaboration interface, combines path weights, hierarchical boundary values ​​and spatial constraints to form multi-dimensional constraint boundaries, and finally outputs the review management control strategy. This strategy requires core promotional tasks to be processed in real time by areas with high knowledge weights, while flexible areas are responsible for auxiliary asynchronous reviews. This ensures that key tasks are carried out efficiently and non-core tasks are flexibly scheduled, thereby improving the overall review efficiency and quality control capabilities in cross-regional development.

[0114] In summary, steps 201 to 204, by introducing key technologies such as geographic association tagging, path weight extraction, version difference resolution priority allocation, dynamic correction mechanisms, and multi-dimensional constraint boundary modeling, achieve integrated management of intelligent scheduling and quality control for cross-regional review tasks. This not only improves review efficiency but also enhances the ability to respond quickly to version differences, effectively reducing communication costs and quality risks caused by geographical, cultural, and time zone differences, and significantly improving the overall quality of software delivery and collaborative efficiency.

[0115] To address the dynamic adaptation problem of resource allocation and spatial constraints in cross-regional software development, this paper extracts the start and end boundary values ​​of the hierarchical intervals of the control parameter sequence and calculates the hierarchical span. A resource allocation base is generated according to a specific ratio, taking into account the time fluctuation characteristics of the defect repair cycle. Hierarchical intervals with strong correlations between path weights and hierarchical start boundary values ​​are selected to form a task triggering condition set. The base is adjusted according to a preset resource allocation ratio to determine the resource allocation threshold for the hierarchical intervals. The triggering conditions are then bound to the thresholds to generate a control rule set with dynamic hierarchical adaptation capabilities. In some embodiments, step 203, mapping the spatial constraint relationship to the hierarchical intervals of the control parameter sequence to obtain the control rule set, includes:

[0116] 301. Extract the starting boundary value and ending boundary value of each level interval of the control parameter sequence, determine the level span value by measuring the distance between the starting boundary value and the ending boundary value, and superimpose the level span value with the time fluctuation characteristics of the defect repair cycle in the spatial constraint relationship according to a preset ratio to generate the resource allocation base of each level interval.

[0117] In step 301, the hierarchical span value refers to the difference between the starting and ending values ​​of adjacent hierarchical intervals in the control parameter sequence, reflecting the quantitative range of the priority span. The resource allocation base value is a basic resource allocation value generated by superimposing the hierarchical span value and the fluctuation characteristics of the defect repair cycle, used to constrain the upper limit of resource allocation for tasks. The initial boundary value refers to the starting point or minimum effective value of the initial condition in the loop structure or data processing of each level of the control parameter sequence. The termination boundary value refers to the maximum effective value or ending point of the termination condition in the loop structure or data processing of each level of the control parameter sequence.

[0118] In this embodiment, firstly, the starting and ending boundary values ​​of each level interval are extracted layer by layer from the control parameter sequence. These boundary values ​​identify the division range between different control levels. Next, based on the starting and ending boundary values, the span value of each level is determined using the Euclidean distance calculation method. This span value reflects the coverage breadth of each level in the control space. Subsequently, the level span value is fused with the temporal fluctuation characteristics of the defect repair cycle contained in the spatial constraint relationship. By setting a weight ratio, the intensity of the temporal fluctuation is superimposed on the level span to reflect the impact of repair efficiency changes over time on resource distribution. Finally, the resource allocation base for each level interval is generated by comprehensively superimposing the results.

[0119] 302. Associate the path weight of the quality assessment node with the starting boundary value of the corresponding hierarchical interval, filter the hierarchical intervals whose association results exceed the preset activation value, and summarize to obtain the task triggering condition set.

[0120] In step 302, the initial boundary value association refers to the logical or mathematical relationship established between the initial boundary value and other variables, rules, or conditions in a specific system or process. The preset activation value refers to setting a threshold to filter out combinations where the association strength between the path weight and the hierarchical interval initial boundary value exceeds this threshold. The task triggering condition set refers to the set of high-priority task allocation conditions filtered out after associating the path weight with the hierarchical interval initial value.

[0121] In this embodiment, the starting boundary value of each hierarchical interval is first numerically correlated with the corresponding path weight to establish a mapping relationship between the boundary value and the quality impact intensity. Then, a preset activation value is set as a judgment threshold, and all hierarchical intervals whose correlation results exceed this threshold are selected, indicating that these hierarchical intervals have a high task priority in the current review process. Finally, all eligible hierarchical intervals are summarized to form a set of task triggering conditions used to drive review task scheduling. This set serves as the logical constraint for subsequent resource allocation.

[0122] 303. Adjust the resource allocation base according to the preset resource allocation ratio, determine the resource allocation threshold of the hierarchical interval, and bind the hierarchical interval of the task triggering condition set with the resource allocation threshold to obtain the control rule set.

[0123] In step 303, the resource allocation ratio refers to the proportional coefficient for allocating limited resources to different levels, tasks, or regions according to specific rules. The level interval refers to the numerical range or logical boundary between different levels in a specific system or process. The resource allocation threshold refers to the upper limit of task allocation resources adjusted based on the resource allocation base, used to control the task load at different levels. The control rule set refers to a set of rules designed to achieve a specific goal, which, through mechanisms such as logical conditions, weight allocation, and dynamic adjustment, constrains, optimizes, or automates the management of system behavior, resource allocation, or process execution.

