Building engineering human resource intelligent management optimization method based on big data analysis

Abstract

The invention discloses a building engineering human resource intelligent management optimization method based on big data analysis, and the method comprises the following steps: S1, collecting building engineering data, and carrying out the preprocessing of the data; s2, fusing multi-modal data, and synchronously generating time-space correlation optimization data; s3, constructing a construction site multi-scale digital twinborn model; s4, constructing a scheduling decision system, and simulating work type, task and resource interaction; s5, optimizing a personnel scheduling configuration strategy in combination with time constraint and skill matching; s6, performing Bayesian reasoning risk assessment, and generating a risk causal analysis report; and S7, the information is fed back to a manager through the multi-terminal display platform. Through big data analysis, digital twinborn modeling, reinforcement learning optimization and adaptive feedback control, intelligent scheduling and dynamic optimization of constructional engineering human resources are realized, construction efficiency is improved, resource waste is reduced, and construction safety and management intelligence level are enhanced.

Description

Technical Field

[0001] The present invention relates to the field of big data technology, and particularly to an intelligent management optimization method for construction project human resources based on big data analysis. Background Art

[0002] In the field of construction projects, human resource management is one of the important factors affecting construction progress, quality and cost. With the expansion of the scale of construction projects and the complexity of the construction environment, traditional human resource scheduling methods are difficult to effectively cope with the dynamic changes during the construction process, resulting in unreasonable human resource allocation, decreased construction efficiency and increased construction costs. The existing human resource management mainly relies on manual experience and rule-making. Personnel arrangements are made by project managers or schedulers based on historical experience. This method still has a certain degree of feasibility in small-scale projects. However, in large and complex construction projects, due to the large personnel mobility, complex task dependencies and rapid changes in the construction environment, the manual management method often has problems such as slow response, resource waste and increased construction risks. In addition, with the progress of intelligent and digital technologies, some construction enterprises have begun to try to introduce information means for human resource management, such as using ERP systems, project management software, etc. to digitally manage construction tasks and personnel scheduling. However, these systems mainly rely on static rule setting, cannot achieve real-time adjustment during the construction process, lack intelligent scheduling optimization capabilities, and make it still difficult for the allocation of human resources to meet the dynamic construction requirements.

[0003] In response to these problems, in recent years, some studies have begun to explore human resource optimization methods based on data analysis and artificial intelligence technologies. For example, some studies adopt optimization methods based on linear programming to calculate the optimal allocation plan of human resources through mathematical modeling. However, such methods usually rely on strict constraint conditions and are difficult to adapt to the frequently changing on-site situations during the construction process. In addition, human resource optimization methods based on heuristic algorithms (such as genetic algorithms, ant colony optimization algorithms, etc.) have improved the flexibility of scheduling to a certain extent. However, due to the complexity of construction tasks and the uncertainty of the environment, these methods have a high computational complexity when dealing with high-dimensional and multi-constrained construction scheduling problems and are difficult to respond to emergencies during the construction process in real time. At the same time, some studies have tried to combine artificial intelligence technologies and use deep learning and reinforcement learning models to optimize human resource allocation, such as using deep neural networks to predict construction progress and combining reinforcement learning for personnel scheduling optimization. However, these methods usually require a large amount of historical data for training, and due to the particularity of the construction industry, the data between different projects often lacks consistency, making these data-driven methods have poor migration in practical applications. In addition, most existing intelligent scheduling systems operate independently, lack the comprehensive perception ability of the construction site, and are difficult to perform dynamic optimization by combining real-time feedback during the construction process.

[0004] The existing technologies mainly have the following defects in the human resource management of construction projects: First, the traditional human resource scheduling methods rely on manual experience and lack intelligent optimization means, making it difficult to adapt to the dynamic changes in complex construction environments. Second, the project management systems based on static rules cannot achieve real-time adjustment during the construction process, resulting in a lack of flexibility in the human resource allocation plan. In addition, although the scheduling methods based on mathematical modeling and heuristic algorithms have improved the optimization efficiency to a certain extent, they still have problems such as high computational complexity and slow response speed when dealing with complex construction scheduling problems. Although the scheduling optimization methods based on deep learning and reinforcement learning have strong learning capabilities, due to the inconsistency of data in the construction industry, these methods have poor applicability in different projects. In addition, the existing human resource management systems lack an effective feedback mechanism and cannot optimize the scheduling by combining the real-time data on the construction site, making it difficult for the scheduling plan to adapt to unexpected situations during the construction process.

[0005] Therefore, how to provide an intelligent management optimization method for the human resources of construction projects based on big data analysis is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] An object of the present invention is to propose an intelligent management optimization method for the human resources of construction projects based on big data analysis. The present invention makes full use of multi-modal data analysis, digital twin technology, reinforcement learning optimization, and adaptive feedback control mechanism. Through big data collection, analysis, and optimization, an efficient and intelligent human resource management system is constructed, aiming to solve the limitations of the existing technologies in human resource scheduling optimization. The present invention details how to dynamically adjust the human resource scheduling plan based on the real-time construction environment, conduct intelligent feedback through a multi-terminal display platform, and adopt an adaptive optimization mechanism to continuously improve the accuracy of the scheduling plan, realizing the efficient matching of construction tasks and personnel allocation.

[0007] The present invention constructs a multi-scale digital twin model of the construction site, uses the reinforcement learning algorithm to optimize the human resource scheduling strategy, and combines the Bayesian risk assessment model to identify potential construction risks, ensuring the accuracy of personnel scheduling and construction safety. In addition, the present invention adopts a multi-agent cooperative game mechanism in the scheduling optimization process, improving the global optimization ability of the scheduling plan, and through the adaptive feedback control mechanism, combining the real-time feedback information of the management personnel, continuously optimizing the scheduling decision-making to make it more flexible and adaptable.

[0008] The present invention has the following advantages: First, through the big data-driven human resource optimization method, it can effectively reduce construction resource waste, improve the utilization rate of human resources, and reduce construction costs. Second, through digital twin technology and dynamic environment perception mechanism, the human resource scheduling plan can be adaptively adjusted to ensure that the human resource allocation at the construction site is always in the optimal state. In addition, the introduction of reinforcement learning and multi-agent game optimization methods enables the scheduling system not only to optimize the current construction tasks, but also to make predictive adjustments based on historical data, improving construction efficiency. Finally, the multi-terminal display platform and intelligent feedback mechanism adopted by the present invention enable managers to intuitively obtain scheduling optimization suggestions and perform manual intervention, greatly improving the intelligent level of human resource scheduling management and enhancing the management ability and decision-making efficiency of construction enterprises in complex engineering projects.

