Software engineering project progress intelligent monitoring system and method

Abstract

The application relates to the technical field of intelligent monitoring of project progress, in particular to an intelligent monitoring system and method for software engineering project progress, which comprises the following steps: constructing a time sequence evolution function, minimizing the deviation degree between the time sequence evolution function and a preset reference progress curve as a target to adjust the project progress, so as to obtain a first compensation parameter and an expected completion time correction value; dynamically optimizing a current task allocation scheme based on the first compensation parameter and the expected completion time correction value, so as to obtain a first task allocation scheme; allocating tasks and monitoring workloads to team members based on the first task allocation scheme, so as to obtain a personnel availability state matrix; and optimizing the first task allocation scheme based on the personnel availability state matrix, so as to obtain a second task allocation scheme. The application can effectively identify the condition deviating from the normal progress through abnormal mode detection of a progress deviation identification model, early warning of project risks and reduction of the possibility of project delay.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring technology for project progress, specifically to an intelligent monitoring system and method for software engineering project progress. Background Technology

[0002] Traditional methods typically rely on manual input or periodic reporting, resulting in untimely data updates. This lag prevents project managers from monitoring project progress in real time, potentially causing them to miss the optimal opportunity to resolve issues. Traditional methods often react reactively to identify and address schedule deviations, lacking proactive early warning mechanisms, which can easily lead to escalation of problems and ultimately impact project schedule and quality. Traditional methods are relatively static in task allocation and resource management, making it difficult to dynamically adjust based on real-time data, potentially leading to inefficient resource utilization and task execution bottlenecks. In traditional methods, personnel workload monitoring is often insufficient, failing to accurately reflect the availability of team members, which may result in uneven task allocation and affect overall team efficiency. Traditional methods lack flexible response strategies to changing working conditions, making it difficult to quickly adjust project plans and resource allocation, thus reducing the project's adaptability. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide an intelligent monitoring system and method for software engineering project progress.

[0004] The technical solution adopted to solve the above-mentioned technical problems is: a software engineering project progress intelligent monitoring method, including: The project uses an integrated multi-source acquisition terminal to collect real-time execution data of the entire lifecycle of the target software engineering project to obtain the original project data, which includes task node status, code submission frequency, resource utilization rate and collaborative interaction logs. A multidimensional feature vector is constructed based on the original project data. Anomaly pattern detection is performed on the multidimensional feature vector based on a pre-trained schedule deviation identification model to obtain a schedule risk identification sequence and a risk confidence score corresponding to the schedule risk identification sequence. Based on the schedule risk identification sequence and risk confidence score, a time-series evolution function for the target software engineering project is constructed. The project schedule is adjusted with the goal of minimizing the deviation between the time-series evolution function and the preset benchmark schedule curve, so as to obtain the first compensation parameter and the expected completion time correction value. The current task allocation scheme is dynamically optimized based on the first compensation parameter and the expected completion time correction value to obtain the first task allocation scheme. Based on the first task allocation scheme, task allocation and workload monitoring are performed on team members to obtain the personnel availability status matrix. The first task allocation scheme is optimized based on the personnel availability state matrix to obtain a second task allocation scheme. Based on the second task allocation scheme, the remaining tasks of the target software engineering project are rearranged and resources are dynamically allocated to obtain a progress correction strategy adapted to the current working conditions.

[0005] Preferably, a multidimensional feature vector is constructed based on the original data of the project, including: The task node status is divided into time windows, and the task completion rate, task delay ratio and task dependency complexity within each time window are extracted to obtain the task status feature sub-vector. The code submission frequency is statistically aggregated over a period of time to calculate the number of code submissions, code review pass rate, number of code rollbacks, and branch merging frequency per unit time, so as to obtain a code development feature sub-vector. The resource utilization rate is sampled to obtain the peak value of computing resource usage, memory usage fluctuation rate, network bandwidth utilization rate and storage resource consumption rate, so as to obtain the resource load feature sub-vector. Semantic parsing and relation extraction are performed on the collaborative interaction logs to identify the communication frequency, problem response latency, number of collaborative requests, and conflict resolution efficiency among team members, so as to obtain collaborative interaction feature sub-vectors. The task status feature vector, code development feature vector, resource load feature vector, and collaborative interaction feature vector are normalized and weighted to obtain the multidimensional feature vector.

[0006] Preferably, the progress deviation identification model adopts a deep learning architecture, including a feature embedding layer, a temporal coding layer, an anomaly detection layer, and a risk assessment layer; The feature embedding layer is used to perform dimensionality reduction mapping on the multidimensional feature vector, projecting the high-dimensional feature space to the low-dimensional latent space, and extracting the latent semantic representation of the features; The temporal coding layer uses a long short-term memory network to perform time series modeling on the output of the feature embedding layer, capturing the temporal dependencies and evolution trends during project execution. The anomaly detection layer calculates the reconstruction error between the current feature representation and the historical normal pattern based on the hidden state representation of the temporal coding layer. When the reconstruction error exceeds a preset threshold, a progress risk identifier is generated. The risk assessment layer calculates the confidence level of the schedule risk identifier and outputs the risk confidence score, which is a continuous value between 0 and 1.

[0007] Preferably, the abnormal pattern detection of the multidimensional feature vector based on the pre-trained progress deviation recognition model includes: The multidimensional feature vector is input into the progress deviation recognition model, and the feature vector of each time sampling point is linearly transformed and nonlinearly activated based on the feature embedding layer to obtain a low-dimensional feature embedding representation. Based on the temporal coding layer, the low-dimensional feature embedding representation is fused with forward and backward temporal information to obtain the hidden state representation of each time sampling point, wherein the hidden state representation includes feature information of the current time and historical time. The Euclidean distance between the hidden state representation and the reference normal mode representation is calculated based on the anomaly detection layer. When the Euclidean distance is greater than a preset distance threshold, the corresponding time sampling point is marked as an anomaly point to obtain the initial progress risk identifier. The initial schedule risk identifiers are subjected to time-series smoothing processing, and noise filtering is performed based on a sliding window voting mechanism to eliminate isolated false detection points, so as to obtain the schedule risk identifier sequence. The confidence score is calculated for each risk identifier in the schedule risk identifier sequence based on the risk assessment layer. The confidence score is obtained by weighting and fusing the magnitude of the reconstruction error, the duration of the anomaly, and the number of anomaly feature dimensions.