[0124] In this embodiment, firstly, based on historical resource utilization data, a linear regression model is used to fit the resource allocation ratio. The resource allocation base is then normalized according to the resource allocation ratio to generate a resource allocation threshold. Subsequently, the task triggering condition set and the resource allocation threshold are subjected to a Cartesian product operation to form all possible resource allocation combinations. A decision tree pruning algorithm is used to bind and eliminate inefficient combinations, retaining high-efficiency, high-coverage resource allocation schemes. The final scheme is transformed into a set of control rules, such as "if the weight of a certain level interval path exceeds the threshold, then allocate the corresponding resources."

[0125] Here is a specific example:

[0126] In the cross-regional e-commerce development of a promotional function, firstly, for the starting and ending values ​​of high-matching-level intervals in promotional tasks, the span value was calculated and combined with the fluctuation characteristics of delay fixing during the testing phase. This was then superimposed according to a preset ratio to generate a high-base resource allocation threshold, used to identify the high resource demand of tasks at that level. Next, the path weights of high-matching levels were correlated with the starting values, filtering out intervals with correlation exceeding the activation threshold, forming the task triggering condition set "high-matching promotional tasks should be prioritized," ensuring priority scheduling for critical tasks. Subsequently, based on the preset resource allocation ratio, the resource threshold for high-matching tasks was dynamically adjusted, generating the rule "high-matching task resource ratio does not exceed the preset ratio, medium and low-matching tasks are dynamically supplemented," achieving hierarchical resource control. Finally, during peak promotional periods, the system automatically limited the resource usage limit for high-matching tasks, prioritizing the efficiency of core function review, while flexibly allocating medium and low-matching tasks to auxiliary areas. This dynamic resource scheduling balanced the processing needs of critical and non-core tasks, ensuring overall development progress and quality.

[0127] In summary, steps 301 to 303, through the dynamic overlay of hierarchical span values ​​and time fluctuation characteristics, combined with the precise matching of path weights and resource allocation thresholds, enable intelligent adaptation and dynamic optimization of resource allocation in cross-regional review tasks. The resource allocation base of hierarchical intervals can reflect regional load and repair efficiency in real time, the task triggering condition set ensures priority scheduling of critical paths, and the resulting set of control rules significantly improves resource utilization and review efficiency while reducing scheduling conflicts and quality risks caused by spatial constraints, providing a systematic solution for complex distributed development scenarios.

[0128] To address the issue of dynamic adaptation between version difference resolution priority and resource allocation in cross-regional collaboration, based on the syntax structure matching depth of version difference resolution priority, the hierarchy is arranged in ascending order of numerical range and segmented into continuous intervals, with interval boundaries defined as hierarchy boundary values. Code change records within each interval are traversed, their correlation values ​​with path weights are calculated, and these are combined proportionally with the hierarchy boundary values ​​to generate a sequence of control parameters. The index mapping between hierarchy boundary values ​​and parameter sequences is implemented through encoding. In some embodiments, step 202, which involves dividing the version difference resolution priority hierarchy into continuous intervals according to the matching depth of syntax structure features, and cross-combining these intervals with path weights to generate a sequence of control parameters indexed by hierarchy boundary values, includes:

[0129] 401. Based on the numerical range of the matching depth of the syntax structure features of the version difference parsing priority, the levels of the version difference parsing priority are arranged in ascending order, and the numerical range is divided into continuous intervals according to a preset segmentation density. The starting and ending values ​​of the intervals are defined as level boundary values.

[0130] In step 401, "arranging levels in ascending order" means sorting different levels in the system from low to high. "Numerical range" refers to the range of values ​​for a certain level or parameter, usually defined by a minimum and maximum value. "Segment density" refers to a preset interval division rule used to control the granularity of priority levels, such as the matching depth range covered by each segment. Further refinement of "arranging levels in ascending order" emphasizes the progressive relationship and sorting logic between levels. For example, in version control, the major version number has the highest priority in ascending order, followed by the minor version number, and finally the revision number. "Level boundary values" refer to the starting and ending values ​​determined when dividing intervals, used to identify the upper and lower limits of the priority level.

[0131] In this embodiment, code change records are first obtained from the version difference resolution priority, and their abstract syntax tree is parsed. Syntactic structure features such as nesting levels, node types, and dependencies are extracted from the abstract syntax tree. An abstract syntax tree parsing tool is used to extract node levels, and a depth-first search traversal is used to calculate the numerical range of matching depths, arranging the levels in ascending order. Subsequently, based on a preset segmentation density, a linear interpolation algorithm is used to divide the numerical range of matching depths into continuous intervals. The starting and ending values ​​of each interval are the level boundary values, used for the subsequent index construction of the control parameter sequence.

[0132] 402. Traverse the code change records within the interval, calculate the correlation value between the code change record and the path weight, combine the correlation value with the hierarchical boundary value according to a preset ratio, encode the combination result in ascending hierarchical order, and generate a sequence of control parameters indexed by the hierarchical boundary value.

[0133] In step 402, the correlation value refers to a quantitative indicator of the matching degree between code change records and path weights, reflecting the collaborative efficiency of task allocation paths and priorities. Encoding in ascending order of hierarchy refers to sorting the combined results from low to high priority and generating an index sequence. Preset ratio refers to a predefined ratio or size configuration in a specific scenario, which users can directly select to quickly adapt to their needs. The control parameter sequence is a set of parameters generated according to rules or algorithms, used to dynamically adjust key levels or indicators in system, process, or resource allocation.