[0009] According to the intelligent management optimization method for construction project human resources based on big data analysis in an embodiment of the present invention, the following steps are included: S1. Collect data on the construction project site and preprocess the data, including data denoising, feature enhancement, and anomaly detection; S2. Perform multi-modal data fusion on the preprocessed data, synchronize the data, and generate spatio-temporal correlation optimization data; S3. Based on the data on the construction project site and the spatio-temporal correlation optimization data, construct a digital twin model of the construction site, perform dynamic modeling using graph neural networks, and update the twin environment in combination with real-time data; S4. On the basis of the constructed digital twin model of the construction site, construct a human resource scheduling decision-making system, simulate the interaction relationships among different types of work, construction tasks, and resource constraints, and establish an adaptive reward mechanism; S5. Use the human resource scheduling decision-making system to adjust the personnel scheduling plan in combination with task time window constraints, worker skill matching degrees, and real-time construction progress, and generate a human resource allocation strategy; S6. Adopt multi-level Bayesian inference to conduct risk assessment on the human resource allocation strategy, construct a risk early warning system, predict efficiency decline, safety hazards, and resource waste, and generate a risk causal analysis report; S7. Feed back the human resource allocation strategy and risk assessment results to the manager through a multi-terminal display platform, and optimize the human resource scheduling decision-making system based on an adaptive feedback control mechanism.

[0010] Optionally, the specific content of S3 includes: S31. Obtain the data on the construction project site and the spatio-temporal correlation optimization data. The data on the construction project site includes personnel location data, operation status data, construction environment data, and equipment operation data; S32. Construct a multi-scale digital twin model for the construction site, adopting a hierarchical modeling method, including a global construction progress model at the macro level, a task allocation model at the meso level, and a personnel behavior simulation model at the micro level, respectively modeling the construction progress, task arrangement, and individual behavior, and defining the information transfer mechanism between different levels; S33. Construct a personnel-task-environment relationship matrix : ; Among them, represents the personnel skill matching degree matrix, represents the task urgency matrix, represents the geographical distance matrix; S34. Use a dynamic graph neural network to model the personnel-task-environment relationship matrix and construct a construction personnel status update model: ; Among them, represents the hidden state matrix of the l-th layer, represents the adjacency matrix under different topological scales, represents the trainable weight matrix, B represents the bias term, represents the weighted coefficient of different scale topological structures, represents the hidden state matrix of the l-th layer, represents the number of different topological scales, represents the non-linear activation function, S35. Combine the modeling results to perform multi-step prediction on the construction personnel scheduling plan, and use a spatio-temporal graph convolutional variational auto-regression model to calculate the changing trends of construction progress and personnel demand: ; Among them, represents the completion degree of construction tasks at time t, and represent the regression coefficients, p represents the time lag step, q represents the latent variable influence duration, represents the completion degree of construction tasks, represents the latent variable, represents the error term; i and j represent indices; S36. Based on the prediction results, use the Bayesian optimization method to update the multi-scale digital twin model of the construction site and construct an objective function : ; Among them, represents the actual observed task completion degree, represents the predicted value, denotes the neural network weight matrix, denotes the regularization coefficient, denotes the time window length, and t denotes the index.

[0011] Optionally, step S4 specifically includes: S41. Based on the constructed multi-scale digital twin model of the construction site, extract the personnel-task-environment relationship matrix and construction status information, and use the dynamic relationship weighting method to construct an optimization matrix : ; wherein, denotes the personnel-task-environment relationship matrix, denotes the construction environment status matrix, denotes the relationship change rate, denotes the weighting coefficient; S42. Based on the optimization matrix, construct a human resource scheduling decision system, use graph reinforcement learning to model the task allocation process of construction workers, and use a multi-scale attention mechanism to define the task matching score: ; wherein, is the allocation score between construction worker i and task j, is the trainable weight, is the element-wise multiplication, is the activation function, is the bias term, is the feature vector of construction worker i, is the feature vector of task j; S43. Use the calculated task matching score to construct a reinforcement learning model based on the graph attention mechanism, and define the scheduling policy update rule: ; wherein, is the scheduling policy at time step t, represents the task allocation probability calculated based on the optimization matrix G, is the learning rate, is the logarithmic function, is the scheduling policy at time step t+1, is the logarithmic gradient of the task allocation probability; S44. Adopt an optimal scheduling game model based on adversarial training, and define the multi-objective optimization function for the allocation of construction workers: ; wherein, is the comprehensive reward at the current time step, is the scheduling smoothness control coefficient, represents the optimized human resource scheduling plan, represents the scheduling plan at time step t, represents the scheduling plan at time step t−1, represents the time window length, Solve the personnel-task allocation matrix A that maximizes the objective function; S45. Using the optimized human resource allocation plan, adopt a hybrid adaptive adjustment mechanism to fine-tune the personnel scheduling plan, and calculate the human resource scheduling matrix based on dynamic task requirements: ; where, is the human resource scheduling matrix, is the task demand gradient, is the adjustment step size, represents the gradient adjustment direction.

[0012] Optionally, the S5 specifically includes: S51. Based on the generated personnel scheduling plan and the multi-scale digital twin model of the construction site, extract historical scheduling data, construction progress status, and worker skill matching information, and construct a multi-modal task matching matrix Z: ; where, is the personnel scheduling plan, is the personnel-task-environment relationship matrix, is the construction environment status matrix, is the task demand gradient regarding the scheduling matrix, ( ) is the multi-modal fusion function; S52. Based on the constructed task matching matrix, adopt a reinforcement learning model optimized by a graph variational autoencoder to perform multi-step reasoning on the construction task allocation plan, and calculate the scheduling optimization objective using the task state transition function: ; where, is the scheduling strategy, is the comprehensive reward at the current time step, represents the structural change distance of the task matching matrix between consecutive time steps, S is the number of scheduling decision iteration steps, is the scheduling smoothness control parameter, is the expected symbol, is the time window length, is the task matching matrix at time step s, is the task matching matrix at time step s−1, represents solving the scheduling strategy that maximizes the objective function ; S53. Based on the generated scheduling strategy, combined with historical task execution data and real-time construction environment information, adopt a multi-agent adaptive attention mechanism for dynamic scheduling optimization, and define the update rule of the human resource scheduling matrix: ; Among them, is the optimized human resource scheduling matrix, is the initial scheduling matrix calculated, is the scheduling stability control parameter, is the task gradient optimization coefficient, is the scheduling strategy distribution, is the policy parameter, is the scheduling strategy, is the number of scheduling schemes; S54. Based on the optimized human resource scheduling matrix, adopt an adaptive adjustment mechanism based on gradient manifold mapping for dynamic optimization: ; Among them, is the gradient manifold mapping function, is the scheduling convergence weight, is the human resource allocation strategy, is the task demand objective function; S55. Based on the human resource allocation strategy, optimize the long-term scheduling stability and global optimality of human resource management.