[0008] Preferably, the time-series evolution function of the target software engineering project is constructed based on the schedule risk identification sequence and risk confidence score, including: The sequence of schedule risk identifiers is arranged in chronological order to obtain a risk time series, wherein each risk identifier corresponds to a time sampling point; The risk confidence score is associated with each risk identifier in the risk time series to obtain a weighted risk time series, wherein the weight is the risk confidence score of the corresponding time sampling point; An initial time series evolution function is constructed based on the weighted risk time series and curve fitting method, wherein the initial time series evolution function represents the actual trajectory of project progress over time; The initial time series evolution function is smoothed using a moving average filtering method to obtain a time series evolution function with noise interference eliminated.

[0009] Preferably, the project schedule is adjusted with the objective of minimizing the deviation between the time-series evolution function and the preset baseline schedule curve, to obtain the first compensation parameter and the expected completion time correction value, including: Calculate the longitudinal difference between the time-series evolution function and the preset baseline progress curve at each time sampling point to obtain the progress deviation sequence, wherein the longitudinal difference is the difference between the actual progress completion percentage and the planned progress completion percentage; A global deviation index is calculated based on the schedule deviation sequence. A schedule adjustment optimization model is established with the minimization of the global deviation index as the optimization objective. The global deviation index is the weighted sum of squares of the schedule deviation sequence, and the weight coefficients are dynamically set according to the importance of the project stage. The schedule adjustment optimization model is solved based on the gradient descent algorithm to obtain the first compensation parameter that minimizes the global deviation index. The first compensation parameter includes the task parallelism adjustment coefficient, the resource input increment coefficient, and the working time elasticity coefficient. Substitute the first compensation parameter into the time-series evolution function for forward extrapolation to calculate the expected progress trajectory under the compensation measures, so as to obtain the expected completion time correction value.

[0010] Preferably, the current task allocation scheme is dynamically optimized based on the first compensation parameter and the expected completion time correction value to obtain a first task allocation scheme, including: The task parallelism adjustment coefficient in the first compensation parameter is analyzed to obtain the combination of tasks that can be executed in parallel in the current task allocation scheme. The logical constraints between tasks are analyzed based on the task dependency graph to obtain the set of parallelizable tasks. The scalable capacity of the current resource pool is evaluated based on the resource input increment coefficient, and the resource demand urgency of each task node is calculated based on the expected completion time correction value to obtain the task priority weight matrix. The working time adjustment range of team members is determined based on the working time flexibility coefficient, and working time flexibility constraints are established based on historical work efficiency data to obtain the feasible domain of personnel scheduling. A multi-objective optimization model is established based on the set of parallelizable tasks, the task priority weight matrix, and the feasible region for personnel scheduling. The optimization objectives are to minimize task completion time and maximize resource utilization to obtain the first task allocation scheme.

[0011] Preferably, based on the first task allocation scheme, task allocation and workload monitoring are performed on team members to obtain a personnel availability status matrix, including: Based on the task allocation results in the first task allocation scheme, extract the task assignment list of each team member to obtain the initial task allocation matrix; Based on the working time elasticity coefficient and the initial task allocation matrix, the expected workload distribution of each team member is calculated to obtain the personnel time-series load curve, wherein the personnel time-series load curve reflects the changes in the work saturation of team members in each time period. Based on the personnel time-series load curves, load balance analysis is performed to calculate the load peak, load trough and load variation coefficient of each team member in order to obtain the statistical characteristics of load distribution. The load distribution statistical characteristics are compared with the preset healthy load threshold range to obtain overload members that exceed the healthy load threshold range and underload members that are below the healthy load threshold range, so as to obtain a load anomaly identifier set. Based on the set of load anomaly identifiers, the initial task allocation matrix is ​​locally adjusted by transferring some tasks of overloaded members to underloaded members to obtain the adjusted task allocation matrix. Based on the adjusted task allocation matrix and the real-time collected collaborative interaction logs, dynamic monitoring is performed to track the task execution progress, actual working hours, and task blocking status of each team member in order to obtain a real-time execution status vector. The real-time execution state vector and the adjusted task allocation matrix are fused to construct the personnel availability state matrix, which includes personnel identifier, task identifier, current load level, skill matching degree, and available time window.

[0012] Preferably, the first task allocation scheme is optimized based on the personnel availability state matrix to obtain a second task allocation scheme, including: Feature extraction is performed on the personnel availability status matrix to identify the distribution characteristics of the current workload level, skill matching degree and available time window of each team member, so as to obtain a set of personnel capability profiles; The matching degree is calculated based on the personnel capability profile set and the task requirement features in the first task allocation scheme. The task requirement features include the required skill type, estimated workload and time constraints, so as to obtain the personnel task fit matrix. Based on the personnel task fit matrix, mismatched task assignment items with fit below a preset threshold are identified. The mismatched task assignment items include skill mismatch tasks, time conflict tasks, and overload tasks, so as to obtain a set of tasks to be optimized. Based on the set of tasks to be optimized and the set of personnel capability profiles, a task redistribution optimization model is established. The optimization objectives are to maximize the global adaptability index and minimize the task switching cost. The model is then iteratively solved to obtain candidate task redistribution schemes. The feasibility of the candidate task redistribution schemes is verified by checking whether the adjusted load levels of each team member are within the healthy load threshold range, so as to obtain a set of feasible redistribution schemes. The scheme with the best global fit index is selected from the set of feasible redistribution schemes as the second task allocation scheme.