[0134] In this embodiment, the code change records within each interval are first traversed, and the syntactic structure features of the code change record locations are extracted. The correlation value, calculated by matrix multiplication, is the product of the syntactic structure matching depth and the path weight. Then, a linear regression model or weighted average algorithm is used to combine the correlation value with the hierarchical boundary value according to a preset ratio. The combined result is encoded in ascending hierarchical order to generate a sequence of control parameters indexed by the hierarchical boundary value.

[0135] Here is a specific example:

[0136] In a cross-regional development of an e-commerce promotion function, the system first divides the hierarchical range based on the matching depth of the syntactic structure features of promotion code changes. The matching depth range is divided into three intervals: low, medium, and high. The low interval corresponds to changes with shallow syntactic nesting, the medium interval to changes with medium complexity, and the high interval to changes with deep nesting. Then, the system traverses the promotion code change records within the high interval, extracts their syntactic structure features, and calculates the association value with the path weight. If a change in a high interval has deep nesting and a high path weight, the association value is obtained by multiplying the matching depth by the path weight. Next, this high association value is linearly superimposed with the high interval boundary value according to a preset ratio to generate the parameter item "High Boundary and High Association". Finally, the system encodes the boundary values ​​and corresponding association values ​​of the low, medium, and high intervals in ascending order of hierarchy, forming an index sequence containing three entries: "Low Boundary and Low Association", "Medium Boundary and Medium Association", and "High Boundary and High Association". This set of control parameters is used to guide the allocation of promotional tasks. Highly correlated parameter items trigger priority processing in core areas, ensuring rapid review and resource allocation for code changes to key promotional logic, thereby improving the development efficiency and stability of core functions in cross-regional collaboration.

[0137] In summary, steps 401 to 402 divide the hierarchical intervals by matching the depth of syntactic structural features, quantifying the granularity of version difference parsing priority. Combined with path weights, a sequence of control parameters is generated to achieve dynamic adaptation of resource allocation and code change priority. In cross-regional collaboration scenarios, this improves the review efficiency of core functional modules, reduces the risk of resource conflicts in auxiliary functional modules, and optimizes the stability and response speed of the overall development process.

[0138] To address the dynamic adaptation issue between version difference resolution priority and code review efficiency, this method involves capturing the syntactic structure features of code change records in real time and decomposing them into feature fragments. These fragments are then matched multidimensionally with the syntactic feature vectors and behavioral patterns of historical review trajectories. Effective associations are filtered based on the matching degree between feature fragments and historical data. The spatiotemporal correlation strength and syntactic complexity are then combined to generate priority weights, which, in conjunction with the version change scope, generate processing instructions that include version difference resolution priority. In some embodiments, step 103 involves synchronously collecting the syntactic structure features of the current software code change records and jointly encoding these features with historical review trajectories in the review knowledge graph to generate processing instructions that include version difference resolution priority, including:

[0139] 501. Capture the syntactic structure features of the current software code change record in real time, decompose the syntactic structure features into feature fragments, and convert the feature fragments into a current feature vector set of uniform dimension;

[0140] In step 501, syntactic structure features refer to the structured information of the abstract syntax tree in the code change record, such as node types, nesting levels, and dependencies. Feature fragments refer to the independent syntactic elements decomposed from the code change record. The current feature vector set refers to the set of numerical representations of feature fragments converted into a unified dimension for quantitative analysis. The current feature vector set refers to the set of numerical vectors of a unified dimension converted from feature fragments by the encoder for subsequent matching calculations.

[0141] In this embodiment, firstly, a parser captures the current software code change record in real time and extracts its syntactic structure features, such as function call relationships, control flow structure, and variable reference patterns. Next, these structured features are segmented according to semantic granularity, forming multiple identifiable feature segments. Each feature segment is then transformed into a uniform-dimensional numerical vector through embedding encoding techniques, such as word vector mapping or graph neural network representation learning, and finally integrated into the current feature vector set. This vector set is used for subsequent comparative analysis with historical data.

[0142] 502. Extract the syntactic structure feature vectors and review behavior patterns associated with the nodes of the historical review trajectory in the review knowledge graph to form a historical feature vector set;

[0143] In step 502, the syntax structure feature vector refers to the abstract syntax tree encoding vector of the code change node in the historical review trajectory. The review behavior pattern refers to the sequence of operations associated with code changes in the historical review. The historical feature vector set refers to the set of syntax feature vectors corresponding to past review behavior patterns stored in the review knowledge graph.

[0144] In this embodiment, node information related to the current code change is retrieved from the review knowledge graph to obtain the associated historical review trajectory, including review records and defect repair paths for similar syntax structures in previous versions. Furthermore, the corresponding syntax structure feature vectors and associated review behavior patterns, such as review frequency, repair cycle, and reviewer preferences, are extracted from these trajectories. Finally, the above information is organized into a historical feature vector set, which is used for joint analysis with the current feature vector set to support intelligent matching and optimization decisions for subsequent review tasks.

[0145] 503. Calculate the matching degree between the feature fragments of the current feature vector set and the corresponding grammatical structure feature vectors in the historical feature vector set, filter the feature fragments with matching degrees exceeding a preset threshold, and generate a feature matching mapping table.

[0146] In step 503, the matching degree calculation measures the similarity between the current feature vector and historical feature vectors, and is used to filter highly relevant segments. The feature matching mapping table records the matching relationship between the current feature segment and historical feature vectors, and is used to filter highly relevant features.