[0013] Optionally, the said S6 specifically includes: S61. Based on the generated human resource allocation strategy and the multi-scale digital twin model of the construction site, extract historical scheduling data, construction progress status and personnel task execution situation, construct the task execution status matrix Q, and adopt an adaptive multi-scale spatio-temporal embedding method for state representation optimization: ; Among them, is the personnel-task-environment relationship matrix, is the construction environment status matrix, is the task demand gradient regarding the scheduling matrix, ( ) is the dynamic state embedding function, is the spatio-temporal correlation weight, is the influence function of the historical state, is the human resource allocation strategy, is the historical window size; S62. Based on the constructed task execution status matrix, adopt a risk assessment model based on Bayesian dynamic inference to perform multi-scale risk prediction on the human resource allocation strategy, and optimize the calculation of the uncertainty of task execution by combining multi-objective confidence intervals: ; Among them, is the posterior probability distribution calculated based on the task execution status matrix , is the construction task risk scoring function, is the risk variable space, is the uncertainty confidence interval calculated during the task execution process, is the uncertainty control parameter, is the construction task risk variable, is the total number of uncertainty factors of the construction task, is the risk score; S63. Based on the calculated risk score, use a generative adversarial network to generate potential high-risk scenarios, and optimize the calculation of the risk impact degree based on graph adversarial games: ; Among them, is the optimal risk generator, represents the change distance of the task execution status at consecutive time steps, is the risk stability control parameter, is the risk game loss calculated by the graph game strategy, is the game balance coefficient, is the expectation operator, The optimization objective is to find the that makes the performance of the risk assessment model optimal, is the current task execution status matrix, is the status matrix at the previous moment, is the high-risk scenario, is the actual task status change; S64. Based on the optimized risk assessment results, combine deep causal reasoning to calculate the key risk impact factors of the scheduling plan, and use causal effect estimation based on the adaptive attention mechanism to calculate the causal weights of personnel scheduling and construction task execution: ; Among them, is the total risk impact degree of the scheduling strategy, is the key event occurrence probability, is the risk weight of the event, N is the total number of risk impact factors, is the interaction relationship between risk factors calculated by the adaptive attention mechanism, is the risk interaction influence weight, is the number of interaction relationships of risk factors; S65. Based on the calculated causal risk assessment results, adopt a multi-level reinforcement game optimization strategy to adjust the human resource allocation strategy, and construct an optimal scheduling strategy based on multi-agent risk control: ; Among them, is the optimized human resource scheduling plan, is the scheduling risk control parameter, is the scheduling adjustment strategy optimized by multi-agent reinforcement game, is the game weight, is the total number of game agents, is the scheduling plan, is the time step.

[0014] Optionally, the S7 specifically includes: S71. Based on the optimized human resource scheduling plan and the risk assessment results , structurally process the human resource allocation plan, task execution status and potential risk data, construct a multi-modal feedback data matrix F, and optimize the data expression through a dynamic feature fusion method: ; Among them, is the task execution status matrix, is the construction environment status matrix, ( ) is the multi-perspective information mapping function, is the mapping weight, represents the cumulative impact of the historical feedback state, is the historical information impact factor, is the total number of feedback data perspectives, is the total number of historical feedback time steps, is the historical feedback state; S72. Based on the generated multi-modal feedback data matrix, use a multi-terminal display platform for data visualization, and combine the operation behaviors and real-time feedback information of managers to construct a feedback optimization model based on cognitive computing, and define a feedback optimization equation: ; Among them, is the feedback optimization goal of the display platform, ( ) is the dynamic mapping function of the feedback behavior of manager m, is the operation characteristics of the manager, is the real-time feedback response data, is the total number of management personnel; S73. Based on the feedback optimization result, adopt an adaptive feedback control mechanism to adjust the human resource scheduling decision system and construct a feedback correction model: ; wherein, is the human resource scheduling plan after feedback optimization and adjustment, is the feedback adjustment step size, is the feedback optimization gradient, is the correction strategy calculated by the multi-agent feedback mechanism, is the feedback weight, is the time step, is the scheduling plan; S74. Based on the optimized human resource scheduling plan, adopt a self-supervised optimization mechanism based on attention enhancement to construct a multi-level feedback learning model and optimize the intelligent decision-making ability of the scheduling system: ; wherein, is the finally optimized human resource scheduling plan, is the feedback optimization coefficient, is the dynamic scheduling correction function based on feedback optimization, is the optimization strategy calculated by the multi-level feedback learning mechanism, is the feedback adaptability weight, is the total number of feedback learning factors, is the parameter of the d-th risk adaptability factor; S75. Based on the optimized human resource scheduling plan, provide data-driven decision-making suggestions through a multi-terminal display platform, combine the feedback information of management personnel, synchronously update the multi-scale digital twin model, scheduling decision system and risk warning system on the construction site, and continuously optimize the human resource intelligent scheduling system based on the long-term feedback mechanism.

[0015] The beneficial effects of the present invention are: The present invention realizes intelligent, efficient, and precise scheduling optimization in the human resource management of construction projects by introducing big data analysis, digital twin technology, reinforcement learning optimization, and an adaptive feedback control mechanism. Compared with traditional management methods based on manual experience or static rules, the present invention can dynamically perceive the real-time state of the construction site, construct a multi-scale digital twin model based on multi-modal data, accurately simulate the human resource requirements and scheduling strategies during the construction process, and ensure the rationality and adaptability of the scheduling plan. By adopting a reinforcement learning and multi-agent collaborative game mechanism, the present invention effectively improves the global optimality of the human resource scheduling plan, enabling the scheduling system to adaptively adjust according to changes in the construction environment, not only optimizing the task allocation efficiency, but also reducing resource waste and improving the construction cost control ability.