[0013] The technical solution adopted to solve the above-mentioned technical problems is: a software engineering project progress intelligent monitoring system, which is applicable to the aforementioned software engineering project progress intelligent monitoring method, including: The data acquisition unit is used to collect the full-cycle execution data of the target software engineering project in real time based on the integrated multi-source acquisition terminal to obtain the original project data, wherein the original project data includes task node status, code submission frequency, resource utilization rate and collaborative interaction log; The progress monitoring unit is used to construct a multi-dimensional feature vector based on the original project data, and to perform abnormal pattern detection on the multi-dimensional feature vector based on a pre-trained progress deviation identification model, so as to obtain a progress risk identification sequence and a risk confidence score corresponding to the progress risk identification sequence. The schedule evolution unit is used to construct the time-series evolution function of the target software engineering project based on the schedule risk identification sequence and risk confidence score, and to adjust the project schedule with the goal of minimizing the deviation between the time-series evolution function and the preset benchmark schedule curve, so as to obtain the first compensation parameter and the expected completion time correction value. The scheme generation unit is used to dynamically optimize the current task allocation scheme based on the first compensation parameter and the expected completion time correction value to obtain the first task allocation scheme, and to allocate tasks and monitor the workload of team members based on the first task allocation scheme to obtain the personnel availability status matrix. The scheme optimization unit is used to optimize the first task allocation scheme based on the personnel availability state matrix to obtain a second task allocation scheme, and to perform task rearrangement and dynamic resource allocation on the remaining tasks of the target software engineering project based on the second task allocation scheme to obtain a progress correction strategy adapted to the current working conditions.

[0014] The beneficial effects of the present invention are as follows: (1) The present invention constructs a multi-dimensional feature vector based on the original project data, which helps to comprehensively analyze all aspects of project progress, rather than being limited to a single indicator. This comprehensive analysis can more accurately identify potential schedule risks; (2) The present invention uses a pre-trained schedule deviation identification model to detect abnormal patterns, which can effectively identify deviations from the normal schedule, provide early warning of project risks, and reduce the possibility of project delays; (3) The present invention analyzes and optimizes the personnel availability status matrix, which can better adapt to the current working conditions and adjust the schedule correction strategy in a timely manner, thereby improving the adaptability and flexibility of the project. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the overall method steps in one embodiment of the present invention; Figure 2This is a schematic diagram of the overall system flow in one embodiment of the present invention; reference numerals: 1, data acquisition unit; 2, progress monitoring unit; 3, progress evolution unit; 4, scheme generation unit; 5, scheme optimization unit. Detailed Implementation

[0016] Example 1, as Figure 1 As shown, the intelligent monitoring method for software engineering project progress proposed in this invention includes: S1. Real-time collection of execution data of the target software engineering project throughout its entire lifecycle is performed using an integrated multi-source acquisition terminal to obtain the original project data, which includes task node status, code submission frequency, resource utilization rate, and collaborative interaction logs. S2. Construct a multi-dimensional feature vector based on the original project data, and perform abnormal pattern detection on the multi-dimensional feature vector based on the pre-trained schedule deviation identification model to obtain the schedule risk identification sequence and the risk confidence score corresponding to the schedule risk identification sequence. S3. Construct a time-series evolution function for the target software engineering project based on the schedule risk identification sequence and risk confidence score. Adjust the project schedule with the goal of minimizing the deviation between the time-series evolution function and the preset baseline schedule curve to obtain the first compensation parameter and the expected completion time correction value. S4. Based on the first compensation parameter and the expected completion time correction value, the current task allocation scheme is dynamically optimized to obtain the first task allocation scheme. Based on the first task allocation scheme, task allocation and workload monitoring are performed on team members to obtain the personnel availability status matrix. S5. Optimize the first task allocation scheme based on the personnel availability state matrix to obtain the second task allocation scheme. Based on the second task allocation scheme, rearrange the remaining tasks of the target software engineering project and dynamically allocate resources to obtain a schedule correction strategy that adapts to the current working conditions.

[0017] In this invention, a multi-source acquisition terminal refers to a device or system that can simultaneously collect data from multiple sources. It is typically used for real-time monitoring and data acquisition. In software engineering projects, such a terminal can acquire various types of execution data. Full-cycle execution data refers to all execution-related data generated throughout the entire lifecycle of a software engineering project, including data from the planning, execution, monitoring, and closure phases. Anomaly pattern detection refers to using model analysis of multi-dimensional feature vectors to identify anomalies compared to normal progress and to pinpoint potential schedule risks. Task rescheduling and dynamic resource allocation refer to prioritizing and reallocating resources for remaining tasks to ensure the project is completed on time.

[0018] Example 2: The intelligent monitoring method for software engineering project progress proposed in this invention, compared to Example 1, further includes constructing a multi-dimensional feature vector based on the original project data, including: A1. Divide the task node status into time windows, and extract the task completion rate, task delay ratio and task dependency complexity within each time window to obtain the task status feature sub-vector. A2. Perform statistical periodic aggregation on code submission frequency to calculate the number of code submissions, code review pass rate, code rollback count, and branch merging frequency per unit time to obtain code development feature sub-vectors. A3. Collect sampling points for resource utilization to obtain peak computing resource usage, memory usage fluctuation rate, network bandwidth utilization, and storage resource consumption rate, so as to obtain resource load feature sub-vectors. A4. Perform semantic parsing and relation extraction on the collaborative interaction logs to identify the communication frequency, problem response latency, number of collaborative requests, and conflict resolution efficiency among team members, so as to obtain collaborative interaction feature sub-vectors. A5. Normalize and weight the task status feature vector, code development feature vector, resource load feature vector, and collaborative interaction feature vector to obtain a multidimensional feature vector.

