A project monitoring method and system based on big data

Abstract

The present invention proposes a project monitoring method and system based on big data, the method includes: obtaining the project data of the required detection project, constructing a project data set according to the project data and uploading it to a cloud server for storage and constructing a project scenario model; applying a dynamic weighted federated average algorithm to perform adaptive distributed joint training based on blockchain on the project scenario model; constructing and using a deep reinforcement learning model, adjusting the prediction project monitoring strategy according to the real-time status and historical data of the project, judging the accuracy of the prediction project strategy in combination with the progress status of the current project, introducing a correction plan according to the accuracy of the prediction project strategy to dynamically adjust the prediction project monitoring strategy to obtain the project monitoring strategy; performing monitoring tasks according to the adjusted monitoring strategy, feeding back the monitoring results to the system, and optimizing the monitoring strategy according to the detection results. The present invention realizes accurate monitoring and early warning, and provides strong support for project management decision-making.

Description

Technical Field

[0001] The present invention belongs to the technical field of project monitoring, and in particular relates to a project monitoring method and system based on big data. Background Art

[0002] In the context of big data, project monitoring faces the contradiction between multi-party data integration and privacy protection. Traditional methods make it difficult to effectively utilize the massive data scattered among different participants while protecting the privacy of all parties' data. In addition, the project environment is complex and changeable, and static models are difficult to adapt to dynamic changes, resulting in delayed or inaccurate monitoring results. How to fully integrate multi-party data to improve model accuracy while protecting privacy? How to make the model adaptive and adjust in real time to adapt to changes in the project environment? Summary of the invention

[0003] The purpose of the present invention is to propose a project monitoring method and system based on big data, which can fully utilize distributed data, protect privacy, and dynamically adapt to environmental changes, while considering data quality differences and dynamically adjusting the contributions of all parties. It also has self-learning capabilities, continuously optimizes prediction strategies based on the real-time status and historical data of the project, and realizes accurate, timely, and safe project monitoring.

[0004] In order to achieve the above object, a first aspect of the present invention provides a project monitoring method based on big data, the method comprising:

[0005] S1. Obtain the project data of the required detection project, preprocess the project data, build a project data set based on the preprocessed project data and upload it to the cloud server for storage, extract and fuse the data set, and build a project scenario model based on the fused multimodal features;

[0006] S2. Apply the dynamic weighted federated average algorithm to conduct blockchain-based adaptive distributed joint training on the project scenario model, update the project scenario model, dynamically adjust the weight of the project scenario model update according to the real-time data of each participating node, and update the aggregated weighted project scenario model to obtain the global project scenario model;

[0007] S3. Build and use a deep reinforcement learning model to adjust the forecast project monitoring strategy according to the real-time status and historical data of the project, judge the accuracy of the forecast project strategy in combination with the progress status of the current project, and introduce a correction plan based on the accuracy of the forecast project strategy to dynamically adjust the forecast project monitoring strategy to obtain the project monitoring strategy;

[0008] S4. Perform monitoring tasks according to the adjusted monitoring strategy, feed back monitoring results to the system, and optimize the monitoring strategy according to the detection results;

[0009] The step of judging the accuracy of the predicted project strategy in combination with the progress status of the current project specifically includes:

[0010] The current status of the project is obtained based on the degree of deviation between the actual project progress and the expected progress, and the specific reasons for the deviation of the project progress are determined by the completion of the key milestones of the project:

[0011] If the actual investment of project resources matches the plan, the impact of resource allocation rationality on schedule deviation is considered. The project risk identification and response effect are used to analyze the impact of potential risks on project schedule. The impact of quality control on project schedule is evaluated by the achievement of project quality goals.

[0012] If the project budget execution is highly compliant, then examine the contribution of budget control to schedule deviation; obtain the satisfaction evaluation results of project stakeholders and determine the feedback of the results on project strategy adjustment; predict the accuracy of project strategy based on the analysis results and obtain suggestions for strategy adjustment;

[0013] If the actual investment of project resources and the project budget obtained do not match the forecast, the strategy will be re-evaluated and a new adjustment strategy will be determined.

[0014] Furthermore, the dataset is subjected to feature extraction and fusion, and a project scenario model is constructed according to the fused multimodal features, including:

[0015] If time attribute data is obtained, then the timestamp, date, and time period information are extracted to obtain the time characteristics of the data. According to the spatial attribute information, the geographic location, coordinates, and regional data are used to determine the spatial characteristics of the data. Through the user attribute data, the user ID, user behavior, and user preference are determined to obtain the characteristic information related to the user.

[0016] If analyzing content attribute data, then extract text content, image content, and audio content to obtain the specific content characteristics of the data; according to statistical attributes, use mean, variance, maximum value, and minimum value to determine the statistical characteristics of the data; through relationship attribute information, determine association rules, network relationships, and hierarchical relationships to obtain the relationship characteristics between data;

[0017] If environmental attribute data is obtained, the weather, temperature, and humidity information are extracted to obtain the environmental characteristics when the data occurs; based on the device attribute information, the device type, device status, and device configuration are used to determine the characteristics of the data collection device; through the behavioral attributes and event attribute data, the click, browse, purchase behavior and event type, event frequency and duration are judged to obtain user behavior and event characteristics in specific scenarios.

[0018] Furthermore, the application of the dynamic weighted federated average algorithm to perform blockchain-based adaptive distributed joint training on the project scenario model to update the project scenario model includes:

[0019] The data of each participating node has been pre-processed locally, and the nodes are securely connected through the blockchain network;

[0020] According to the quality and quantity of local data, the weight of each node is calculated using a dynamic weighting mechanism to obtain a weight list of nodes participating in training;

[0021] The node performs weighted average calculation on the model parameters according to the weight list to obtain updated local model parameters; if it is found during the training process that the model accuracy of the node does not meet the expected indicators, the model parameters are dynamically modified or the training strategy is optimized through adaptive adjustment strategy to obtain optimized model parameters;

[0022] Through the consensus mechanism of the blockchain network, the node broadcasts the optimized model parameters to other nodes, determines whether other nodes accept the model update, and obtains the model version recognized by consensus;

[0023] Get decentralized training control flow;

[0024] Through the transparent logs recorded in the blockchain, nodes obtain detailed records of each model update from the chain; if business needs change, the number and configuration of nodes participating in training are adjusted according to scalability properties to obtain a node group that meets the current business scale;

[0025] Through the federated learning algorithm, the final model parameters of each node are integrated to obtain the global model, which is stored on the blockchain.

[0026] Furthermore, dynamically adjusting the weight of updating the project scenario model according to the real-time data of each participating node, and updating the aggregated weighted project scenario model to obtain a global project scenario model, includes:

[0027] Based on the latest data provided by each participating node, the following analysis is performed:

[0028] If the node data is valid, the system will dynamically adjust the weight of each node based on the node contribution and data accuracy;

[0029] If the weight adjustment parameters are determined, the system will apply these parameters to weight the project scenario models of each node to obtain the weighted results of each model;

[0030] If the weighted results of each model are obtained, the system obtains a preliminary global project scenario model through the aggregate weighting process;

[0031] Based on the preliminary global project scenario model obtained, the system will conduct a model evaluation to determine whether the model's performance indicators meet the preset standards, including:

[0032] If the model evaluation is judged to be below the standard, the system will obtain feedback information from each node on the global model through the feedback mechanism;

[0033] If the model parameters are adjusted through the feedback mechanism, the system will determine the model update frequency again;

[0034] If the model update frequency is determined, the system uses a communication protocol for data synchronization;

[0035] After data synchronization is complete, continue to maintain the parameters of the global project scenario model.