[0016] During the scheduling process, the present invention combines a Bayesian dynamic inference model for risk assessment, which can identify key risk factors that may affect the construction progress or safety in advance, and optimize the scheduling for high-risk areas to reduce potential safety hazards at the construction site. In addition, the adaptive feedback control mechanism adopted by the present invention enables the human resource scheduling plan to be optimized in combination with the operation feedback of the management personnel. The scheduling plan, construction status, and risk assessment results are intuitively presented through a multi-terminal display platform, and the management personnel can make adjustments based on the visual data, improving the intelligence level and manual intervention ability of the scheduling system, thereby enhancing the flexibility and decision-making efficiency of construction management.

[0017] Compared with the prior art, the present invention realizes the automation, intelligence, and adaptive optimization of human resource scheduling, enabling construction enterprises to manage human resources more precisely, and enhancing the controllability and reliability of the construction process. Through multi-modal data fusion, dynamic optimization of human resources, construction risk early warning, and an intelligent feedback mechanism, the present invention can effectively improve construction efficiency, reduce resource waste, lower management costs, and enhance the intelligent scheduling ability of enterprises in complex construction environments. Ultimately, the present invention not only improves the accuracy and flexibility of human resource management, but also enhances the intelligence level of the construction process, providing strong technical support for the digital transformation of the future construction engineering field. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. They are used together with the embodiments of the present invention to explain the present invention, and do not constitute a limitation to the present invention. In the drawings: Figure 1 is a flowchart of the intelligent management optimization method for construction project human resources based on big data analysis proposed by the present invention; Figure 2Schematic diagram of multi-modal data fusion and construction of a multi-scale digital twin model for the intelligent management optimization method of construction project human resources based on big data analysis proposed by the present invention. Detailed implementation manners

[0019] Now, the present invention will be further described in detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, so they only show the components related to the present invention.

[0020] Refer to Figure 1 and Figure 2 , the intelligent management optimization method of construction project human resources based on big data analysis includes the following steps: S1. Collect data at the construction project site and preprocess the data, including data denoising, feature enhancement, and anomaly detection; S2. Perform multi-modal data fusion on the preprocessed data, synchronize the data, and generate spatio-temporal correlation optimized data; S3. Based on the data at the construction project site and the spatio-temporal correlation optimized data, construct a digital twin model of the construction site, perform dynamic modeling using graph neural networks, and update the twin environment in combination with real-time data; S4. On the basis of the constructed digital twin model of the construction site, construct a human resource scheduling decision-making system, simulate the interaction relationships among different work types, construction tasks, and resource constraints, and establish an adaptive reward mechanism; S5. Use the human resource scheduling decision-making system, combine task time window constraints, worker skill matching degrees, and real-time construction progress to adjust the personnel scheduling plan, and generate a human resource allocation strategy; S6. Adopt multi-level Bayesian inference to conduct risk assessment on the human resource allocation strategy, construct a risk early warning system, predict efficiency decline, safety hazards, and resource waste, and generate a risk causal analysis report; S7. Feed back the human resource allocation strategy and risk assessment results to the management personnel through a multi-terminal display platform, and optimize the human resource scheduling decision-making system based on an adaptive feedback control mechanism.

[0021] In this embodiment, the specific content of S3 includes: S31. Obtain the data at the construction project site and the spatio-temporal correlation optimized data. The data at the construction project site includes personnel location data, operation status data, construction environment data, and equipment operation data; S32. Construct a multi-scale digital twin model of the construction site, adopt a hierarchical modeling method, including a global construction progress model at the macro level, a task assignment model at the middle level, and a personnel behavior simulation model at the micro level, respectively model the construction progress, task arrangement, and individual behavior, and define the information transfer mechanism between different levels; S33. Construct the personnel-task-environment relationship matrix : ; Among them, represents the construction personnel skill matching degree matrix, represents the task urgency matrix, represents the geographical distance matrix; S34. Use a dynamic graph neural network to model the personnel-task-environment relationship matrix and construct a construction personnel status update model: ; Among them, represents the hidden state matrix of the l-th layer, represents the adjacency matrix under different topological scales, represents the trainable weight matrix, B represents the bias term, represents the weighted coefficient of different scale topological structures, represents the hidden state matrix of the l-th layer, represents the number of different topological scales, represents the non-linear activation function, S35. Combine the modeling results to perform multi-step prediction on the construction personnel scheduling plan, and use a spatio-temporal graph convolutional variational auto-regression model to calculate the construction progress and the changing trend of personnel requirements: ; Among them, represents the completion degree of the construction task at time t, and represent the regression coefficients, p represents the time lag step, q represents the latent variable influence duration, represents the completion degree of the construction task, represents the latent variable, represents the error term; i and j represent indices; S36. Based on the prediction results, use the Bayesian optimization method to update the multi-scale digital twin model of the construction site and construct the objective function : ; Among them, represents the actual observed task completion degree, represents the predicted value, represents the neural network weight matrix, represents the regularization coefficient, represents the time window length, t represents the index.

[0022] In this embodiment, the S4 specifically includes: S41. Based on the constructed multi-scale digital twin model of the construction site, extract the personnel-task-environment relationship matrix and construction status information, and construct an optimization matrix using the dynamic relationship weighting method : ; Among them, represents the personnel-task-environment relationship matrix, represents the construction environment status matrix, represents the relationship change rate, represents the weighting coefficient; S42. Based on the optimization matrix, construct a human resource scheduling decision-making system, use graph reinforcement learning to model the task allocation process of construction workers, and define task matching scores using a multi-scale attention mechanism: ; Among them, is the allocation score between construction worker i and task j, is the trainable weight, is the element-wise multiplication, is the activation function, is the bias term, is the feature vector of construction worker i, is the feature vector of task j; S43. Use the calculated task matching scores to construct a reinforcement learning model based on the graph attention mechanism, and define the scheduling policy update rule: ; Among them, is the scheduling policy at time step t, represents the task allocation probability calculated based on the optimization matrix G, is the learning rate, is the logarithmic function, is the scheduling policy at time step t + 1, is the logarithmic gradient of the task allocation probability; S44. Adopt an optimal scheduling game model based on adversarial training, and define a multi-objective optimization function for the allocation of construction workers: ; Among them, is the comprehensive reward at the current time step, is the scheduling smoothness control coefficient, represents the optimized human resource scheduling plan, represents the scheduling plan at time step t, represents the scheduling plan at time step t - 1, represents the time window length, Solve for the personnel task allocation matrix A that maximizes the objective function; S45. Using the optimized human resource allocation plan, adopt a hybrid adaptive adjustment mechanism to fine-tune the personnel scheduling plan, and calculate the human resource scheduling matrix based on dynamic task requirements: ; where, is the human resource scheduling matrix, is the task demand gradient, is the adjustment step size, represents the gradient adjustment direction.