[0019] In this embodiment, the task completion rate refers to the ratio of the number of completed tasks to the total number of tasks within a specific time window, used to measure the progress of task execution; the task status feature sub-vector describes the overall status of tasks within a certain time window; statistical period aggregation refers to summarizing and calculating data within a certain time period to obtain representative statistical results; code commit volume refers to the number of codes submitted by developers in the version control system per unit time, reflecting the frequency and intensity of development activities; branch merging frequency refers to the frequency of merging different code branches into the main branch, affecting the collaborative efficiency of the development process; the resource load feature sub-vector is a vector composed of peak computing resource usage, memory usage fluctuation rate, network bandwidth utilization rate, and storage resource consumption rate, describing the resource load. Source usage; the number of collaboration requests refers to the number of collaboration requests initiated by team members, indicating the collaboration needs within the team; the collaborative interaction feature sub-vector is a vector composed of communication frequency, problem response latency, number of collaboration requests, and conflict resolution efficiency, describing the characteristics of team collaboration; normalization processing refers to transforming data of different dimensions or ranges so that they can be compared under the same standard, usually used for data preprocessing; weight allocation refers to weighting each feature according to its importance so that the contribution of different features can be reflected in the final multidimensional feature vector; the multidimensional feature vector is a vector that integrates the task status feature sub-vector, code development feature sub-vector, resource load feature sub-vector, and collaborative interaction feature sub-vector, comprehensively reflecting multiple aspects of information.

[0020] In an optional embodiment, the schedule deviation identification model employs a deep learning architecture, including a feature embedding layer, a temporal coding layer, an anomaly detection layer, and a risk assessment layer. B1. The feature embedding layer is used to reduce the dimensionality of multidimensional feature vectors, projecting the high-dimensional feature space to the low-dimensional latent space and extracting the latent semantic representation of the features. B2. The temporal coding layer uses a long short-term memory network to perform time series modeling on the output of the feature embedding layer, capturing the temporal dependencies and evolutionary trends during project execution. B3. The anomaly detection layer calculates the reconstruction error between the current feature representation and the historical normal pattern based on the hidden state representation of the temporal coding layer. When the reconstruction error exceeds a preset threshold, a progress risk identifier is generated. B4. The risk assessment layer calculates the confidence level of the schedule risk identifier and outputs a risk confidence score. The risk confidence score ranges from 0 to 1 as a continuous value.

[0021] It should be noted that reconstruction error refers to the difference between the current feature representation reconstructed by the model and the actual feature representation in anomaly detection. A large reconstruction error usually means that the current state is significantly different from the historical normal pattern, which may indicate a risk. Risk label is a marker generated when an anomaly is detected, used to indicate that there is a risk of delay in project progress. The risk assessment layer is responsible for further analyzing the risk labels generated in project progress monitoring, calculating the confidence score of the risk, and quantifying the severity of the risk.

[0022] In an optional embodiment, abnormal pattern detection is performed on the multidimensional feature vector based on a pre-trained progress deviation recognition model, including: C1. Input the multidimensional feature vector into the progress deviation recognition model, and perform linear transformation and nonlinear activation on the feature vector of each time sampling point based on the feature embedding layer to obtain a low-dimensional feature embedding representation. C2. Based on the temporal coding layer, perform forward and backward temporal information fusion on the low-dimensional feature embedding representation to obtain the hidden state representation of each time sampling point, wherein the hidden state representation includes feature information of the current time and historical time. C3. Calculate the Euclidean distance between the hidden state representation and the reference normal mode representation based on the anomaly detection layer. When the Euclidean distance is greater than the preset distance threshold, mark the corresponding time sampling point as an anomaly point to obtain the initial progress risk identifier. C4. Perform time-series smoothing on the initial schedule risk identifiers, filter noise based on the sliding window voting mechanism, eliminate isolated false detection points, and obtain the schedule risk identifier sequence. C5. Based on the risk assessment layer, the confidence level of each risk identifier in the schedule risk identifier sequence is calculated. The confidence level calculation is based on the weighted fusion of the magnitude of the reconstruction error, the duration of the anomaly, and the number of anomaly feature dimensions to obtain the risk confidence score.

[0023] It should be noted that the initial schedule risk marker refers to the marker generated based on the Euclidean distance results during the anomaly detection process, used to indicate which points in time may pose a risk to the project schedule; time-series smoothing is a method of processing the initial schedule risk marker to reduce the impact of noise. Through a sliding window voting mechanism, risk markers at consecutive time points can be integrated, thereby filtering out isolated false detection points; the schedule risk marker sequence refers to a series of risk markers obtained after time-series smoothing, reflecting the risk status of the project schedule at different points in time.

[0024] In an optional embodiment, a time-series evolution function for the target software engineering project is constructed based on a schedule risk identification sequence and a risk confidence score, including: D1. Arrange the schedule risk identifier sequence in chronological order to obtain the risk time series, where each risk identifier corresponds to a time sampling point; D2. Associate and map the risk confidence score with each risk identifier in the risk time series to obtain a weighted risk time series, where the weight is the risk confidence score of the corresponding time sampling point; D3. Construct an initial time series evolution function based on weighted risk time series and curve fitting methods, where the initial time series evolution function represents the actual trajectory of project progress over time; D4. The initial time series evolution function is smoothed using the moving average filtering method to obtain a time series evolution function with noise interference eliminated.