[0036] Furthermore, the construction uses a deep reinforcement learning model to adjust the predictive project monitoring strategy based on the real-time status and historical data of the project, including:

[0037] According to the real-time status and historical data of the project, the initial input parameters of the project are obtained by acquiring the basic information of the project; wherein the basic information of the project includes the project name, type, scale, budget and planned duration;

[0038] By analyzing the initial input parameters and combining the completed workload, actual cost, schedule deviation and resource usage in the real-time status of the project, the current progress of the project can be judged, including:

[0039] If the progress of the project deviates from expectations, compare historical data, combine the execution status, risk events, response measures and effects of similar projects, determine the potential risks faced by the current project and their impact, and obtain the additional impact on the project based on changes in external environmental factors, including changes in policies and regulations, market demand fluctuations and changes in raw material prices, and determine the specific impact of these factors on the project progress;

[0040] By analyzing the project organizational structure and personnel capabilities, combined with team member composition, skill level and work performance, the project team's execution capabilities are evaluated to determine whether team adjustments or skill enhancements are needed;

[0041] Adopt project management systems and tools, combine management systems, process specifications, information systems and communication and coordination mechanisms, optimize project management processes and information flows, and ensure the effectiveness of information transmission within the project team;

[0042] Build a deep reinforcement learning model and set the parameters of the reinforcement learning model, including state space, action space, reward function, learning rate and discount factor, so that the deep reinforcement learning model can learn the dynamic changes of the project;

[0043] Through deep reinforcement learning model training and evaluation, the deep reinforcement learning model is trained using training data sets, validation data sets, combined with evaluation indicators and number of iterations to obtain its prediction accuracy in different project scenarios;

[0044] If the training results of the deep reinforcement learning model meet expectations, the prediction results are interpreted and applied, the results are visualized, and the optimization direction of the project monitoring strategy is determined through decision-making recommendations and adjustment plans. If a risk warning is obtained through the prediction results, the project monitoring strategy is adjusted in real time to ensure that the project always remains on the expected track under changes in the external environment and internal resource adjustments.

[0045] Furthermore, the method of introducing a correction plan according to the accuracy of the forecast project strategy to dynamically adjust the forecast project monitoring strategy to obtain the project monitoring strategy includes:

[0046] If the prediction accuracy of the project strategy is higher than the set threshold, obtain the parameter adjustment plan of the prediction model:

[0047] By analyzing the coverage of the forecast project strategy, we can obtain the areas with low coverage and determine the parameter weights of these areas in the forecast model; if the adjustment range of the revised plan exceeds the predetermined standard, determine the impact of the adjustment range on the forecast accuracy and obtain the impact assessment results; according to the adjustment frequency of the revised plan, obtain the correlation data between the adjustment frequency and the forecast accuracy and determine the impact of the frequency on the accuracy; use the dynamic update cycle data of the project monitoring strategy to determine the effect of the update cycle on the monitoring accuracy; by analyzing the monitoring indicators of the project monitoring strategy, obtain the key monitoring indicators related to the forecast project and determine the relationship between the indicators and the forecast accuracy; according to the early warning threshold of the project monitoring strategy, obtain the impact of the threshold setting on the effectiveness of the project early warning; through the feedback mechanism of the project monitoring strategy, determine whether the feedback mechanism is perfect and its effect on the forecast accuracy and dynamic adjustment of the monitoring strategy; according to the accuracy of the forecast project strategy and the results of the dynamic adjustment of the revised plan, obtain the final adjustment plan of the project monitoring strategy;

[0048] Comprehensively determine the implementation plan of the project monitoring strategy to form a complete predictive monitoring process.

[0049] Furthermore, the parameter adjustment scheme of the prediction model includes:

[0050] If the input dataset is large, calculate efficiency and model convergence speed by determining an appropriate batch size;

[0051] Use the initial learning rate to obtain the initial performance indicators of the training model, and observe whether the learning rate needs to be adjusted;

[0052] By adding momentum terms, determine whether the convergence speed of the model can be improved;

[0053] Adjust the number of hidden layer units according to the model's expressiveness requirements;

[0054] Select different optimizers, determine their impact on model performance, and determine the final optimizer solution based on actual performance.

[0055] Furthermore, the weight of each node is calculated using a dynamic weighting mechanism to obtain a weight list of nodes participating in the training, including:

[0056] Obtaining current status information of at least one node, and if the current status of the node is online and fault-free, evaluating the node, otherwise skipping the node and obtaining status information of the next node;

[0057] Evaluate the computing power of the node, obtain the CPU and GPU performance indicators of the node, and determine the computing power score of the node according to the preset computing power evaluation rules for subsequent calculation of the node weight;

[0058] Acquire data quality information of the node, the data quality information including data volume and annotation accuracy, and determine the data quality score of the node according to a preset data quality assessment rule;

[0059] Obtaining historical performance data of the node, including model accuracy and convergence speed, and determining a historical performance score of the node according to a preset historical performance evaluation rule;

[0060] Scoring based on the computing power of the node , data quality score and historical performance ratings , using the preset weight calculation formula to obtain the preliminary weight value of the node , which is expressed as follows:

[0061]

[0062] in, , and Score computing ability , data quality score and historical performance ratings The weight of

[0063] Get the current status information of the node, if the node is online and the computing power score of the node Above the preset threshold , then adjust the preliminary weight value of the node according to the preset weight adjustment rule; wherein the weight adjustment rule is expressed as follows:

[0064]

[0065] in, represents the updated weight, represents the weight coefficient;

[0066] For the preliminary weight values ​​of all nodes, a preset normalization method is used to obtain the standardized weight value of each node;

[0067] According to the calculated node weight value list, the preset dynamic weighting mechanism is used to sort the nodes participating in the training, and finally determine the weight distribution scheme of each node in the model aggregation process.

[0068] Furthermore, the node weight value list obtained by calculation adopts a preset dynamic weighting mechanism to sort the nodes participating in the training, and finally determines the weight distribution scheme of each node in the model aggregation process, including:

[0069] According to the calculated node weight value list, the preset dynamic weighting mechanism is used to sort the nodes participating in the training, and finally determine the weight distribution scheme of each node in the model aggregation process;

[0070] According to the node ID, obtain the node weight value list, and obtain the preliminary weight data of each node through the data processing mechanism;

[0071] Determine the status of each node. If the node status is normal, retain the node weight value:

[0072] If the node status is abnormal, the node weight value is adjusted to zero. The nodes are sorted by the node weight value to obtain a node sequence arranged from large to small according to the weight value. According to the sorted node sequence, the cumulative weight sum of each node is calculated to obtain a cumulative weight list. A dynamic weighting mechanism is used to allocate the corresponding aggregation weight percentage according to the cumulative weight list to obtain the weight distribution ratio of each node. According to the aggregation round, the participation of the current node is judged. If the node participates in the current round, its weight distribution ratio is retained;

[0073] If the node does not participate, the weight distribution ratio is set to zero. The aggregation weight is distributed according to the node weight distribution ratio to obtain the actual weight of each node in the model aggregation process. According to the actual weight, the contribution of the node in the model aggregation process is judged, and the initial weight value of the node in the next round is adjusted according to the contribution.