[0023] In this embodiment, the specific steps of S5 include: S51. Based on the generated personnel scheduling plan and the multi-scale digital twin model of the construction site, extract historical scheduling data, construction progress status, and worker skill matching information, and construct a multi-modal task matching matrix Z: ; where, is the personnel scheduling plan, is the personnel-task-environment relationship matrix, is the construction environment status matrix, is the task demand gradient regarding the scheduling matrix, ( ) is the multi-modal fusion function; S52. Based on the constructed task matching matrix, adopt a reinforcement learning model optimized by a graph variational autoencoder to perform multi-step reasoning on the construction task allocation plan, and calculate the scheduling optimization objective using the task state transition function: ; where, is the scheduling policy, is the comprehensive reward at the current time step, represents the structural change distance of the task matching matrix between consecutive time steps, S is the number of scheduling decision iteration steps, is the scheduling smoothness control parameter, is the expected symbol, is the time window length, is the task matching matrix at time step s, is the task matching matrix at time step s−1, represents solving for the scheduling policy that maximizes the objective function ; S53. Based on the generated scheduling policy, combined with historical task execution data and real-time construction environment information, adopt a multi-agent adaptive attention mechanism for dynamic scheduling optimization, and define the update rule of the human resource scheduling matrix: ; Among them, is the optimized human resource scheduling matrix, is the initial scheduling matrix calculated, is the scheduling stability control parameter, is the task gradient optimization coefficient, is the scheduling policy distribution, is the policy parameter, is the scheduling policy, is the number of scheduling schemes; S54. Based on the optimized human resource scheduling matrix, adopt an adaptive adjustment mechanism based on gradient manifold mapping for dynamic optimization: ; Among them, is the gradient manifold mapping function, is the scheduling convergence weight, is the human resource allocation strategy, is the task demand objective function; S55. Based on the human resource allocation strategy, optimize the long-term scheduling stability and global optimality of human resource management.

[0024] In this embodiment, the specific content of S6 includes: S61. Based on the generated human resource allocation strategy and the multi-scale digital twin model of the construction site, extract historical scheduling data, construction progress status, and personnel task execution status, construct a task execution status matrix Q, and adopt an adaptive multi-scale spatio-temporal embedding method for state representation optimization: ; Among them, is the personnel-task-environment relationship matrix, is the construction environment status matrix, is the task demand gradient regarding the scheduling matrix, ( ) is the dynamic state embedding function, is the spatio-temporal correlation weight, is the influence function of the historical state, is the human resource allocation strategy, is the historical window size; S62. Based on the constructed task execution status matrix, adopt a risk assessment model based on Bayesian dynamic inference to conduct multi-scale risk prediction on the human resource allocation strategy, and combine multi-objective confidence intervals to optimize and calculate the uncertainty of task execution: ; Among them, is based on the task execution status matrix The calculated posterior probability distribution is the risk scoring function for construction tasks is the risk variable space is the uncertainty confidence interval calculated during task execution is the uncertainty control parameter is the risk variable for construction tasks is the total number of uncertainty factors for construction tasks is the risk score S63. Based on the calculated risk score, use a generative adversarial network to generate potential high - risk scenarios, and optimize the calculation of risk impact degree based on graph adversarial game: ; Among them, is the optimal risk generator represents the change distance of the task execution state at consecutive time steps is the risk stability control parameter is the risk game loss calculated by the graph game strategy is the game balance coefficient is the expectation operator The optimization objective is to find the that makes the performance of the risk assessment model optimal is the current task execution state matrix is the state matrix at the previous moment is the high - risk scenario is the actual task state change S64. Based on the optimized risk assessment results, combine deep causal reasoning to calculate the key risk impact factors of the scheduling plan, and use causal effect estimation based on the adaptive attention mechanism to calculate the causal weights between personnel scheduling and construction task execution: ; Among them, is the total risk impact degree of the scheduling strategy is the key event The probability of occurrence is the risk weight of the event. N is the total number of risk impact factors is the interaction relationship between risk factors calculated by the adaptive attention mechanism is the risk interaction impact weight is the number of interaction relationships between risk factors S65. Based on the calculated causal risk assessment results, use a multi - level reinforcement game optimization strategy to adjust the human resource allocation strategy, and construct an optimal scheduling strategy based on multi - agent risk control: ; Among them, is the optimized human resource scheduling plan, is the scheduling risk control parameter, is the scheduling adjustment strategy optimized by multi-agent reinforcement game, is the game weight, is the total number of game agents, is the scheduling plan, is the time step.

[0025] In this embodiment, the S7 specifically includes: S71. Based on the optimized human resource scheduling plan and the risk assessment result , structure the human resource allocation plan, task execution status and potential risk data, construct the multi-modal feedback data matrix F, and optimize the data expression through the dynamic feature fusion method: ; Among them, is the task execution status matrix, is the construction environment status matrix, ( ) is the multi-perspective information mapping function, is the mapping weight, represents the cumulative impact of the historical feedback state, is the historical information impact factor, is the total number of feedback data perspectives, is the total number of historical feedback time steps, is the historical feedback state; S72. Based on the generated multi-modal feedback data matrix, use the multi-terminal display platform for data visualization, and combine the operation behaviors and real-time feedback information of the management personnel to construct a feedback optimization model based on cognitive computing, and define the feedback optimization equation: ; Among them, is the feedback optimization goal of the display platform, ( ) is the dynamic mapping function of the feedback behavior of the management personnel m, is the operation characteristics of the management personnel, is the real-time feedback response data, is the total number of management personnel; S73. Based on the feedback optimization result, adopt the adaptive feedback control mechanism to adjust the human resource scheduling decision system and construct the feedback correction model: ; Among them, is the human resource scheduling plan adjusted based on the feedback optimization, is the feedback adjustment step size, is the feedback optimization gradient, is the correction strategy calculated by the multi-agent feedback mechanism, is the feedback weight, is the time step, is the scheduling plan; S74. Based on the optimized human resource scheduling plan, an attention-enhanced self-supervised optimization mechanism is adopted to construct a multi-level feedback learning model to optimize the intelligent decision-making ability of the scheduling system: ; Among them, is the finally optimized human resource scheduling plan, is the feedback optimization coefficient, is the dynamic scheduling correction function based on feedback optimization, is the optimization strategy calculated by the multi-level feedback learning mechanism, is the feedback adaptability weight, is the total number of feedback learning factors, is the parameter of the d-th risk adaptability factor; S75. Based on the optimized human resource scheduling plan, data-driven decision-making suggestions are provided through a multi-terminal display platform. Combining the feedback information of managers, the multi-scale digital twin model, scheduling decision-making system, and risk warning system of the construction site are synchronously updated, and the human resource intelligent scheduling system is continuously optimized based on the long-term feedback mechanism.