[0025] It should be noted that the initial time series evolution function refers to a function constructed based on weighted risk time series and curve fitting methods, used to describe the actual trajectory of project progress changes over time. By mathematically modeling the risk time series, a continuous function representing the evolution trend of project progress can be obtained. The curve fitting method is a mathematical tool used to find an optimal curve that fits the given data points as closely as possible. In project progress monitoring, curve fitting methods can help analyze and predict changes in project progress. The moving average filtering method reduces noise interference by calculating the average value of data points within a certain time window. This method can help extract the main trend in the signal and eliminate short-term fluctuations, so as to more clearly observe the changes in the time series evolution function. The time series evolution function refers to the smoothed function, which shows the trend of project progress changes over time. After removing noise interference, it more realistically reflects the project progress.

[0026] In an optional embodiment, the project schedule is adjusted with the objective of minimizing the deviation between the time-series evolution function and the preset baseline schedule curve, to obtain a first compensation parameter and an expected completion time correction value, including: E1. Calculate the longitudinal difference between the time-series evolution function and the preset baseline schedule curve at each time sampling point to obtain the schedule deviation sequence, where the longitudinal difference is the difference between the actual progress completion percentage and the planned progress completion percentage. E2. Calculate the global deviation index based on the schedule deviation sequence, and establish a schedule adjustment optimization model with the goal of minimizing the global deviation index. The global deviation index is the weighted sum of squares of the schedule deviation sequence, and the weight coefficients are dynamically set according to the importance of the project stage. E3. Solve the schedule adjustment optimization model based on the gradient descent algorithm to obtain the first compensation parameter that minimizes the global deviation index. The first compensation parameter includes the task parallelism adjustment coefficient, the resource input increment coefficient, and the working time elasticity coefficient. E4. Substitute the first compensation parameter into the time evolution function for forward extrapolation, calculate the expected progress trajectory under the action of compensation measures, and obtain the expected completion time correction value.

[0027] It should be noted that the schedule deviation sequence refers to the sequence obtained by calculating the longitudinal difference between the project schedule intelligent monitoring time series evolution function (actual progress) and the preset baseline schedule curve (planned progress) at each time sampling point. This difference reflects the difference between the actual progress completion percentage and the planned progress completion percentage, which can help identify whether the project progress meets expectations. The longitudinal difference refers to the difference between the actual progress completion percentage and the planned progress completion percentage at a specific time point. A positive value indicates that the project progress is ahead of the plan, and a negative value indicates that the project progress is behind the plan. The global deviation index is measured by weighted root mean square error. The weight coefficient is set exponentially according to the time distance of each time sampling point from the current time. The closer the sampling point is to the current time, the higher the weight. The smaller the global deviation index, the closer the project progress is to the plan and the smaller the deviation. Forward extrapolation refers to substituting the adjusted first compensation parameter into the time series evolution function for calculation to predict the change in project progress after the implementation of compensation measures. The expected completion time correction value is the correction value of the expected completion time of the project calculated after considering compensation measures (such as adjustments to task parallelism, resource input, and working time).

[0028] In an optional embodiment, the current task allocation scheme is dynamically optimized based on a first compensation parameter and an expected completion time correction value to obtain a first task allocation scheme, including: F1. Analyze the task parallelism adjustment coefficient in the first compensation parameter to obtain the combination of tasks that can be executed in parallel in the current task allocation scheme. Analyze the logical constraints between tasks based on the task dependency graph to obtain the set of parallelizable tasks. F2. Evaluate the scalable capacity of the current resource pool based on the incremental coefficient of resource input, and calculate the urgency of resource demand for each task node based on the expected completion time correction value to obtain the task priority weight matrix. F3. Determine the range of adjustment for team members' working hours based on the working time flexibility coefficient, and establish working time flexibility constraints based on historical work efficiency data to obtain the feasible domain for personnel scheduling. F4. Based on the set of parallelizable tasks, the task priority weight matrix, and the feasible region for personnel scheduling, a multi-objective optimization model is established. The optimization objectives are to minimize task completion time and maximize resource utilization to obtain the first task allocation scheme. It should be noted that the task parallelism adjustment coefficient, used in project schedule adjustments, represents the number or proportion of tasks that can be executed in parallel within the current task allocation scheme. This coefficient helps identify which tasks can be performed simultaneously, thereby improving overall project efficiency. The resource input increment coefficient assesses the quantity or extent of additional resources (such as manpower, equipment, and funds) added to the project. This coefficient helps determine the scalability of the current resource pool to meet project needs. The resource demand urgency refers to the degree of urgency of the resources required for each task node, calculated based on the expected completion time adjustment value. The more urgent the task demand, the higher its resource requirement, which is crucial for determining task priority. The working time flexibility coefficient determines the team members' ability to adjust their working hours, indicating the flexibility of team members. The working time range is a coefficient that helps to rationally arrange personnel working hours when adjusting project plans. The personnel scheduling feasible region refers to the personnel scheduling range obtained by combining the working time elasticity coefficient and the working time elasticity constraint. It represents the feasible space for the team members' shift scheduling and working hours arrangement under all constraints. The multi-objective optimization model is an optimization model established based on the set of parallelizable tasks, the task priority weight matrix, and the personnel scheduling feasible region. This model aims to optimize multiple objectives simultaneously, such as minimizing task completion time and improving resource utilization. By solving this model, the optimal task allocation scheme can be found. The first task allocation scheme refers to the task allocation scheme obtained by solving the multi-objective optimization model. This scheme specifies the specific task allocation, personnel arrangement, and their execution order to achieve the optimization objective.