[0074] Obtain the final node weight distribution plan and feed it back to the aggregation system.

[0075] In a second aspect of the present invention, a project monitoring system based on big data is provided, the system comprising:

[0076] A project data acquisition unit is used to acquire the project data of the required detection project, preprocess the project data, construct a project data set based on the preprocessed project data and upload it to the cloud server for storage, extract and fuse the features of the data set, and construct a project scenario model based on the fused multi-modal features;

[0077] The project monitoring unit is used to apply the dynamic weighted federated average algorithm to perform blockchain-based adaptive distributed joint training on the project scenario model, update the project scenario model, dynamically adjust the weight of the project scenario model update according to the real-time data of each participating node, and update the aggregated weighted project scenario model to obtain the global project scenario model;

[0078] The project adjustment unit is used to build a deep reinforcement learning model, adjust the forecast project monitoring strategy according to the real-time status and historical data of the project, judge the accuracy of the forecast project strategy in combination with the progress status of the current project, and introduce a correction plan based on the accuracy of the forecast project strategy to dynamically adjust the forecast project monitoring strategy to obtain the project monitoring strategy;

[0079] The project optimization unit is used to perform monitoring tasks according to the adjusted monitoring strategy, feed back the monitoring results to the system, and optimize the monitoring strategy according to the test results;

[0080] The step of judging the accuracy of the predicted project strategy in combination with the progress status of the current project specifically includes:

[0081] The current status of the project is obtained based on the degree of deviation between the actual project progress and the expected progress, and the specific reasons for the deviation of the project progress are determined by the completion of the key milestones of the project:

[0082] If the actual investment of project resources matches the plan, the impact of resource allocation rationality on schedule deviation is considered. The project risk identification and response effect are used to analyze the impact of potential risks on project schedule. The impact of quality control on project schedule is evaluated by the achievement of project quality goals.

[0083] If the project budget execution is highly compliant, then examine the contribution of budget control to schedule deviation; obtain the satisfaction evaluation results of project stakeholders and determine the feedback of the results on project strategy adjustment; predict the accuracy of project strategy based on the analysis results and obtain suggestions for strategy adjustment;

[0084] If the actual investment of project resources and the project budget obtained do not match the forecast, the strategy will be re-evaluated and a new adjustment strategy will be determined.

[0085] The beneficial technical effects of the present invention are at least as follows:

[0086] The present invention discloses a project monitoring method and system based on big data, which applies a dynamic weighted federated average algorithm to perform adaptive distributed joint training on project scenario models on the blockchain. Through a deep reinforcement learning model, the prediction strategy is adjusted according to the real-time status and historical data of the project, and a correction plan is introduced to dynamically optimize the monitoring strategy. During the training process, the present invention uses a dynamic weighting mechanism to calculate the node weight according to the quality and quantity of each node data, and updates the model parameters by weighted average accordingly. This method can effectively integrate multi-party data, improve model accuracy and generalization ability, and protect data privacy at the same time. By dynamically adjusting weights and strategies, the present invention can adapt to changes in the project environment, achieve accurate monitoring and early warning, and provide strong support for project management decisions. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0088] Figure 1 This is a flow chart of a project monitoring method based on big data in the present invention.

[0089] Figure 2 This is a framework diagram of a project monitoring system based on big data in the present invention. DETAILED DESCRIPTION

[0090] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0091] like Figure 1 As shown, an embodiment of the present invention provides a project monitoring method based on big data, and the method includes the following steps S1-S4.

[0092] S1. Obtain the project data of the required detection project, preprocess the project data, build a project data set based on the preprocessed project data and upload it to the cloud server for storage, extract and fuse the data set, and build a project scenario model based on the fused multimodal features;

[0093] S2. Apply the dynamic weighted federated average algorithm to conduct blockchain-based adaptive distributed joint training on the project scenario model, update the project scenario model, dynamically adjust the weight of the project scenario model update according to the real-time data of each participating node, and update the aggregated weighted project scenario model to obtain the global project scenario model;

[0094] S3. Build and use a deep reinforcement learning model to adjust the forecast project monitoring strategy according to the real-time status and historical data of the project, judge the accuracy of the forecast project strategy in combination with the progress status of the current project, and introduce a correction plan based on the accuracy of the forecast project strategy to dynamically adjust the forecast project monitoring strategy to obtain the project monitoring strategy;

[0095] S4. Perform monitoring tasks according to the adjusted monitoring strategy, feed back monitoring results to the system, and optimize the monitoring strategy according to the detection results;

[0096] The step of judging the accuracy of the predicted project strategy in combination with the progress status of the current project specifically includes:

[0097] The current status of the project is obtained based on the degree of deviation between the actual project progress and the expected progress, and the specific reasons for the deviation of the project progress are determined by the completion of the key milestones of the project:

[0098] If the actual investment of project resources matches the plan, the impact of resource allocation rationality on schedule deviation is considered. The project risk identification and response effect are used to analyze the impact of potential risks on project schedule. The impact of quality control on project schedule is evaluated by the achievement of project quality goals.

[0099] If the project budget execution is highly compliant, then examine the contribution of budget control to schedule deviation; obtain the satisfaction evaluation results of project stakeholders and determine the feedback of the results on project strategy adjustment; predict the accuracy of project strategy based on the analysis results and obtain suggestions for strategy adjustment;

[0100] If the actual investment of project resources and the project budget obtained do not match the forecast, the strategy will be re-evaluated and a new adjustment strategy will be determined.

[0101] Furthermore, the dataset is subjected to feature extraction and fusion, and a project scenario model is constructed according to the fused multimodal features, including:

[0102] If time attribute data is obtained, then the timestamp, date, and time period information are extracted to obtain the time characteristics of the data. According to the spatial attribute information, the geographic location, coordinates, and regional data are used to determine the spatial characteristics of the data. Through the user attribute data, the user ID, user behavior, and user preference are determined to obtain the characteristic information related to the user.

[0103] If analyzing content attribute data, then extract text content, image content, and audio content to obtain the specific content characteristics of the data; according to statistical attributes, use mean, variance, maximum value, and minimum value to determine the statistical characteristics of the data; through relationship attribute information, determine association rules, network relationships, and hierarchical relationships to obtain the relationship characteristics between data;

[0104] If environmental attribute data is obtained, the weather, temperature, and humidity information are extracted to obtain the environmental characteristics when the data occurs; based on the device attribute information, the device type, device status, and device configuration are used to determine the characteristics of the data collection device; through the behavioral attributes and event attribute data, the click, browse, purchase behavior and event type, event frequency and duration are judged to obtain user behavior and event characteristics in specific scenarios.