[0147] In this embodiment, the similarity between each feature segment of the current feature vector set and the corresponding dimension vector in the historical feature vector set is first calculated using the cosine similarity formula. Then, a threshold is set, and segment pairs with similarity higher than the threshold are retained. If the similarity exceeds the preset threshold, the segment is marked as a high match. For example, if the current interface call chain segment has a high similarity to the historical high-concurrency interface mode vector, a mapping entry "interface call chain and high-concurrency mode" is generated. All matching pairs are traversed, low-similarity segments are filtered, and finally, a feature matching mapping table is generated.

[0148] 504. Based on the matching feature fragments in the feature matching mapping table, extract the spatiotemporal correlation strength of the review behavior pattern, and calculate the spatiotemporal correlation strength and the syntactic structure complexity of the matching feature fragments by superimposing them to generate priority weight parameters for each matching feature fragment.

[0149] In step 504, the spatiotemporal correlation strength refers to the correlation strength value of historical review behavior patterns in the temporal and spatial dimensions. Syntactic structure complexity refers to the complexity indicators such as the nesting level of the abstract syntax tree of the feature fragment and the number of nodes. The priority weight parameter refers to a comprehensive score combining spatiotemporal correlation strength and complexity, used to determine processing priority.

[0150] In this embodiment, firstly, matched feature fragments are extracted from the feature matching mapping table. These fragments represent the semantically similar parts between the current code change and historical review records. Next, for each matched feature fragment, combined with its associated review behavior pattern in the review knowledge graph, a spatiotemporal correlation analysis method is used to extract the activity level of this pattern in different development stages and geographical regions, forming a spatiotemporal correlation strength value. This strength value reflects the degree of attention the feature fragment received during historical reviews. Subsequently, the syntactic structural complexity corresponding to each matched feature fragment is obtained. This complexity is quantified using indicators such as the depth of the abstract syntax tree or the number of control flow paths to measure the difficulty and risk of the fragment in code implementation. Then, the spatiotemporal correlation strength and syntactic structural complexity are superimposed according to a set fusion ratio to form a comprehensive impact score. Finally, this score is normalized to generate priority weight parameters for each matched feature fragment, providing a basis for the scheduling and resource allocation of subsequent review tasks.

[0151] 505. The priority weight parameter is proportionally fused with the version change coverage of the corresponding feature segment to generate a processing instruction that includes the version difference parsing priority.

[0152] In step 505, the version change coverage refers to the range of functional modules affected by the code change. Proportional fusion is a method that linearly combines multiple independent parameters into a single composite value by setting different weight ratios for each parameter. Processing instructions are specific operational guidelines generated based on the proportional fusion results, used to allocate review resources, adjust priorities, or trigger automated processes.

[0153] In this embodiment, firstly, the number of files, lines modified, and module levels involved in the current code change are counted. For example, metadata of the change records is obtained through the version control system interface, and the coverage value is calculated. Subsequently, the priority weight parameters are proportionally merged with the version coverage, the priority tags of all fragments are integrated, and a processing instruction is generated, such as "core interface change requires urgent processing".

[0154] Here is a specific example:

[0155] In the development of a promotional campaign for an e-commerce platform, a version control system was used to capture code change records in real time, analyze the "points calculation logic refactoring" feature fragment, and convert it into a four-dimensional vector using a word embedding model. Subsequently, feature vectors of historical "high-concurrency points review modes" were extracted from the review knowledge graph to generate a mapping entry "distributed locks and high-concurrency modes." Next, weight parameters were generated by proportionally superimposing the spatiotemporal correlation strength and syntactic complexity of historical behavior patterns. Finally, combining the coverage and weight parameters, a processing instruction priority was generated by proportionally fusing them, outputting the instruction "High-concurrency points changes require cross-regional real-time collaborative review," ensuring that review resources are prioritized for high-risk modules.

[0156] In summary, steps 501 to 505, through real-time capture of the syntactic structural features of code changes and joint encoding with historical review data, quantify the priority of version difference parsing, thereby achieving dynamic adaptation of code review resources. In cross-regional collaboration scenarios, this method improves the review efficiency of core modules, reduces the risk of resource conflicts in auxiliary modules, and optimizes the stability and response speed of the overall development process.

[0157] To address the issue of dynamic adaptation of code review resource allocation and version difference parsing priorities, this method decomposes the feature fragments of current code changes and historical feature vectors by dimension, calculating the difference score between independent dimensions and historical dimensions. Feature fragments with more than a set standard of dimension differences below a threshold are selected as valid matching fragments. A mapping table is generated based on identifier correspondences to establish the association between feature fragments and historical syntactic features. In some embodiments, step 503, which involves calculating the matching degree between feature fragments of the current feature vector set and the corresponding syntactic structure feature vectors of the historical feature vector set, selecting feature fragments with matching degrees exceeding a preset threshold, and generating a feature matching mapping table, includes:

[0158] 601. Decompose the feature segments in the current feature vector set into independent dimension values ​​according to the dimension, and decompose the corresponding grammatical structure feature vectors in the historical feature vector set into historical dimension values ​​according to the same dimension.

[0159] In step 601, the independent dimension value refers to the numerical representation of a single feature segment in a certain dimension of the current feature vector set. The historical dimension value refers to the numerical representation of the syntactic structure feature vectors in the corresponding dimension of the historical feature vector set.

[0160] In this embodiment, firstly, each feature fragment in the current feature vector set is decomposed according to its vector dimension to obtain independent dimension values ​​for each dimension. These independent dimension values ​​represent the specific manifestation of the current code change on different syntactic structure features. Next, the syntactic structure feature vectors corresponding to the current feature fragment in the historical feature vector set are decomposed using the same dimensionality partitioning method to extract the corresponding historical dimension values. This process ensures that current and historical data are comparable within the same semantic space. This decomposition operation provides fundamental data support for subsequent calculations of the differences between the two datasets.