[0029] In an optional embodiment, task allocation and workload monitoring are performed on team members based on a first task allocation scheme to obtain a personnel availability status matrix, including: G1. Based on the task allocation results in the first task allocation scheme, extract the task assignment list of each team member to obtain the initial task allocation matrix; G2. Calculate the expected workload distribution of each team member based on the working time elasticity coefficient and the initial task allocation matrix to obtain the personnel time-series load curve, which reflects the changes in the work saturation of team members in each time period. G3. Based on the personnel time-series load curve, perform load balance analysis, calculate the load peak, load trough and load variation coefficient of each team member, so as to obtain the statistical characteristics of load distribution; G4. Compare the load distribution statistical characteristics with the preset healthy load threshold range to obtain overload members that exceed the healthy load threshold range and underload members that are below the healthy load threshold range, so as to obtain a load anomaly identifier set. G5. Based on the load anomaly identifier set, the initial task allocation matrix is ​​locally adjusted, and some tasks of the overloaded members are transferred to the underloaded members to obtain the adjusted task allocation matrix. G6. Based on the adjusted task allocation matrix and real-time collected collaborative interaction logs, dynamic monitoring is performed to track the task execution progress, actual working hours and task blocking status of each team member in order to obtain a real-time execution status vector. G7. The real-time execution state vector and the adjusted task allocation matrix are merged to construct a personnel availability state matrix that includes personnel identification, task identification, current load level, skill matching degree, and available time window.

[0030] It should be noted that the expected workload distribution refers to the workload of each team member calculated based on the work time elasticity coefficient and the initial task allocation matrix. The expected workload reflects the amount of work that each member should undertake within a specific time period. The personnel time-series load curve is a graph obtained by calculating the changes in the workload saturation of each team member in each time period. This curve shows the changes in the work intensity and load of team members in different time periods, which can help identify overwork or underwork. The adjusted task allocation matrix refers to the task allocation matrix after local adjustments, reflecting the new allocation status of tasks among team members, which can optimize workload and improve work efficiency. The real-time execution status vector is a vector obtained based on the adjusted task allocation matrix and the real-time collected collaborative interaction logs. It contains information such as the current task execution progress, actual working hours, and task blocking status of each team member. This vector helps to understand the implementation effect of the task.

[0031] In an optional embodiment, the first task allocation scheme is optimized based on the personnel availability state matrix to obtain a second task allocation scheme, including: H1. Extract features from the personnel availability status matrix to identify the distribution characteristics of the current load level, skill matching degree and available time window of each team member, so as to obtain a set of personnel capability profiles. H2. Based on the personnel capability profile set and the task requirement features in the first task allocation scheme, the matching degree is calculated. The task requirement features include the required skill type, estimated workload and time constraints, so as to obtain the personnel task fit matrix. H3. Identify mismatched task assignments with a matching degree lower than a preset threshold based on the personnel task matching degree matrix. Mismatched task assignments include skill mismatch tasks, time conflict tasks, and overload tasks, so as to obtain a set of tasks to be optimized. H4. Based on the set of tasks to be optimized and the set of personnel capability profiles, establish a task redistribution optimization model. With the optimization objectives of maximizing the global adaptability index and minimizing the task switching cost, iteratively solve the model to obtain candidate task redistribution schemes. H5. Verify the feasibility of candidate task redistribution schemes, and check whether the adjusted load levels of each team member are within the healthy load threshold range, so as to obtain a set of feasible redistribution schemes. H6. Select the scheme with the best global fit index from the set of feasible redistribution schemes as the second task allocation scheme.

[0032] It should be noted that the personnel capability profile set refers to a collection formed based on the results of feature extraction, recording the capability characteristics of each team member. These profiles showcase the member's work ability, skill expertise, and available working hours. Matching degree calculation refers to comparing the personnel capability profile set with the task requirement features to calculate the suitability of each team member for each task, forming a personnel-task suitability matrix. This matrix helps identify which tasks and which members have the best match. The task set to be optimized refers to the set of mismatched task assignments. These tasks need to be reassigned or optimized to improve the overall task suitability. Candidate task reassignment schemes refer to a series of possible task reassignment schemes obtained through iterative solutions. These schemes aim to optimize task allocation to achieve the best suitability and the lowest switching cost. The scheme with the best global suitability index is the best scheme selected from the set of feasible reassignment schemes. This scheme performs best in terms of suitability index and serves as the final second task assignment scheme. This scheme will be applied to subsequent task execution to improve the overall team's work efficiency.

[0033] Example 3, as Figure 2 As shown, the intelligent monitoring system for software engineering project progress proposed in this invention, and its applicable intelligent monitoring method for software engineering project progress, include: Data acquisition unit 1 is used to collect the full life cycle execution data of the target software engineering project in real time based on the integrated multi-source acquisition terminal to obtain the original project data, which includes task node status, code submission frequency, resource utilization rate and collaborative interaction logs. The progress monitoring unit 2 is used to construct a multi-dimensional feature vector based on the original project data, and to perform abnormal pattern detection on the multi-dimensional feature vector based on the pre-trained progress deviation identification model, so as to obtain the progress risk identification sequence and the risk confidence score corresponding to the progress risk identification sequence. Schedule evolution unit 3 is used to construct the time-series evolution function of the target software engineering project based on the schedule risk identification sequence and risk confidence score. The project schedule is adjusted with the goal of minimizing the deviation between the time-series evolution function and the preset baseline schedule curve, so as to obtain the first compensation parameter and the expected completion time correction value. The scheme generation unit 4 is used to dynamically optimize the current task allocation scheme based on the first compensation parameter and the expected completion time correction value to obtain the first task allocation scheme. Based on the first task allocation scheme, the team members are assigned tasks and their workload is monitored to obtain the personnel availability status matrix. The scheme optimization unit 5 is used to optimize the first task allocation scheme based on the personnel availability state matrix to obtain the second task allocation scheme. Based on the second task allocation scheme, the remaining tasks of the target software engineering project are rearranged and resources are dynamically allocated to obtain a schedule correction strategy that adapts to the current working conditions.