[0026] Example 1: To verify the feasibility of the present invention in implementation, the present invention is applied to a large commercial complex construction project. The project covers an area of 50,000 square meters, with a total construction period of 18 months, involving 1,200 construction workers, covering multiple types of work such as steel workers, concrete workers, bricklayers, electricians, and welders. Due to the complex construction tasks and large personnel mobility, the traditional scheduling method mainly relies on the experience of the construction manager for arrangement, often resulting in unreasonable personnel arrangements, uneven task distribution, construction delays, etc., leading to a decrease in construction efficiency and an increase in resource waste and management costs.

[0027] This project introduces an intelligent management optimization method for construction project human resources based on big data analysis. Multimodal data acquisition devices are used to obtain the real-time positions, attendance data, and construction progress information of construction workers, and combined with historical scheduling data, a multi-scale digital twin model of the construction site is constructed. Through this model, the distribution of different types of work personnel, task completion progress, and construction site status can be simulated, providing data support for optimizing human resource scheduling.

[0028] During the concrete pouring stage, 150 workers need to be coordinated, including 80 concrete workers, 30 steel bar workers, and 40 auxiliary workers. The traditional scheduling method has a lag in response when encountering emergencies (such as equipment failures, personnel absences, etc.), resulting in low labor-hour utilization rate. Some personnel are in a waiting state, affecting the overall construction progress. However, based on the reinforcement learning and multi-agent game optimization algorithm, the present invention can adjust the personnel allocation in real time. For example, when construction in a certain area is suspended due to weather reasons, the system will automatically transfer the workers in this area to other tasks, improving the labor-hour utilization rate and ensuring the smooth progress of construction.

[0029] In addition, during the construction process, a Bayesian dynamic inference model is combined for risk assessment. For example, during high-altitude operations, the system analyzes historical data and finds that during peak hours (8:00 - 11:00 in the morning, 14:00 - 17:00 in the afternoon), due to factors such as personnel fatigue and poor cooperation, the incidence of construction safety accidents is relatively high. Therefore, the system dynamically adjusts the personnel allocation plan during high-risk periods and increases the frequency of safety inspections. Eventually, the safety accident rate at the construction site is significantly reduced.

[0030] In terms of construction management, the present invention adopts a multi-terminal display platform to feedback the human resource allocation plan and construction progress information to the management personnel in real time. Through the intelligent visualization system, project management personnel can intuitively view the construction progress, personnel scheduling situation, and risk assessment results, and make necessary adjustments. For example, during the basic structure construction stage, the management personnel can monitor the distribution of personnel in each work type in real time and optimize the personnel scheduling in combination with factors such as weather and equipment conditions, avoiding construction delays caused by improper personnel arrangements.

[0031] To verify the actual effect of the present invention, the performance of the optimization method of the present invention and the traditional manual scheduling method in terms of construction efficiency was compared respectively, and the recorded data are as follows: Table 1 Comparison of construction efficiency before and after construction scheduling optimization Construction stage Utilization rate of traditional manual scheduling man-hours (%) Utilization rate of man-hours using the present invention (%) Improvement rate (%) Steel bar binding 76.5 89.2 16.6 Concrete pouring 72.3 91.5 26.7 Masonry construction 78.1 88.7 13.6 Mechanical and electrical installation 74.8 90.2 20.5 Fine decoration 80.2 92.3 15.1 It can be seen from the above data that the present invention has significant advantages in construction scheduling optimization, effectively improving the labor-hour utilization rate in each construction stage and enhancing the construction efficiency. Among them, during the concrete pouring stage, due to the complex tasks, large personnel requirements, and great influence of external factors such as weather, the labor-hour utilization rate of the traditional manual scheduling method is only 72.3%. After adopting the optimization method of the present invention, the labor-hour utilization rate in this stage is increased to 91.5%, an increase of 26.7%, indicating that the present invention can effectively reduce the waiting time of personnel and improve the construction efficiency by intelligent scheduling and optimizing task allocation.

[0032] During the steel bar binding stage, the labor-hour utilization rate of the traditional scheduling method was 76.5%, while after optimization by the present invention, it increased to 89.2%, with an increase of 16.6%. The optimization in this stage mainly benefits from the application of digital twin technology, which can simulate the construction progress in real time and dynamically adjust the personnel allocation in combination with the reinforcement learning algorithm, thereby reducing the idle time of workers caused by poor task connection and improving the continuity of construction. In addition, during the masonry construction stage, the optimized labor-hour utilization rate increased from 78.1% to 88.7%, an increase of 13.6%, indicating that the present invention can also achieve good optimization effects in construction stages with high personnel density and fine task division.

[0033] During the mechanical and electrical installation stage, due to a large number of equipment commissioning and cross-trade collaborations involved, there are often situations where workers are idle due to waiting for equipment commissioning or insufficient material supply during the construction process. Under the traditional scheduling mode, the labor-hour utilization rate in this stage was 74.8%, while after optimization using the present invention, the labor-hour utilization rate increased to 90.2%, an increase of 20.5%. This shows that the present invention can not only optimize the scheduling of a single trade but also improve the efficiency of cross-trade collaborative operations, making the connection between equipment installation and manual construction closer and reducing the waste of labor hours caused by waiting.

[0034] During the fine decoration stage, the labor-hour utilization rate under the traditional scheduling mode was 80.2%, and after optimization, it increased to 92.3%, an increase of 15.1%. The improvement in this stage is mainly reflected in the precise allocation of the time window for construction tasks. The present invention combines dynamic prediction of the construction progress and the reinforcement learning scheduling optimization mechanism, and adjusts the worker task allocation in real time according to the construction progress and the feedback information of the management personnel, enabling the workers to always be in a highly efficient working state and further improving the construction efficiency.