[0161] 602. Calculate the difference between the independent dimension value and the historical dimension value, generate the difference score for each dimension, and filter the number of dimensions with the difference score below a preset threshold in the feature fragment. When the number of dimensions exceeds the preset number, mark the corresponding feature fragment as a valid matching fragment.

[0162] In step 602, the difference score refers to the quantitative indicator of the difference between the current independent dimension value and the historical dimension value within the same dimension. The preset threshold is used to filter dimensions with low difference scores, and the preset quantity limits the minimum dimension matching requirement for valid matching segments. Valid matching segments refer to feature segments with difference scores below the threshold on multiple dimensions and that meet the quantity requirements.

[0163] In this embodiment, based on the independent dimension values ​​and historical dimension values ​​obtained in step 601, the difference between the two is calculated for each dimension one by one, and algorithms such as Euclidean distance or cosine similarity are used to generate a difference score for each dimension. Subsequently, the number of dimensions with a difference score lower than a preset threshold is counted in each feature segment, and this number is compared with the set minimum number of matching dimensions. If the number of dimensions with low difference scores in a feature segment exceeds this number, the feature segment is determined to have a high degree of matching with the historical records and is marked as a valid matching segment.

[0164] 603. Match the identifier of the corresponding syntactic structure feature vector with the identifier of the valid matching segment, and generate a feature matching mapping table that reflects the mapping relationship between the feature segment and the syntactic structure feature vector.

[0165] In step 603, identifier matching refers to associating the current feature fragment with the historical syntactic structure feature vector using a unique identifier. The feature matching mapping table refers to associating identifiers to reflect the matching relationship between the current and historical feature fragments.

[0166] In this embodiment, based on the valid matching segments marked in step 602, their unique identifiers are extracted, and the identifiers of the matching grammatical structure feature vectors are searched in the historical feature vector set. By establishing the association between the two, a table containing the mapping relationship between feature segments and historical grammatical structure feature vectors is formed, namely, a feature matching mapping table. This mapping table provides key data for subsequent analysis of review behavior patterns and generation of priority weight parameters, realizing a complete process connection from original feature recognition to historical pattern matching.

[0167] Here is a specific example:

[0168] In developing the points calculation module for e-commerce platform promotional functions, the team needed to address the issue of prioritizing code changes under high-concurrency scenarios. The system first captures current code change records in real-time through a version control system, parses out the distributed lock feature fragments for points, and decomposes them into independent values ​​for three dimensions: function call depth, nested loop level, and number of external dependencies, generating a current feature vector. Then, it extracts feature vectors of historical high-concurrency points patterns from the review knowledge graph, correspondingly decomposing them into historical dimensions of function call depth, nested loop level, and number of external dependencies. By calculating the difference in each dimension, if the difference in function call depth is below a threshold, the difference in nested loop level meets the standard, the difference in the number of external dependencies meets the standard, and the cumulative number of dimensions meeting the standard exceeds a preset number, the feature fragment is marked as a valid match. When generating the mapping table, the identifiers of the distributed lock fragments for points and historical high-concurrency points pattern vectors are extracted using a hash algorithm, establishing a key-value pair mapping relationship. Recording this entry triggers a high-priority task allocation mechanism. Finally, the priority weights and coverage are proportionally integrated to generate processing instructions, guiding the review team to prioritize the changes and ensure the stability of the points module and efficient resource allocation during promotional activities.

[0169] In summary, steps 601 to 603, by comparing the independent dimension values ​​of the feature vector with historical dimension values ​​and combining this with a difference-based scoring and filtering mechanism, achieve precise matching between current code changes and historical review records. For example, in the development of promotional functions on e-commerce platforms, this method significantly improves the efficiency of determining the review priority of core modules, reduces resource waste caused by code changes with low matching degrees, and enhances the traceability and collaboration of the review process through an identifier mapping table.

[0170] To address the issues of spatiotemporal correlation analysis and knowledge transfer path modeling in code review behavior during software development, this paper constructs geographic density parameters and temporal fluctuation parameters by quantifying the geographic distribution density of code review comments and the temporal fluctuation of defect repair cycles. Using the development stage as the evolutionary axis, these two parameters are integrated to generate review behavior pattern parameters and mapped to independent nodes. Spatiotemporal correlation paths are established based on the geographic density difference and temporal fluctuation correlation between adjacent nodes. Knowledge transfer links are dynamically generated by binding reviewer knowledge weight vectors, ultimately integrating them into a review knowledge graph with spatiotemporal correlation and knowledge flow characteristics. In some embodiments, step 102, which involves constructing a review knowledge graph with the software development stage as the evolutionary axis based on the spatiotemporal correlation characteristics of the code review dataset, includes:

[0171] 701. Convert the geographical distribution density of review comments in the code review dataset into a geographical density parameter, convert the time fluctuation characteristics of the defect repair cycle into a time fluctuation parameter, and decompose the domain knowledge coverage of reviewers into a knowledge weight vector.

[0172] In step 701, the geographic density parameter refers to the quantified value of the density of review comments distributed across different regions. The time fluctuation parameter refers to the quantified value of the fluctuation in the defect repair cycle over time. The knowledge weight vector is a numerical representation of the reviewer's knowledge coverage in a specific technical field.