[0105] Furthermore, the application of the dynamic weighted federated average algorithm to perform blockchain-based adaptive distributed joint training on the project scenario model to update the project scenario model includes:

[0106] The data of each participating node has been pre-processed locally, and the nodes are securely connected through the blockchain network;

[0107] According to the quality and quantity of local data, the weight of each node is calculated using a dynamic weighting mechanism to obtain a weight list of nodes participating in the training; wherein the weight is expressed as follows:

[0108]

[0109] in, represents the weight of node i, represents the data quality of node i, represents the amount of data for node i, K represents the total number of nodes involved in training, k represents node k, represents the data quality of node k, Indicates the amount of data for node k;

[0110] The node performs weighted average calculation on the model parameters according to the weight list to obtain updated local model parameters; if it is found during the training process that the model accuracy of the node does not meet the expected indicators, the model parameters are dynamically modified or the training strategy is optimized through adaptive adjustment strategy to obtain optimized model parameters; wherein, the model parameters are expressed as follows:

[0111]

[0112] in, represents the global model parameters after t+1 rounds of training, i represents node i, represents the global model parameters of node i after t rounds of training; where the loss function of node i is It is expressed as follows:

[0113]

[0114] in, represents the number of data samples of node i, j represents the jth sample, represents the true label of the jth sample, represents the input features of the jth sample, represents the model parameters of node i, represents the model function;

[0115] Through the consensus mechanism of the blockchain network, the node broadcasts the optimized model parameters to other nodes, determines whether other nodes accept the model update, and obtains the model version recognized by consensus, which is expressed as follows:

[0116]

[0117] in, Represents the consensus result, Indicates a broadcast operation. represents the global model parameters after t+1 rounds of training;

[0118] Get decentralized training control flow;

[0119] Through the transparent logs recorded in the blockchain, the node obtains detailed records of each model update from the chain; if the business needs change, the number and configuration of nodes participating in the training are adjusted according to the scalability attributes to obtain the node group that meets the current business scale, as shown below:

[0120]

[0121] in, Represents a node group, Indicates the nodes involved in training, and M indicates the number of nodes under the current business scale;

[0122] Through the federated learning algorithm, the final model parameters of each node are integrated to obtain the global model, which is stored on the blockchain.

[0123] Furthermore, dynamically adjusting the weight of updating the project scenario model according to the real-time data of each participating node, and updating the aggregated weighted project scenario model to obtain a global project scenario model, includes:

[0124] Based on the latest data provided by each participating node, the following analysis is performed:

[0125] If the node data is valid, the system will dynamically adjust the weight of each node based on the node contribution and data accuracy;

[0126] If the weight adjustment parameters are determined, the system will apply these parameters to weight the project scenario models of each node to obtain the weighted results of each model;

[0127] If the weighted results of each model are obtained, the system obtains a preliminary global project scenario model through the aggregate weighting process, which is expressed as follows:

[0128]

[0129] in, represents the weighted global project scenario model, represents the project scenario model of node i;

[0130] Based on the preliminary global project scenario model obtained, the system will conduct a model evaluation to determine whether the model's performance indicators meet the preset standards, including:

[0131] If the model evaluation is judged to be below the standard, the system will obtain feedback information from each node on the global model through the feedback mechanism;

[0132] If the model parameters are adjusted through the feedback mechanism, the system will determine the model update frequency again;

[0133] If the model update frequency is determined, the system uses a communication protocol for data synchronization;

[0134] After data synchronization is complete, continue to maintain the parameters of the global project scenario model.

[0135] Furthermore, the construction uses a deep reinforcement learning model to adjust the predictive project monitoring strategy based on the real-time status and historical data of the project, including:

[0136] According to the real-time status and historical data of the project, the initial input parameters of the project are obtained by acquiring the basic information of the project; wherein the basic information of the project includes the project name, type, scale, budget and planned duration;

[0137] By analyzing the initial input parameters and combining the completed workload, actual cost, schedule deviation and resource usage in the real-time status of the project, the current progress of the project can be judged, including:

[0138] If the progress of the project deviates from expectations, compare historical data, combine the execution status, risk events, response measures and effects of similar projects, determine the potential risks faced by the current project and their impact, and obtain the additional impact on the project based on changes in external environmental factors, including changes in policies and regulations, market demand fluctuations and changes in raw material prices, and determine the specific impact of these factors on the project progress;

[0139] By analyzing the project organizational structure and personnel capabilities, combined with team member composition, skill level and work performance, the project team's execution capabilities are evaluated to determine whether team adjustments or skill enhancements are needed;

[0140] Adopt project management systems and tools, combine management systems, process specifications, information systems and communication and coordination mechanisms, optimize project management processes and information flows, and ensure the effectiveness of information transmission within the project team;

[0141] Build a deep reinforcement learning model and set the parameters of the reinforcement learning model, including state space, action space, reward function, learning rate and discount factor, so that the deep reinforcement learning model can learn the dynamic changes of the project;

[0142] The definitions are as follows:

[0143]

[0144]

[0145] in, represents the state space, represents the nth state, represents the action space, Indicates the nth action;

[0146] Defining the reward function as follows:

[0147]

[0148] in, represents the reward function for taking action a in state s, represents the reward obtained at time t+1, and E represents the expectation;

[0149] The deep learning model is:

[0150]

[0151] in, represents the Q value of taking action a in state s, represents the discount factor, represents the reward obtained at time t+1+k;

[0152] The strategy is optimized as follows:

[0153]

[0154] in, represents the probability of selecting action a in state s, represents the temperature parameter;

[0155] The loss function is:

[0156]

[0157] in, represents the loss function, represents the model parameters, r represents the reward, represents the discount factor, represents the Q value of taking action a in state s, Represents the parameters of the target network; the parameters are updated as follows:

[0158]

[0159] in, represents the update amount of the model parameters, represents the learning rate, Represents the loss function L(\theta) on the model parameters gradient.

[0160] Through deep reinforcement learning model training and evaluation, the deep reinforcement learning model is trained using training data sets, validation data sets, combined with evaluation indicators and number of iterations to obtain its prediction accuracy in different project scenarios;

[0161] If the training results of the deep reinforcement learning model meet expectations, the prediction results are interpreted and applied, the results are visualized, and the optimization direction of the project monitoring strategy is determined through decision-making recommendations and adjustment plans. If a risk warning is obtained through the prediction results, the project monitoring strategy is adjusted in real time to ensure that the project always remains on the expected track under changes in the external environment and internal resource adjustments.