[0035] Generally speaking, through multi-modal data analysis, reinforcement learning optimization scheduling, digital twin simulation, and adaptive feedback control mechanism, the present invention has significantly improved the labor-hour utilization rate of construction personnel and reduced the time waste caused by unreasonable manual scheduling. In different construction stages, the increase in the optimized labor-hour utilization rate ranges from 13.6% to 26.7%. Especially in construction stages with high task complexity and large personnel requirements (such as concrete pouring and mechanical and electrical installation), the optimization effect of the present invention is particularly significant. These data indicate that the present invention can play an important role in the human resource management of building construction, provide a more accurate, efficient, and intelligent scheduling scheme for construction enterprises, and contribute to the digital transformation and intelligent upgrading of the building engineering field.

[0036] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, making equivalent substitutions or changes should be covered within the protection scope of the present invention.

Claims

1. The intelligent management optimization method of construction engineering human resources based on big data analysis is characterized by: The steps include: S1. Collect data from construction sites and pre-process the data, including data noise reduction, feature enhancement, and anomaly detection; S2, perform multimodal data fusion and data synchronization on the preprocessed data to generate spatiotemporal correlation optimized data; S3. Based on the data of the construction site and the spatiotemporal correlation optimization data, a digital twin model of the construction site is constructed, and graph neural networks are used for dynamic modeling, and the twin environment is updated in combination with real-time data; S4. Based on the constructed digital twin model of the construction site, a human resource scheduling decision-making system is constructed to simulate the interactive relationship between different types of work, construction tasks and resource constraints, and establish an adaptive reward mechanism; S5. Use the human resource scheduling decision system to adjust the personnel scheduling plan in combination with the task time window constraints, worker skill matching and real-time construction progress to generate a human resource allocation strategy; S6. Use multi-level Bayesian reasoning to conduct risk assessment on human resource allocation strategies, build a risk early warning system, predict efficiency decline, safety hazards and resource waste, and generate risk causal analysis reports; S7. Feedback the human resource allocation strategy and risk assessment results to management personnel through a multi-terminal display platform, and optimize the human resource scheduling decision-making system based on an adaptive feedback control mechanism.

2. The construction engineering human resources intelligent management optimization method based on big data analysis according to claim 1 is characterized in that: The S3 specifically includes: S31, obtaining construction project site data and spatiotemporal correlation optimization data, the construction project site data including personnel location data, operation status data, construction environment data and equipment operation data; S32. Construct a multi-scale digital twin model of the construction site, using a hierarchical modeling approach, including a global construction progress model at the macro level, a task allocation model at the meso level, and a personnel behavior simulation model at the micro level. The construction progress, task arrangement, and individual behavior are modeled separately, and the information transmission mechanism between different levels is defined. S33. Build a Personnel-Task-Environment Relationship Matrix : ； in, represents the construction personnel skill matching matrix, represents the task urgency matrix, represents a geographic distance matrix; S34. Use dynamic graph neural network to model the personnel-task-environment relationship matrix and build a construction personnel status update model: ； in, represents the hidden state matrix of the lth layer, represents the adjacency matrix at different topological scales, represents the trainable weight matrix, B represents the bias term, Represents the weight coefficients of topological structures at different scales, Indicates The hidden state matrix of the layer, represents the number of different topological scales, represents a nonlinear activation function, S35. Combined with the modeling results, a multi-step prediction of the construction personnel scheduling plan is carried out, and the construction progress and personnel demand change trend are calculated using the spatiotemporal graph convolution variational autoregression model: ； in, represents the completion degree of the construction task at time t, and represents the regression coefficient, p represents the time lag step, q represents the duration of latent variable influence, Indicates the completion of the construction task. represents a hidden variable, represents the error term; i and j represent indexes; S36. Based on the prediction results, the Bayesian optimization method is used to update the multi-scale digital twin model of the construction site and construct the objective function : ； in, Indicates the actual observation task completion degree, represents the predicted value, represents the neural network weight matrix, represents the regularization coefficient, represents the time window length, and t represents the index.

3. The construction engineering human resources intelligent management optimization method based on big data analysis according to claim 1 is characterized in that: The S4 specifically includes: S41. Based on the multi-scale digital twin model of the construction site, the personnel-task-environment relationship matrix and construction status information are extracted, and the optimization matrix is ​​constructed using the dynamic relationship weighting method. : ； in, Represents the personnel-task-environment relationship matrix, represents the construction environment state matrix, represents the rate of change of the relationship, represents the weighting coefficient; S42. Build a human resource scheduling decision system based on the optimization matrix, use graph reinforcement learning to model the construction personnel task allocation process, and use a multi-scale attention mechanism to define the task matching score: ； in, Score the allocation between construction worker i and task j, are trainable weights, is element-wise multiplication, is the activation function, is the bias term, is the characteristic vector of construction worker i, is the feature vector of task j; S43. Using the calculated task matching scores, a reinforcement learning model based on the graph attention mechanism is constructed to define the scheduling strategy update rules: ； in, is the scheduling strategy at time step t, represents the task assignment probability calculated based on the optimization matrix G, is the learning rate, is a logarithmic function, is the scheduling strategy at time step t+1, The log gradient of the probability assigned to the task; S44. Adopt the optimal scheduling game model based on adversarial training to define the multi-objective optimization function of construction personnel allocation: ； in, is the comprehensive reward of the current time step, is the scheduling smoothness control coefficient, represents the optimized human resource scheduling plan, represents the scheduling plan at time step t, represents the scheduling plan at time step t−1, represents the time window length, Solve the personnel task allocation matrix A that maximizes the objective function; S45. Using the optimized human resource allocation plan, a hybrid adaptive adjustment mechanism is used to fine-tune the personnel scheduling plan, and the human resource scheduling matrix is ​​calculated based on dynamic task requirements: ； in, is the human resource scheduling matrix, is the task requirement gradient, To adjust the step size, Indicates the gradient adjustment direction.