[0173] In this embodiment, firstly, the geographical distribution density of review comments in the code review dataset is numerically processed and converted into a geographical density parameter that can be used for calculation. This parameter reflects the concentration of review activities in different regions. Next, considering the temporal fluctuation characteristics of the defect repair cycle, time series analysis is used to extract its periodic and sudden change patterns, which are then converted into time fluctuation parameters to measure the trend of repair efficiency over time. Subsequently, the domain knowledge coverage of reviewers is decomposed according to the distribution of technical fields, and a weighted vector representation is used to generate a knowledge weight vector. Each dimension corresponds to a technical direction, and the weight value reflects the reviewer's professional competence in that direction. The construction of these three parameters provides a foundation for subsequent behavioral pattern modeling.

[0174] 702. Using the time sequence of the software development stage as the evolution axis, combine the geographic density parameter and the time fluctuation parameter according to a preset weight ratio to generate review behavior pattern parameters;

[0175] In step 702, the review behavior pattern parameter refers to a composite parameter combining geographical density and temporal fluctuation characteristics, reflecting the review behavior patterns at a specific development stage. The evolution axis refers to the time sequence of stages such as requirements analysis, design, development, testing, and deployment, and the preset weight ratio is used to balance the impact of geographical density and temporal fluctuations.

[0176] In this embodiment, based on the chronological order of the software development stages, the geographic density parameter and temporal fluctuation parameter obtained in step 701 are fused and calculated according to a preset weight ratio to form a comprehensive review behavior pattern parameter. This parameter combines spatial distribution and temporal evolution through mathematical weighting, reflecting the typical characteristics of review behavior in different development stages. For example, the design and implementation stage may exhibit a combination of high geographic density and low temporal fluctuation. This behavior pattern parameter serves as the basic unit for map construction, used to characterize the dynamic performance of review activities in the spatiotemporal dimensions.

[0177] 703. Map the review behavior pattern parameters to independent nodes, and establish spatiotemporal association paths between nodes based on the difference in geographic density parameters between adjacent nodes and the correlation of time fluctuation parameters.

[0178] In step 703, the spatiotemporal correlation path refers to the evolutionary relationship path between review behavior pattern nodes based on the correlation between geographical density differences and temporal fluctuations. The change in the difference of the geographical density parameter reflects the migration trend of the distribution of review opinions in adjacent stages. The correlation of the temporal fluctuation parameter reflects the dynamic relationship of the repair cycle in adjacent stages.

[0179] In this embodiment, the review behavior pattern parameters generated in step 702 are mapped to independent nodes in a knowledge graph, with each node representing the review behavior state at a specific development stage. Subsequently, the correlation between the differences in geographical density parameters and temporal fluctuation parameters between adjacent nodes is analyzed. Using correlation coefficient calculation and distance measurement methods, spatiotemporal association paths between nodes are constructed. This path describes the evolution logic of review behavior across different stages, enhancing the temporal coherence and spatial interpretability of the knowledge graph, and laying a structural foundation for subsequent knowledge transfer mechanisms.

[0180] 704. Dynamically bind the knowledge weight vector with the review behavior mode parameters to obtain the knowledge transfer links between nodes, and generate a review knowledge graph by integrating the nodes, spatiotemporal association paths and knowledge transfer links.

[0181] In step 704, the knowledge transfer link refers to the dynamic binding relationship between the reviewer's knowledge weight vector and the review behavior pattern node, reflecting the path of knowledge coverage's influence on review behavior. The review knowledge graph is a semantic network built based on code review data. By integrating the spatiotemporal correlation characteristics of review behavior with the reviewer's knowledge coverage, it forms a dynamic knowledge model with the software development stage as the evolution axis.

[0182] In this embodiment, the knowledge weight vector obtained in step 701 is dynamically bound to the review behavior mode parameters corresponding to each node to simulate the flow of review knowledge between different review behavior states, forming a knowledge transfer link. This link reflects how the professional competence of reviewers influences the evolution of review behavior. Finally, all nodes, spatiotemporal association paths, and knowledge transfer links are integrated to construct a complete review knowledge graph.

[0183] Here is a specific example:

[0184] In the high-concurrency development of payment modules for e-commerce platforms, the review knowledge graph integrates the spatiotemporal correlation characteristics of code review datasets to achieve precise management of cross-regional collaboration and knowledge transfer. The system first calculates the ratio of review comments to area in a given region, generating a geographic density parameter; analyzes the fluctuations in defect repair cycles to generate time fluctuation parameters; and generates a knowledge weight vector based on the reviewers' historical contributions to module types and frequencies. Subsequently, using the development phase's chronological order as an axis, it calculates review behavior pattern parameters for the testing phase according to preset weight ratios, combining the changes in geographic density differences between adjacent phases with the correlation of time fluctuations to generate path weights. Finally, the pattern parameters are bound to the knowledge weight vector, and knowledge transfer links are generated through matrix operations, integrating nodes, paths, and links to form the review knowledge graph. This graph identifies the high volatility of the testing phase and the concentrated distribution of cross-regional review comments, triggering a dynamic resource adjustment mechanism to guide experts to prioritize payment module reviews, significantly improving code quality and cross-regional collaboration efficiency.

[0185] In summary, steps 701 to 704, through quantitative analysis of geographical density and temporal fluctuation parameters combined with dynamic binding of knowledge weight vectors, enable this method to achieve spatiotemporal correlation modeling of code review behavior and visualization of knowledge transfer paths. In the development of e-commerce payment modules, the system can accurately identify high-risk stages, optimize cross-regional review resource allocation, and guide experts to participate in the review of key modules through knowledge transfer links, significantly improving review efficiency and code quality.