[0162] Furthermore, the method of introducing a correction plan according to the accuracy of the forecast project strategy to dynamically adjust the forecast project monitoring strategy to obtain the project monitoring strategy includes:

[0163] If the prediction accuracy of the project strategy is higher than the set threshold, obtain the parameter adjustment plan of the prediction model:

[0164] By analyzing the coverage of the forecast project strategy, we can obtain the areas with low coverage and determine the parameter weights of these areas in the forecast model; if the adjustment range of the revised plan exceeds the predetermined standard, determine the impact of the adjustment range on the forecast accuracy and obtain the impact assessment results; according to the adjustment frequency of the revised plan, obtain the correlation data between the adjustment frequency and the forecast accuracy and determine the impact of the frequency on the accuracy; use the dynamic update cycle data of the project monitoring strategy to determine the effect of the update cycle on the monitoring accuracy; by analyzing the monitoring indicators of the project monitoring strategy, obtain the key monitoring indicators related to the forecast project and determine the relationship between the indicators and the forecast accuracy; according to the early warning threshold of the project monitoring strategy, obtain the impact of the threshold setting on the effectiveness of the project early warning; through the feedback mechanism of the project monitoring strategy, determine whether the feedback mechanism is perfect and its effect on the forecast accuracy and dynamic adjustment of the monitoring strategy; according to the accuracy of the forecast project strategy and the results of the dynamic adjustment of the revised plan, obtain the final adjustment plan of the project monitoring strategy;

[0165] Comprehensively determine the implementation plan of the project monitoring strategy to form a complete predictive monitoring process.

[0166] Furthermore, the parameter adjustment scheme of the prediction model includes:

[0167] If the input dataset is large, calculate efficiency and model convergence speed by determining an appropriate batch size;

[0168] Use the initial learning rate to obtain the initial performance indicators of the training model, and observe whether the learning rate needs to be adjusted;

[0169] By adding momentum terms, determine whether the convergence speed of the model can be improved;

[0170] Adjust the number of hidden layer units according to the model's expressiveness requirements;

[0171] Select different optimizers, determine their impact on model performance, and determine the final optimizer solution based on actual performance.

[0172] Furthermore, the weight of each node is calculated using a dynamic weighting mechanism to obtain a weight list of nodes participating in the training, including:

[0173] Obtaining current status information of at least one node, and if the current status of the node is online and fault-free, evaluating the node, otherwise skipping the node and obtaining status information of the next node;

[0174] Evaluate the computing power of the node, obtain the CPU and GPU performance indicators of the node, and determine the computing power score of the node according to the preset computing power evaluation rules for subsequent calculation of the node weight;

[0175] Acquire data quality information of the node, the data quality information including data volume and annotation accuracy, and determine the data quality score of the node according to a preset data quality assessment rule;

[0176] Obtaining historical performance data of the node, including model accuracy and convergence speed, and determining a historical performance score of the node according to a preset historical performance evaluation rule;

[0177] Scoring based on the computing power of the node , data quality score and historical performance ratings , using the preset weight calculation formula to obtain the preliminary weight value of the node , which is expressed as follows:

[0178]

[0179] in, , and Score computing ability , data quality score and historical performance ratings The weight of

[0180] Get the current status information of the node, if the node is online and the computing power score of the node Above the preset threshold , then adjust the preliminary weight value of the node according to the preset weight adjustment rule; wherein the weight adjustment rule is expressed as follows:

[0181]

[0182] in, represents the updated weight, represents the weight coefficient;

[0183] For the preliminary weight values ​​of all nodes, a preset normalization method is used to obtain the standardized weight value of each node;

[0184] According to the calculated node weight value list, the preset dynamic weighting mechanism is used to sort the nodes participating in the training, and finally determine the weight distribution scheme of each node in the model aggregation process.

[0185] It can be understood that the dynamic weighting mechanism is:

[0186] A dynamic weighting mechanism is set according to business objectives. First, the initial weight value configuration is obtained to obtain the initial importance of each object.

[0187] If the weight distribution method is statically fixed, the initial configuration is used directly;

[0188] If it is a dynamic change, select a dynamic allocation method based on business needs;

[0189] Use historical data and current environment information to determine the most suitable weight adjustment strategy and obtain the weight value to be adjusted;

[0190] According to the determined weight update frequency, set the weight adjustment period and update the weight value in real time or in stages;

[0191] By collecting feedback results and environmental change information, a weighted calculation method is performed to obtain a weighted average or product value;

[0192] If normalization is required, the weight value range is processed according to the normalization algorithm so that it is between 0 and 1;

[0193] For the adjusted weight values, further analyze the impact of weight changes on business objectives and obtain feedback results;

[0194] Based on the feedback results and the set adjustment strategy, determine whether the weight needs to be readjusted. If adjustment is required, enter the next cycle of weight adjustment;

[0195] If the impact of the weight adjustment meets business expectations, a feedback correction process using a dynamic weighting mechanism is adopted to obtain the final impact assessment of the adjusted weight on business objectives;

[0196] Analyze the final evaluation results and use information processing technology to obtain new business goal adjustment suggestions;

[0197] If the business goal adjustment suggestion is reasonable, the business goal is updated to form a new business strategy input;

[0198] Combined with the new business policy input, reset the initial weight value configuration to ensure that the importance of each object is consistent with the updated business policy;

[0199] The entire dynamic weighting mechanism process is automatically detected through technical means to ensure effective connection and data flow in each link.

[0200] Furthermore, the node weight value list obtained by calculation adopts a preset dynamic weighting mechanism to sort the nodes participating in the training, and finally determines the weight distribution scheme of each node in the model aggregation process, including:

[0201] According to the calculated node weight value list, the preset dynamic weighting mechanism is used to sort the nodes participating in the training, and finally determine the weight distribution scheme of each node in the model aggregation process;

[0202] According to the node ID, obtain the node weight value list, and obtain the preliminary weight data of each node through the data processing mechanism;

[0203] Determine the status of each node. If the node status is normal, retain the node weight value:

[0204] If the node status is abnormal, the node weight value is adjusted to zero. The nodes are sorted by the node weight value to obtain a node sequence arranged from large to small according to the weight value. According to the sorted node sequence, the cumulative weight sum of each node is calculated to obtain a cumulative weight list. A dynamic weighting mechanism is used to allocate the corresponding aggregation weight percentage according to the cumulative weight list to obtain the weight distribution ratio of each node. According to the aggregation round, the participation of the current node is judged. If the node participates in the current round, its weight distribution ratio is retained;

[0205] If the node does not participate, the weight distribution ratio is set to zero. The aggregation weight is distributed according to the node weight distribution ratio to obtain the actual weight of each node in the model aggregation process. According to the actual weight, the contribution of the node in the model aggregation process is judged, and the initial weight value of the node in the next round is adjusted according to the contribution.

[0206] Obtain the final node weight distribution plan and feed it back to the aggregation system.

[0207] To better implement the above method, Figure 2 As shown, the present invention also proposes a project monitoring system based on big data, the system comprising:

[0208] The project data acquisition unit 101 is used to acquire the project data of the required detection project, pre-process the project data, construct a project data set according to the pre-processed project data and upload it to the cloud server for storage, extract and fuse the features of the data set, and construct a project scenario model according to the fused multi-modal features;

[0209] The project monitoring unit 102 is used to apply the dynamic weighted federated average algorithm to perform blockchain-based adaptive distributed joint training on the project scenario model, update the project scenario model, dynamically adjust the weight of the project scenario model update according to the real-time data of each participating node, and update the aggregated weighted project scenario model to obtain a global project scenario model;

[0210] The project adjustment unit 103 is used to construct and use a deep reinforcement learning model to adjust the predicted project monitoring strategy according to the real-time status and historical data of the project, judge the accuracy of the predicted project strategy in combination with the progress status of the current project, and introduce a correction plan according to the accuracy of the predicted project strategy to dynamically adjust the predicted project monitoring strategy to obtain the project monitoring strategy;

[0211] The project optimization unit 104 is used to perform monitoring tasks according to the adjusted monitoring strategy, feed back the monitoring results to the system, and optimize the monitoring strategy according to the detection results;

[0212] The step of judging the accuracy of the predicted project strategy in combination with the progress status of the current project specifically includes:

[0213] The current status of the project is obtained based on the degree of deviation between the actual project progress and the expected progress, and the specific reasons for the deviation of the project progress are determined by the completion of the key milestones of the project:

[0214] If the actual investment of project resources matches the plan, the impact of resource allocation rationality on schedule deviation is considered. The project risk identification and response effect are used to analyze the impact of potential risks on project schedule. The impact of quality control on project schedule is evaluated by the achievement of project quality goals.