4. The construction engineering human resources intelligent management optimization method based on big data analysis according to claim 1 is characterized in that: The S5 specifically includes: S51. Based on the generated personnel scheduling plan and the multi-scale digital twin model of the construction site, historical scheduling data, construction progress status and worker skill matching information are extracted to construct a multimodal task matching matrix Z: ； in, To provide personnel scheduling solutions, For the Person-Task-Environment Relationship Matrix, is the construction environment state matrix, is the task demand gradient with respect to the scheduling matrix, ( ) is the multimodal fusion function; S52. Based on the constructed task matching matrix, a reinforcement learning model based on graph variational autoencoder optimization is used to perform multi-step reasoning on the construction task allocation plan, and the task state transfer function is used to calculate the scheduling optimization target: ； in, is the scheduling strategy, is the comprehensive reward of the current time step, represents the structural change distance of the task matching matrix between consecutive time steps, S is the number of scheduling decision iterations, is the scheduling smoothness control parameter, is the expected symbol, is the time window length, is the task matching matrix at time step s, is the task matching matrix at time step s−1, Represents the scheduling strategy that solves the objective function to the maximum value ; S53. Based on the generated scheduling strategy, combined with historical task execution data and real-time construction environment information, a multi-agent adaptive attention mechanism is used for dynamic scheduling optimization, and the human resource scheduling matrix update rules are defined: ； in, is the optimized human resource scheduling matrix, is the initial scheduling matrix for calculation, is the scheduling stability control parameter, is the task gradient optimization coefficient, is the scheduling strategy distribution, is the strategy parameter, is the scheduling strategy, is the number of scheduling solutions; S54. Based on the optimized human resource scheduling matrix, an adaptive adjustment mechanism based on gradient manifold mapping is used for dynamic optimization: ； in, is the gradient manifold mapping function, is the scheduling convergence weight, Strategies for human resource allocation, is the objective function required by the task; S55. Based on human resource allocation strategy, optimize the long-term scheduling stability and global optimality of human resource management.

5. The construction engineering human resources intelligent management optimization method based on big data analysis according to claim 1 is characterized in that: The S6 specifically includes: S61. Based on the generated human resource allocation strategy and the multi-scale digital twin model of the construction site, the historical scheduling data, construction progress status and personnel task execution status are extracted to construct the task execution state matrix Q, and the adaptive multi-scale spatiotemporal embedding method is used to optimize the state representation: ； in, For the Person-Task-Environment Relationship Matrix, is the construction environment state matrix, is the task demand gradient with respect to the scheduling matrix, ( ) is the dynamic state embedding function, is the spatiotemporal association weight, is the influence function of the historical state, Strategies for human resource allocation, is the history window size; S62. Based on the constructed task execution status matrix, a risk assessment model based on Bayesian dynamic reasoning is used to perform multi-scale risk prediction on the human resource allocation strategy, and the uncertainty of task execution is calculated in combination with multi-objective confidence interval optimization: ； in, Based on the task execution state matrix Compute the posterior probability distribution, is the construction task risk scoring function, is the risk variable space, is the uncertainty confidence interval calculated during task execution, is the uncertainty control parameter, is the construction task risk variable, is the total number of uncertain factors of the construction task, Score the risk; S63. Based on the calculated risk score, a generative adversarial network is used to generate potential high-risk scenarios, and the risk impact is calculated based on graph adversarial game optimization: ； in, is the optimal risk generator, Indicates the change distance of the task execution state in consecutive time steps, is the risk stability control parameter, The risk game loss calculated for the graph game strategy, is the game equilibrium coefficient, is the expectation operator, The optimization goal is to find the optimal risk assessment model. , is the current task execution state matrix, is the state matrix at the previous moment, For high-risk scenarios, For actual task status changes; S64. Based on the optimized risk assessment results, deep causal reasoning is combined to calculate the key risk influencing factors of the scheduling plan, and the causal effect estimation based on the adaptive attention mechanism is used to calculate the causal weights of personnel scheduling and construction task execution: ； in, is the total risk impact of the scheduling strategy, For key events The probability of occurrence, is the risk weight of the event, N is the total number of risk influencing factors, The interactions between risk factors calculated for the adaptive attention mechanism, is the risk interaction weight, is the number of interactions of risk factors; S65. Based on the calculated causal risk assessment results, a multi-level reinforcement game optimization strategy is used to adjust the human resource allocation strategy and construct an optimal scheduling strategy based on multi-agent risk control: ； in, To optimize the human resource scheduling plan, is the scheduling risk control parameter, Scheduling adjustment strategies for multi-agent reinforcement game optimization, is the game weight, is the total number of game agents, For the scheduling plan, is the time step.

6. The construction engineering human resources intelligent management optimization method based on big data analysis according to claim 1 is characterized in that: The S7 specifically includes: S71, based on the optimized human resource scheduling plan and risk assessment results , the human resource allocation plan, task execution status and potential risk data are structured, a multimodal feedback data matrix F is constructed, and the data expression is optimized through the dynamic feature fusion method: ； in, is the task execution state matrix, is the construction environment state matrix, ( ) is the multi-view information mapping function, is the mapping weight, represents the cumulative impact of historical feedback states, is the historical information impact factor, is the total number of feedback data perspectives, is the total number of historical feedback time steps, is the historical feedback status; S72. Based on the generated multimodal feedback data matrix, a multi-terminal display platform is used for data visualization. In combination with the management personnel's operation behavior and real-time feedback information, a feedback optimization model based on cognitive computing is constructed, and the feedback optimization equation is defined: ； in, To optimize the feedback goals of the display platform, ( ) is the dynamic mapping function of the feedback behavior of manager m, Operate the features for managers, To provide real-time feedback response data, is the total number of managers; S73. Based on the feedback optimization results, an adaptive feedback control mechanism is used to adjust the human resource scheduling decision-making system and build a feedback correction model: ； in, To optimize and adjust the human resource scheduling plan based on feedback, is the feedback adjustment step size, Optimize gradients for feedback, Computing correction strategies for multi-agent feedback mechanisms, is the feedback weight, is the time step, For scheduling scheme; S74. Based on the optimized human resource scheduling scheme, a self-supervisory optimization mechanism based on attention enhancement is adopted to build a multi-level feedback learning model to optimize the intelligent decision-making ability of the scheduling system: ； in, For the final optimized human resource scheduling plan, is the feedback optimization coefficient, is the dynamic scheduling correction function based on feedback optimization, Optimization strategies for computing multi-level feedback learning mechanisms, is the feedback adaptability weight, is the total number of feedback learning factors, is the parameter of the dth risk adaptability factor; S75. Based on the optimized human resource scheduling plan, data-driven decision-making suggestions are provided through the multi-terminal display platform. Combined with the feedback information from managers, the multi-scale digital twin model of the construction site, the scheduling decision system and the risk warning system are synchronously updated, and the human resource intelligent scheduling system is continuously optimized based on the long-term feedback mechanism.

Images