[0186] Figure 2 This application provides a schematic diagram of the structure of a software development system management system based on big data, as shown in the embodiments of this application. Figure 2 As shown, the system includes:

[0187] The acquisition module 21 acquires the code review dataset generated during the remote collaborative software development process. The code review dataset includes the geographical distribution density of review comments, the time fluctuation characteristics of defect repair cycles, and the domain knowledge coverage of reviewers.

[0188] Module 22 constructs a review knowledge graph with the software development stage as the evolution axis based on the spatiotemporal correlation characteristics of the code review dataset. The review knowledge graph contains review behavior patterns associated with the geographical distribution density and temporal fluctuation characteristics.

[0189] Encoding module 23 synchronously collects the syntactic structure features of the current software code change record, and jointly encodes the syntactic structure features with the historical review trajectory in the review knowledge graph to generate processing instructions that include the priority of version difference parsing;

[0190] Matching module 24 dynamically matches the processing instructions with the domain knowledge coverage to determine the allocation strategy for cross-regional review tasks. The allocation strategy includes quality assessment nodes that are path-related to the review behavior pattern.

[0191] The generation module 25 generates a review management and control strategy adapted to the software development system based on the allocation strategy, version difference parsing priority, and the mapping relationship between quality assessment nodes.

[0192] Figure 2 The aforementioned big data-based software development system management system can execute... Figure 1 The implementation principle and technical effects of the software development system management method based on big data described in the illustrated embodiment will not be repeated here. The specific methods by which each module and unit of the software development system management system based on big data in the above embodiments perform operations have been described in detail in the embodiments related to this method, and will not be elaborated upon here.

[0193] In one possible design, Figure 2 The software development system management system based on big data in the illustrated embodiment can be implemented as a computing device, such as... Figure 3 As shown, the computing device may include a storage component 31 and a processing component 32;

[0194] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are invoked and executed by the processing component 32.

[0195] The processing component 32 is used for the above Figure 1 The embodiment describes a software development system management method based on big data.

[0196] The processing component 32 may include one or more processors to execute computer instructions to complete all or part of the steps in the above-described method. Alternatively, the processing component may be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above-described method.

[0197] Storage component 31 is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0198] Of course, computing devices may also include other components, such as input / output interfaces, display components, communication components, etc.

[0199] Input / output interfaces provide interfaces between processing components and peripheral interface modules, which can be output devices, input devices, etc.

[0200] The communication components are configured to facilitate wired or wireless communication between computing devices and other devices.

[0201] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform. In this case, the computing device can refer to a cloud server, and the aforementioned processing components, storage components, etc., can be basic server resources rented or purchased from the cloud computing platform.

[0202] This application also provides a computer storage medium storing a computer program, which, when executed by a computer, can perform the above-described functions. Figure 1 The illustrated embodiment is a software development system management method based on big data.

[0203] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0204] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0205] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0206] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A big data-based software development system management method, characterized by, The method comprises the following steps: acquire a code review dataset generated in a remote collaborative software development process, the code review dataset containing a geographical distribution density of review opinions, a time fluctuation feature of a defect repair cycle, and a domain knowledge coverage of reviewers; based on the spatio-temporal correlation characteristics of the code review dataset, construct a review knowledge graph with the software development stage as the evolution axis, the review knowledge graph containing review behavior patterns associated with the geographical distribution density and the time fluctuation feature; synchronously collect the syntax structure features of the current software code change records, and jointly encode the syntax structure features and the historical review trajectory in the review knowledge graph to generate processing instructions containing version difference analysis priorities; dynamically match the processing instructions with the domain knowledge coverage to determine a distribution strategy of cross-regional review tasks, the distribution strategy containing quality evaluation nodes path-associated with the review behavior patterns; generate a review management regulation strategy adapted to the software development system according to the mapping relationship among the distribution strategy, the version difference analysis priority and the quality evaluation node.

2. The method of claim 1, wherein, The method of generating a review management regulation strategy adapted to the software development system according to the mapping relationship among the distribution strategy, the version difference analysis priority and the quality evaluation node comprises: based on the geographical correlation markers of the cross-regional review tasks in the distribution strategy, extract the path weight of the quality evaluation node; divide the version difference analysis priority level into continuous intervals according to the matching depth of the syntax structure features, and cross-combine the intervals with the path weight to generate a regulation parameter sequence indexed by level boundary values; according to the distribution characteristics of the level boundary values, dynamically correct the geographical correlation markers, and form a spatial constraint relationship between the corrected geographical correlation markers and the time fluctuation feature of the defect repair cycle, and map the spatial constraint relationship to the level interval of the regulation parameter sequence to obtain a regulation rule set; embed the regulation rule set into the collaboration interface of the software development system, traverse the domain knowledge coverage of the reviewers through the collaboration interface, and combine the path weight, the level boundary value and the spatial constraint relationship into a multi-dimensional constraint boundary according to the traversal result to output the review management regulation strategy.

3. The method of claim 2, wherein, The method of mapping the spatial constraint relationship to the level interval of the regulation parameter sequence to obtain a regulation rule set comprises: extract the starting boundary value and the ending boundary value of each level interval of the regulation parameter sequence, determine the level span value through the distance measure of the starting boundary value and the ending boundary value, superimpose the level span value and the time fluctuation feature of the defect repair cycle in the spatial constraint relationship according to a preset proportion to generate a resource allocation basis for each level interval; associate the path weight of the quality evaluation node with the starting boundary value of the corresponding level interval, filter the level intervals whose associated results exceed a preset activation value, and obtain a task triggering condition set by summarizing. According to the preset resource allocation ratio, the resource allocation base is adjusted to determine a resource allocation threshold of the hierarchical interval, the hierarchical interval in the task triggering condition set is bound to the resource allocation threshold, and a regulation set is obtained.