[0034] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A method for intelligent monitoring of software engineering project progress, characterized in that, include: The project uses an integrated multi-source acquisition terminal to collect real-time execution data of the entire lifecycle of the target software engineering project to obtain the original project data, which includes task node status, code submission frequency, resource utilization rate and collaborative interaction logs. A multidimensional feature vector is constructed based on the original project data. Anomaly pattern detection is performed on the multidimensional feature vector based on a pre-trained schedule deviation identification model to obtain a schedule risk identification sequence and a risk confidence score corresponding to the schedule risk identification sequence. Based on the schedule risk identification sequence and risk confidence score, a time-series evolution function for the target software engineering project is constructed. The project schedule is adjusted with the goal of minimizing the deviation between the time-series evolution function and the preset benchmark schedule curve, so as to obtain the first compensation parameter and the expected completion time correction value. The current task allocation scheme is dynamically optimized based on the first compensation parameter and the expected completion time correction value to obtain the first task allocation scheme. Based on the first task allocation scheme, task allocation and workload monitoring are performed on team members to obtain the personnel availability status matrix. The first task allocation scheme is optimized based on the personnel availability state matrix to obtain a second task allocation scheme. Based on the second task allocation scheme, the remaining tasks of the target software engineering project are rearranged and resources are dynamically allocated to obtain a progress correction strategy adapted to the current working conditions.

2. The intelligent monitoring method for software engineering project progress according to claim 1, characterized in that, A multidimensional feature vector is constructed based on the original data of the project, including: The task node status is divided into time windows, and the task completion rate, task delay ratio and task dependency complexity within each time window are extracted to obtain the task status feature sub-vector. The code submission frequency is statistically aggregated over a period of time to calculate the number of code submissions, code review pass rate, number of code rollbacks, and branch merging frequency per unit time, so as to obtain a code development feature sub-vector. The resource utilization rate is sampled to obtain the peak value of computing resource usage, memory usage fluctuation rate, network bandwidth utilization rate and storage resource consumption rate, so as to obtain the resource load feature sub-vector. Semantic parsing and relation extraction are performed on the collaborative interaction logs to identify the communication frequency, problem response latency, number of collaborative requests, and conflict resolution efficiency among team members, so as to obtain collaborative interaction feature sub-vectors. The task status feature vector, code development feature vector, resource load feature vector, and collaborative interaction feature vector are normalized and weighted to obtain the multidimensional feature vector.

3. The intelligent monitoring method for software engineering project progress according to claim 2, characterized in that, The progress deviation identification model adopts a deep learning architecture. It includes a feature embedding layer, a temporal coding layer, an anomaly detection layer, and a risk assessment layer; The feature embedding layer is used to perform dimensionality reduction mapping on the multidimensional feature vector, projecting the high-dimensional feature space to the low-dimensional latent space, and extracting the latent semantic representation of the features; The temporal coding layer uses a long short-term memory network to perform time series modeling on the output of the feature embedding layer, capturing the temporal dependencies and evolution trends during project execution. The anomaly detection layer calculates the reconstruction error between the current feature representation and the historical normal pattern based on the hidden state representation of the temporal coding layer. When the reconstruction error exceeds a preset threshold, a progress risk identifier is generated. The risk assessment layer calculates the confidence level of the schedule risk identifier and outputs the risk confidence score, which is a continuous value between 0 and 1.

4. The intelligent monitoring method for software engineering project progress according to claim 3, characterized in that, Anomaly pattern detection is performed on the multidimensional feature vector based on a pre-trained progress deviation recognition model, including: The multidimensional feature vector is input into the progress deviation recognition model, and the feature vector of each time sampling point is linearly transformed and nonlinearly activated based on the feature embedding layer to obtain a low-dimensional feature embedding representation. Based on the temporal coding layer, the low-dimensional feature embedding representation is fused with forward and backward temporal information to obtain the hidden state representation of each time sampling point, wherein the hidden state representation includes feature information of the current time and historical time. The Euclidean distance between the hidden state representation and the reference normal mode representation is calculated based on the anomaly detection layer. When the Euclidean distance is greater than a preset distance threshold, the corresponding time sampling point is marked as an anomaly point to obtain the initial progress risk identifier. The initial schedule risk identifiers are subjected to time-series smoothing processing, and noise filtering is performed based on a sliding window voting mechanism to eliminate isolated false detection points, so as to obtain the schedule risk identifier sequence. The confidence score is calculated for each risk identifier in the schedule risk identifier sequence based on the risk assessment layer. The confidence score is obtained by weighting and fusing the magnitude of the reconstruction error, the duration of the anomaly, and the number of anomaly feature dimensions.

5. The intelligent monitoring method for software engineering project progress according to claim 4, characterized in that, Based on the aforementioned schedule risk identification sequence and risk confidence score, a time-series evolution function for the target software engineering project is constructed, including: The sequence of schedule risk identifiers is arranged in chronological order to obtain a risk time series, wherein each risk identifier corresponds to a time sampling point; The risk confidence score is associated with each risk identifier in the risk time series to obtain a weighted risk time series, wherein the weight is the risk confidence score of the corresponding time sampling point; An initial time series evolution function is constructed based on the weighted risk time series and curve fitting method, wherein the initial time series evolution function represents the actual trajectory of project progress over time; The initial time series evolution function is smoothed using a moving average filtering method to obtain a time series evolution function with noise interference eliminated.

6. The intelligent monitoring method for software engineering project progress according to claim 5, characterized in that, The project schedule is adjusted with the objective of minimizing the deviation between the time-series evolution function and the preset baseline schedule curve, to obtain the first compensation parameter and the expected completion time correction value, including: Calculate the longitudinal difference between the time-series evolution function and the preset baseline progress curve at each time sampling point to obtain the progress deviation sequence, wherein the longitudinal difference is the difference between the actual progress completion percentage and the planned progress completion percentage; A global deviation index is calculated based on the schedule deviation sequence. A schedule adjustment optimization model is established with the minimization of the global deviation index as the optimization objective. The global deviation index is the weighted sum of squares of the schedule deviation sequence, and the weight coefficients are dynamically set according to the importance of the project stage. The schedule adjustment optimization model is solved based on the gradient descent algorithm to obtain the first compensation parameter that minimizes the global deviation index. The first compensation parameter includes the task parallelism adjustment coefficient, the resource input increment coefficient, and the working time elasticity coefficient. Substitute the first compensation parameter into the time-series evolution function for forward extrapolation to calculate the expected progress trajectory under the compensation measures, so as to obtain the expected completion time correction value.