[0215] If the project budget execution is highly compliant, then examine the contribution of budget control to schedule deviation; obtain the satisfaction evaluation results of project stakeholders and determine the feedback of the results on project strategy adjustment; predict the accuracy of project strategy based on the analysis results and obtain suggestions for strategy adjustment;

[0216] If the actual investment of project resources and the project budget obtained do not match the forecast, the strategy will be re-evaluated and a new adjustment strategy will be determined.

[0217] Unless otherwise specifically stated, the relative steps, numerical expressions and values ​​of the components and steps set forth in these embodiments do not limit the scope of the present application.

[0218] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a non-volatile computer-readable storage medium that is executable by a processor. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0219] In the description of the present application, it should be noted that the terms "upper", "lower", etc. indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, or the orientations or positional relationships in which the inventive product is usually placed when in use. These are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present application.

[0220] In the description of this application, it should also be noted that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0221] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A project monitoring method based on big data, characterized in that: The method comprises: S1. Obtain the project data of the required detection project, preprocess the project data, build a project data set based on the preprocessed project data and upload it to the cloud server for storage, extract and fuse the data set, and build a project scenario model based on the fused multimodal features; S2. Apply the dynamic weighted federated average algorithm to conduct blockchain-based adaptive distributed joint training on the project scenario model, update the project scenario model, dynamically adjust the weight of the project scenario model update according to the real-time data of each participating node, and update the aggregated weighted project scenario model to obtain the global project scenario model; S3. Build and use a deep reinforcement learning model to adjust the forecast project monitoring strategy according to the real-time status and historical data of the project, judge the accuracy of the forecast project strategy in combination with the progress status of the current project, and introduce a correction plan based on the accuracy of the forecast project strategy to dynamically adjust the forecast project monitoring strategy to obtain the project monitoring strategy; S4. Perform monitoring tasks according to the adjusted monitoring strategy, feed back monitoring results to the system, and optimize the monitoring strategy according to the detection results; The step of judging the accuracy of the predicted project strategy in combination with the progress status of the current project specifically includes: The current status of the project is obtained based on the degree of deviation between the actual progress of the project and the expected progress. The specific reasons for the deviation of the project progress are determined by the completion of key milestones of the project: If the actual investment of project resources matches the plan, the impact of resource allocation rationality on schedule deviation is considered, and the impact of potential risks on project schedule is analyzed by project risk identification and response effectiveness; the impact of quality control on project schedule is evaluated by the achievement of project quality goals; If the project budget execution is highly compliant, then examine the contribution of budget control to schedule deviation; obtain the satisfaction evaluation results of project stakeholders and determine the feedback of the results on project strategy adjustment; predict the accuracy of project strategy based on the analysis results and obtain suggestions for strategy adjustment; If the actual input of project resources and the project budget obtained do not match the forecast, re-evaluate the strategy and determine a new adjustment strategy; The application of the dynamic weighted federated average algorithm to perform blockchain-based adaptive distributed joint training on the project scenario model to update the project scenario model includes: The data of each participating node has been pre-processed locally, and the nodes are securely connected through the blockchain network; According to the quality and quantity of local data, the weight of each node is calculated using a dynamic weighting mechanism to obtain a weight list of nodes participating in training; The node performs weighted average calculation on the model parameters according to the weight list to obtain updated local model parameters; if it is found during the training process that the model accuracy of the node does not meet the expected indicators, the model parameters are dynamically modified or the training strategy is optimized through adaptive adjustment strategy to obtain optimized model parameters; Through the consensus mechanism of the blockchain network, the node broadcasts the optimized model parameters to other nodes, determines whether other nodes accept the model update, and obtains the model version recognized by consensus; Get decentralized training control flow; Through the transparent logs recorded in the blockchain, nodes obtain detailed records of each model update from the chain; if business needs change, the number and configuration of nodes participating in training are adjusted according to scalability properties to obtain a node group that meets the current business scale; Through the federated learning algorithm, the final model parameters of each node are integrated to obtain the global model, which is stored on the blockchain; The method of dynamically adjusting the weight of updating the project scenario model according to the real-time data of each participating node, and updating the aggregated weighted project scenario model to obtain a global project scenario model, includes: Based on the latest data provided by each participating node, the following analysis is performed: If the node data is valid, the system will dynamically adjust the weight of each node based on the node contribution and data accuracy; If the weight adjustment parameters are determined, the system will apply these parameters to weight the project scenario models of each node to obtain the weighted results of each model; If the weighted results of each model are obtained, the system obtains a preliminary global project scenario model through the aggregate weighting process; Based on the preliminary global project scenario model obtained, the system will conduct a model evaluation to determine whether the model's performance indicators meet the preset standards, including: If the model evaluation is judged to be below the standard, the system will obtain feedback information from each node on the global model through the feedback mechanism; If the model parameters are adjusted through the feedback mechanism, the system will determine the model update frequency again; If the model update frequency is determined, the system uses a communication protocol for data synchronization; After data synchronization is completed, continue to maintain the parameters of the global project scenario model; The construction uses a deep reinforcement learning model to adjust the predictive project monitoring strategy based on the real-time status and historical data of the project, including: According to the real-time status and historical data of the project, the initial input parameters of the project are obtained by acquiring the basic information of the project; wherein the basic information of the project includes the project name, type, scale, budget and planned duration; By analyzing the initial input parameters and combining the completed workload, actual cost, schedule deviation and resource usage in the real-time status of the project, the current progress of the project can be judged, including: If the progress of the project deviates from expectations, compare historical data, combine the execution status, risk events, response measures and effects of similar projects, determine the potential risks faced by the current project and their impact, and obtain the additional impact on the project based on changes in external environmental factors, including changes in policies and regulations, market demand fluctuations and changes in raw material prices, and determine the specific impact of these factors on the project progress; By analyzing the project organizational structure and personnel capabilities, combined with team member composition, skill level and work performance, the project team's execution capabilities are evaluated to determine whether team adjustments or skill enhancements are needed; Adopt project management systems and tools, combine management systems, process specifications, information systems and communication and coordination mechanisms, optimize project management processes and information flows, and ensure the effectiveness of information transmission within the project team; Build a deep reinforcement learning model and set the parameters of the reinforcement learning model, including state space, action space, reward function, learning rate and discount factor, so that the deep reinforcement learning model can learn the dynamic changes of the project; Through deep reinforcement learning model training and evaluation, the deep reinforcement learning model is trained using training data sets, validation data sets, combined with evaluation indicators and number of iterations to obtain its prediction accuracy in different project scenarios; If the training results of the deep reinforcement learning model meet expectations, the prediction results are interpreted and applied to visualize the results, and the optimization direction of the project monitoring strategy is determined through decision-making suggestions and adjustment plans; if risk warnings are obtained through the prediction results, the project monitoring strategy is adjusted in real time to ensure that the project always runs within the expected track under changes in the external environment and adjustments in internal resources; The method of introducing a correction plan according to the accuracy of the forecast project strategy and dynamically adjusting the forecast project monitoring strategy to obtain the project monitoring strategy includes: If the prediction accuracy of the project strategy is higher than the set threshold, obtain the parameter adjustment plan of the prediction model: By analyzing the coverage of the forecast project strategy, we can obtain the areas with low coverage and determine the parameter weights of these areas in the forecast model; if the adjustment range of the revised plan exceeds the predetermined standard, determine the impact of the adjustment range on the forecast accuracy and obtain the impact assessment results; according to the adjustment frequency of the revised plan, obtain the correlation data between the adjustment frequency and the forecast accuracy and determine the impact of the frequency on the accuracy; use the dynamic update cycle data of the project monitoring strategy to determine the effect of the update cycle on the monitoring accuracy; by analyzing the monitoring indicators of the project monitoring strategy, obtain the key monitoring indicators related to the forecast project and determine the relationship between the indicators and the forecast accuracy; according to the early warning threshold of the project monitoring strategy, obtain the impact of the threshold setting on the effectiveness of the project early warning; through the feedback mechanism of the project monitoring strategy, determine whether the feedback mechanism is perfect and its effect on the forecast accuracy and dynamic adjustment of the monitoring strategy; according to the accuracy of the forecast project strategy and the results of the dynamic adjustment of the revised plan, obtain the final adjustment plan of the project monitoring strategy; Comprehensively determine the implementation plan of the project monitoring strategy to form a complete predictive monitoring process.