4. The method of claim 2, wherein, The hierarchical interval of the version difference analysis priority is divided into continuous intervals according to the matching depth of the syntax structure feature, and the intervals are cross-combined with the path weight to generate a regulation parameter sequence indexed by the hierarchical boundary value, including: According to the numerical range of the matching depth of the syntax structure feature of the version difference analysis priority, the hierarchical interval of the version difference analysis priority is arranged in ascending order, and the numerical range is divided into continuous intervals according to a preset segmentation density, and the starting value and the ending value of the interval are defined as the hierarchical boundary value; Traverse the code change record in the interval, calculate the association value of the code change record and the path weight, and combine the association value with the hierarchical boundary value according to a preset proportion, encode the combination result in ascending order of the hierarchical interval to generate a regulation parameter sequence indexed by the hierarchical boundary value.

5. The method of claim 1, wherein, The syntax structure feature of the current software code change record is synchronously collected, and the syntax structure feature is jointly encoded with the historical review track in the review knowledge graph to generate a processing instruction containing the version difference analysis priority, including: Real-time capture the syntax structure feature of the current software code change record, decompose the syntax structure feature into feature segments, and convert the feature segments into a set of current feature vectors of uniform dimensions; Extract the syntax structure feature vector and the review behavior mode associated with the node of the historical review track in the review knowledge graph to form a set of historical feature vectors; Calculate the matching degree of the feature segments of the current feature vector set and the syntax structure feature vectors of the corresponding dimensions in the historical feature vector set, filter the feature segments with a matching degree exceeding a preset threshold, and generate a feature matching mapping table; Based on the matching feature segments in the feature matching mapping table, extract the spatio-temporal association strength of the review behavior mode, superimpose calculate the syntax structure complexity of the matching feature segments, and generate the priority weight parameter of each matching feature segment; Proportionally fuse the priority weight parameter and the version change coverage range of the corresponding feature segment to generate a processing instruction containing the version difference analysis priority.

6. The method of claim 5, wherein, The feature segments of the current feature vector set and the syntax structure feature vectors of the corresponding dimensions in the historical feature vector set are matched to calculate the matching degree, the feature segments with a matching degree exceeding a preset threshold are filtered, and a feature matching mapping table is generated, including: The feature segments in the current feature vector set are independently decomposed into dimension values according to the dimensions, and the syntax structure feature vectors of the corresponding dimensions in the historical feature vector set are decomposed into historical dimension values according to the same dimensions; Calculate the difference degree of the independent dimension value and the historical dimension value, generate the difference degree score of each dimension, and screen the number of dimensions with a difference degree score lower than a preset threshold in the feature segment; when the number of dimensions exceeds a preset number, mark the corresponding feature segment as an effective matching segment; According to the identifier of the effective matching segment, match the identifier of the corresponding syntax structure feature vector, and generate a feature matching mapping table reflecting the mapping relationship between the feature segment and the syntax structure feature vector.

7. The method of claim 1, wherein, Based on the spatio-temporal correlation characteristics of the code review dataset, a review knowledge graph is constructed with the software development stage as the evolution axis, including: Convert the geographical distribution density of the review opinions in the code review dataset into a geographical density parameter, convert the time fluctuation characteristics of the defect repair cycle into a time fluctuation parameter, and decompose the domain knowledge coverage range of the reviewers into a knowledge weight vector; Taking the time sequence of the software development stage as the evolution axis, the geographical density parameter and the time fluctuation parameter are combined according to a preset weight ratio to generate a review behavior mode parameter; Map the review behavior mode parameter to an independent node, and based on the difference value change of the geographical density parameter and the correlation degree of the time fluctuation parameter between adjacent nodes, establish the spatio-temporal correlation path between the nodes; Dynamically bind the knowledge weight vector and the review behavior mode parameter to obtain the knowledge transmission link between the nodes, and generate the review knowledge graph by integrating the nodes, the spatio-temporal correlation path and the knowledge transmission link. 8.A big data-based software development system management system, characterized by comprising: It includes: An acquisition module acquires a code review dataset generated in a remote collaborative software development process, and the code review dataset includes geographical distribution density of review opinions, time fluctuation characteristics of defect repair cycle, and domain knowledge coverage range of reviewers; A construction module constructs a review knowledge graph with the software development stage as the evolution axis based on the spatio-temporal correlation characteristics of the code review dataset, and the review knowledge graph includes a review behavior mode related to the geographical distribution density and the time fluctuation characteristics; An encoding module synchronously acquires syntax structure features of current software code change records, and jointly encodes the syntax structure features and historical review tracks in the review knowledge graph to generate a processing instruction containing version difference analysis priority; A matching module dynamically matches the processing instruction with the domain knowledge coverage range to determine a distribution strategy of cross-regional review tasks, and the distribution strategy includes a quality evaluation node path-associated with the review behavior mode; A generation module generates a review management regulation strategy adapted to a software development system according to the mapping relationship among the distribution strategy, the version difference analysis priority and the quality evaluation node.

9. A computing device, comprising: It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the software development system management method based on big data in any one of claims 1-7.

10. A computer storage medium, characterized in that, The computer program is stored and is executed by a computer to realize the software development system management method based on big data according to any one of claims 1-7.

Images