7. The intelligent monitoring method for software engineering project progress according to claim 6, characterized in that, The current task allocation scheme is dynamically optimized based on the first compensation parameter and the expected completion time correction value to obtain a first task allocation scheme, including: The task parallelism adjustment coefficient in the first compensation parameter is analyzed to obtain the combination of tasks that can be executed in parallel in the current task allocation scheme. The logical constraints between tasks are analyzed based on the task dependency graph to obtain the set of parallelizable tasks. The scalable capacity of the current resource pool is evaluated based on the resource input increment coefficient, and the resource demand urgency of each task node is calculated based on the expected completion time correction value to obtain the task priority weight matrix. The working time adjustment range of team members is determined based on the working time flexibility coefficient, and working time flexibility constraints are established based on historical work efficiency data to obtain the feasible domain of personnel scheduling. A multi-objective optimization model is established based on the set of parallelizable tasks, the task priority weight matrix, and the feasible region for personnel scheduling. The optimization objectives are to minimize task completion time and maximize resource utilization to obtain the first task allocation scheme.

8. The intelligent monitoring method for software engineering project progress according to claim 7, characterized in that, Based on the first task allocation scheme, tasks are allocated to team members and workload is monitored to obtain a personnel availability status matrix, including: Based on the task allocation results in the first task allocation scheme, extract the task assignment list of each team member to obtain the initial task allocation matrix; Based on the working time elasticity coefficient and the initial task allocation matrix, the expected workload distribution of each team member is calculated to obtain the personnel time-series load curve, wherein the personnel time-series load curve reflects the changes in the work saturation of team members in each time period. Based on the personnel time-series load curves, load balance analysis is performed to calculate the load peak, load trough and load variation coefficient of each team member in order to obtain the statistical characteristics of load distribution. The load distribution statistical characteristics are compared with the preset healthy load threshold range to obtain overload members that exceed the healthy load threshold range and underload members that are below the healthy load threshold range, so as to obtain a load anomaly identifier set. Based on the load anomaly identifier set, the initial task allocation matrix is ​​locally adjusted by transferring some tasks of overloaded members to underloaded members to obtain the adjusted task allocation matrix; Based on the adjusted task allocation matrix and the real-time collected collaborative interaction logs, dynamic monitoring is performed to track the task execution progress, actual working hours, and task blocking status of each team member in order to obtain a real-time execution status vector. The real-time execution state vector and the adjusted task allocation matrix are fused to construct the personnel availability state matrix, which includes personnel identifier, task identifier, current load level, skill matching degree, and available time window.

9. The intelligent monitoring method for software engineering project progress according to claim 8, characterized in that, The first task allocation scheme is optimized based on the personnel availability state matrix to obtain a second task allocation scheme, including: Feature extraction is performed on the personnel availability status matrix to identify the distribution characteristics of the current workload level, skill matching degree and available time window of each team member, so as to obtain a set of personnel capability profiles; The matching degree is calculated based on the personnel capability profile set and the task requirement features in the first task allocation scheme. The task requirement features include the required skill type, estimated workload and time constraints, so as to obtain the personnel task fit matrix. Based on the personnel task fit matrix, mismatched task assignment items with fit below a preset threshold are identified. The mismatched task assignment items include skill mismatch tasks, time conflict tasks, and overload tasks, so as to obtain a set of tasks to be optimized. Based on the set of tasks to be optimized and the set of personnel capability profiles, a task redistribution optimization model is established. The optimization objectives are to maximize the global adaptability index and minimize the task switching cost. The model is then iteratively solved to obtain candidate task redistribution schemes. The feasibility of the candidate task redistribution schemes is verified by checking whether the adjusted load levels of each team member are within the healthy load threshold range, so as to obtain a set of feasible redistribution schemes. The scheme with the best global fit index is selected from the set of feasible redistribution schemes as the second task allocation scheme.

10. A software engineering project progress intelligent monitoring system, applicable to the software engineering project progress intelligent monitoring method according to any one of claims 1-9, characterized in that, include: The data acquisition unit is used to collect the full-cycle execution data of the target software engineering project in real time based on the integrated multi-source acquisition terminal to obtain the original project data, wherein the original project data includes task node status, code submission frequency, resource utilization rate and collaborative interaction log; The progress monitoring unit is used to construct a multi-dimensional feature vector based on the original project data, and to perform abnormal pattern detection on the multi-dimensional feature vector based on a pre-trained progress deviation identification model, so as to obtain a progress risk identification sequence and a risk confidence score corresponding to the progress risk identification sequence. The schedule evolution unit is used to construct the time-series evolution function of the target software engineering project based on the schedule risk identification sequence and risk confidence score, and to adjust the project schedule with the goal of minimizing the deviation between the time-series evolution function and the preset benchmark schedule curve, so as to obtain the first compensation parameter and the expected completion time correction value. The scheme generation unit is used to dynamically optimize the current task allocation scheme based on the first compensation parameter and the expected completion time correction value to obtain the first task allocation scheme, and to allocate tasks and monitor the workload of team members based on the first task allocation scheme to obtain the personnel availability status matrix. The scheme optimization unit is used to optimize the first task allocation scheme based on the personnel availability state matrix to obtain a second task allocation scheme, and to perform task rearrangement and dynamic resource allocation on the remaining tasks of the target software engineering project based on the second task allocation scheme to obtain a progress correction strategy adapted to the current working conditions.

Images