2. A project monitoring method based on big data according to claim 1, characterized in that: The data set is subjected to feature extraction and fusion, and a project scenario model is constructed based on the fused multimodal features, including: If time attribute data is obtained, then the timestamp, date, and time period information are extracted to obtain the time characteristics of the data. According to the spatial attribute information, the geographic location, coordinates, and regional data are used to determine the spatial characteristics of the data. Through the user attribute data, the user ID, user behavior, and user preference are determined to obtain the characteristic information related to the user. If analyzing content attribute data, then extract text content, image content, and audio content to obtain the specific content characteristics of the data; according to statistical attributes, use mean, variance, maximum value, and minimum value to determine the statistical characteristics of the data; through relationship attribute information, determine association rules, network relationships, and hierarchical relationships to obtain the relationship characteristics between data; If environmental attribute data is obtained, the weather, temperature, and humidity information are extracted to obtain the environmental characteristics when the data occurs; based on the device attribute information, the device type, device status, and device configuration are used to determine the characteristics of the data collection device; through the behavioral attributes and event attribute data, the click, browse, purchase behavior and event type, event frequency and duration are judged to obtain user behavior and event characteristics in specific scenarios.

3. A project monitoring method based on big data according to claim 1, characterized in that: The parameter adjustment scheme of the prediction model includes: If the input dataset is large, calculate the efficiency and model convergence speed by determining the appropriate batch size; Use the initial learning rate to obtain the initial performance indicators of the training model, and observe whether the learning rate needs to be adjusted; By adding momentum terms, determine whether the convergence speed of the model can be improved; Adjust the number of hidden layer units according to the model's expressiveness requirements; Select different optimizers, determine their impact on model performance, and determine the final optimizer solution based on actual performance.

4. A project monitoring method based on big data according to claim 3, characterized in that: The dynamic weighting mechanism is used to calculate the weight of each node to obtain a weight list of nodes participating in the training, including: Obtaining current status information of at least one node, and if the current status of the node is online and fault-free, evaluating the node, otherwise skipping the node and obtaining status information of the next node; Evaluate the computing power of the node, obtain the CPU and GPU performance indicators of the node, and determine the computing power score of the node according to the preset computing power evaluation rules for subsequent calculation of the node weight; Acquire data quality information of the node, the data quality information including data volume and annotation accuracy, and determine the data quality score of the node according to a preset data quality assessment rule; Obtaining historical performance data of the node, including model accuracy and convergence speed, and determining a historical performance score of the node according to a preset historical performance evaluation rule; Scoring based on the computing power of the node , data quality score and historical performance ratings , using the preset weight calculation formula to obtain the preliminary weight value of the node , which is expressed as follows: in, , and Score computing ability , data quality score and historical performance ratings The weight of Get the current status information of the node, if the node is online and the computing power score of the node Above the preset threshold , then adjust the preliminary weight value of the node according to the preset weight adjustment rule; wherein the weight adjustment rule is expressed as follows: in, represents the updated weight, represents the weight coefficient; For the preliminary weight values ​​of all nodes, a preset normalization method is used to obtain the standardized weight value of each node; According to the calculated node weight value list, the preset dynamic weighting mechanism is used to sort the nodes participating in the training, and finally determine the weight distribution scheme of each node in the model aggregation process.

5. A project monitoring method based on big data according to claim 4, characterized in that: The node weight value list obtained by calculation is used to sort the nodes participating in the training using a preset dynamic weighting mechanism, and finally determine the weight distribution scheme of each node in the model aggregation process, including: According to the calculated node weight value list, the preset dynamic weighting mechanism is used to sort the nodes participating in the training, and finally determine the weight distribution scheme of each node in the model aggregation process; According to the node ID, obtain the node weight value list, and obtain the preliminary weight data of each node through the data processing mechanism; Determine the status of each node. If the node status is normal, retain the node weight value: If the node status is abnormal, the node weight value is adjusted to zero. The nodes are sorted by the node weight value to obtain a node sequence arranged from large to small according to the weight value. According to the sorted node sequence, the cumulative weight sum of each node is calculated to obtain a cumulative weight list. A dynamic weighting mechanism is used to allocate the corresponding aggregation weight percentage according to the cumulative weight list to obtain the weight distribution ratio of each node. According to the aggregation round, the participation of the current node is judged. If the node participates in the current round, its weight distribution ratio is retained; If the node does not participate, the weight distribution ratio is set to zero. The aggregation weight is distributed according to the node weight distribution ratio to obtain the actual weight of each node in the model aggregation process. According to the actual weight, the contribution of the node in the model aggregation process is judged, and the initial weight value of the node in the next round is adjusted according to the contribution. Obtain the final node weight distribution plan and feed it back to the aggregation system.

6. A project monitoring system based on big data, characterized in that: The system comprises: A project data acquisition unit is used to acquire the project data of the required detection project, preprocess the project data, construct a project data set based on the preprocessed project data and upload it to the cloud server for storage, extract and fuse the features of the data set, and construct a project scenario model based on the fused multi-modal features; The project monitoring unit is used to apply the dynamic weighted federated average algorithm to perform blockchain-based adaptive distributed joint training on the project scenario model, update the project scenario model, dynamically adjust the weight of the project scenario model update according to the real-time data of each participating node, and update the aggregated weighted project scenario model to obtain the global project scenario model; The project adjustment unit builds and uses a deep reinforcement learning model to adjust the forecast project monitoring strategy according to the real-time status and historical data of the project, judge the accuracy of the forecast project strategy in combination with the progress status of the current project, and introduce a correction plan based on the accuracy of the forecast project strategy to dynamically adjust the forecast project monitoring strategy to obtain the project monitoring strategy; The project optimization unit performs monitoring tasks according to the adjusted monitoring strategy, feeds back the monitoring results to the system, and optimizes the monitoring strategy based on the test results; The step of judging the accuracy of the predicted project strategy in combination with the progress status of the current project specifically includes: The current status of the project is obtained based on the degree of deviation between the actual progress of the project and the expected progress. The specific reasons for the deviation of the project progress are determined by the completion of key milestones of the project: If the actual investment of project resources matches the plan, the impact of resource allocation rationality on schedule deviation is considered, and the impact of potential risks on project schedule is analyzed by project risk identification and response effectiveness; the impact of quality control on project schedule is evaluated by the achievement of project quality goals; If the project budget execution is highly compliant, then examine the contribution of budget control to schedule deviation; obtain the satisfaction evaluation results of project stakeholders and determine the feedback of the results on project strategy adjustment; predict the accuracy of project strategy based on the analysis results and obtain suggestions for strategy adjustment; If the actual input of project resources and the project budget obtained do not match the forecast, re-evaluate the strategy and determine a new adjustment strategy; The application of the dynamic weighted federated average algorithm to perform blockchain-based adaptive distributed joint training on the project scenario model to update the project scenario model includes: The data of each participating node has been pre-processed locally, and the nodes are securely connected through the blockchain network; According to the quality and quantity of local data, the weight of each node is calculated using a dynamic weighting mechanism to obtain a weight list of nodes participating in training; The node performs weighted average calculation on the model parameters according to the weight list to obtain updated local model parameters; if it is found during the training process that the model accuracy of the node does not meet the expected indicators, the model parameters are dynamically modified or the training strategy is optimized through adaptive adjustment strategy to obtain optimized model parameters; Through the consensus mechanism of the blockchain network, the node broadcasts the optimized model parameters to other nodes, determines whether other nodes accept the model update, and obtains the model version recognized by consensus; Get decentralized training control flow; Through the transparent logs recorded in the blockchain, nodes obtain detailed records of each model update from the chain; if business needs change, the number and configuration of nodes participating in training are adjusted according to scalability properties to obtain a node group that meets the current business scale; Through the federated learning algorithm, the final model parameters of each node are integrated to obtain the global model, which is stored on the blockchain; The method of dynamically adjusting the weight of updating the project scenario model according to the real-time data of each participating node, and updating the aggregated weighted project scenario model to obtain a global project scenario model, includes: Based on the latest data provided by each participating node, the following analysis is performed: If the node data is valid, the system will dynamically adjust the weight of each node based on the node contribution and data accuracy; If the weight adjustment parameters are determined, the system will apply these parameters to weight the project scenario models of each node to obtain the weighted results of each model; If the weighted results of each model are obtained, the system obtains a preliminary global project scenario model through the aggregate weighting process; Based on the preliminary global project scenario model obtained, the system will conduct a model evaluation to determine whether the model's performance indicators meet the preset standards, including: If the model evaluation is judged to be below the standard, the system will obtain feedback information from each node on the global model through the feedback mechanism; If the model parameters are adjusted through the feedback mechanism, the system will determine the model update frequency again; If the model update frequency is determined, the system uses a communication protocol for data synchronization; After data synchronization is completed, continue to maintain the parameters of the global project scenario model; The construction uses a deep reinforcement learning model to adjust the predictive project monitoring strategy based on the real-time status and historical data of the project, including: According to the real-time status and historical data of the project, the initial input parameters of the project are obtained by acquiring the basic information of the project; wherein the basic information of the project includes the project name, type, scale, budget and planned duration; By analyzing the initial input parameters and combining the completed workload, actual cost, schedule deviation and resource usage in the real-time status of the project, the current progress of the project can be judged, including: If the progress of the project deviates from expectations, compare historical data, combine the execution status, risk events, response measures and effects of similar projects, determine the potential risks faced by the current project and their impact, and obtain the additional impact on the project based on changes in external environmental factors, including changes in policies and regulations, market demand fluctuations and changes in raw material prices, and determine the specific impact of these factors on the project progress; By analyzing the project organizational structure and personnel capabilities, combined with team member composition, skill level and work performance, the project team's execution capabilities are evaluated to determine whether team adjustments or skill enhancements are needed; Adopt project management systems and tools, combine management systems, process specifications, information systems and communication and coordination mechanisms, optimize project management processes and information flows, and ensure the effectiveness of information transmission within the project team; Build a deep reinforcement learning model and set the parameters of the reinforcement learning model, including state space, action space, reward function, learning rate and discount factor, so that the deep reinforcement learning model can learn the dynamic changes of the project; Through deep reinforcement learning model training and evaluation, the deep reinforcement learning model is trained using training data sets, validation data sets, combined with evaluation indicators and number of iterations to obtain its prediction accuracy in different project scenarios; If the training results of the deep reinforcement learning model meet expectations, the prediction results are interpreted and applied to visualize the results, and the optimization direction of the project monitoring strategy is determined through decision-making suggestions and adjustment plans; if risk warnings are obtained through the prediction results, the project monitoring strategy is adjusted in real time to ensure that the project always runs within the expected track under changes in the external environment and adjustments in internal resources; The method of introducing a correction plan according to the accuracy of the forecast project strategy and dynamically adjusting the forecast project monitoring strategy to obtain the project monitoring strategy includes: If the prediction accuracy of the project strategy is higher than the set threshold, obtain the parameter adjustment plan of the prediction model: By analyzing the coverage of the forecast project strategy, we can obtain the areas with low coverage and determine the parameter weights of these areas in the forecast model; if the adjustment range of the revised plan exceeds the predetermined standard, determine the impact of the adjustment range on the forecast accuracy and obtain the impact assessment results; according to the adjustment frequency of the revised plan, obtain the correlation data between the adjustment frequency and the forecast accuracy and determine the impact of the frequency on the accuracy; use the dynamic update cycle data of the project monitoring strategy to determine the effect of the update cycle on the monitoring accuracy; by analyzing the monitoring indicators of the project monitoring strategy, obtain the key monitoring indicators related to the forecast project and determine the relationship between the indicators and the forecast accuracy; according to the early warning threshold of the project monitoring strategy, obtain the impact of the threshold setting on the effectiveness of the project early warning; through the feedback mechanism of the project monitoring strategy, determine whether the feedback mechanism is perfect and its effect on the forecast accuracy and dynamic adjustment of the monitoring strategy; according to the accuracy of the forecast project strategy and the results of the dynamic adjustment of the revised plan, obtain the final adjustment plan of the project monitoring strategy; Comprehensively determine the implementation plan of the project monitoring strategy to form a complete predictive monitoring process.

Images