Dynamic optimization method for big data in project management

Abstract

The present invention belongs to the field of project management technologies, and discloses a dynamic optimization method for big data in project management, including: obtaining project management information; constructing a project knowledge graph, and performing graph traversal on the project knowledge graph based on a real-time task set of a to-be-managed project, to obtain a corresponding dynamic subgraph; performing image inference based on the obtained dynamic subgraph, to obtain corresponding project expectation information, and determining, based on the project expectation information, whether there is a project risk in the to-be-managed project; and if there is a risk, generating corresponding risk early-warning information; and constructing a set of backup response policies based on the risk early-warning information, and performing policy selection optimization based on dependences between data nodes in the dynamic subgraph, to obtain a corresponding optimal response policy.

Description

A Dynamic Optimization Method for Project Management Based on Big Data Technical Field

[0001] This invention relates to the field of project management technology, and more specifically, to a dynamic optimization method for project management big data. Background Technology

[0002] In today's complex and ever-changing business environment and the rapid development of information technology, project management faces unprecedented challenges. The limitations of traditional project management methods are becoming increasingly apparent, urgently requiring innovative solutions, specifically in the following aspects:

[0003] 1. Traditional data collection methods are often limited to a single or a few data sources, resulting in incomplete and untimely data acquisition, which makes it difficult to reflect the true status of the project. Moreover, in the data processing process, there is a lack of effective integration and standardization methods, resulting in chaotic data formats and inconsistent quality, making it difficult for managers to extract valuable information from the data and failing to provide strong support for project decision-making, thus increasing the risk of decision-making errors.

[0004] 2. The lack of tools that can intuitively and comprehensively display the overall project structure and task relationships makes it difficult for managers to grasp the overall picture of the project and to have a sufficient understanding of the project's progress, potential problems, and the synergistic effects between various tasks. This not only affects the efficiency of communication and collaboration among team members, but also easily leads to problems such as poor task coordination and unreasonable resource allocation, which in turn delays the project schedule and increases project costs.

[0005] 3. Traditional risk assessment methods mainly rely on subjective experience and simple analytical models, which cannot accurately predict potential risks. They often only respond passively after risks occur, missing the best time for prevention and control. In addition, when facing risks, there is a lack of systematic response strategy generation and optimization mechanisms. The response measures taken are often not targeted and have poor effects, making it difficult to effectively reduce risk losses and ensure the smooth progress of the project.

[0006] 4. Traditional project management methods lack the ability to optimize and dynamically adjust strategies in real time, and cannot respond promptly to changes in the actual situation and environment of the project. This makes the project's response strategies lagging behind when facing unexpected situations or new problems, unable to adapt to the dynamic development needs of the project, and affecting the achievement of project goals.

[0007] In view of this, the present invention proposes a dynamic optimization method for project management big data to solve the above problems. Technical issues

[0008] Traditional project management methods rely on decentralized tools, making it difficult to fully consider the complex interactions between projects; this leads to low project management efficiency and insufficient risk control. Technical solutions

[0009] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution:

[0010] A project management big data dynamic optimization method includes:

[0011] Step 1: Obtain project management information and perform data preprocessing on the corresponding project management information based on the pre-built ontology database to obtain the corresponding project optimization information;

[0012] Step 2: Construct a project knowledge graph and perform graph traversal based on the real-time task set of the project to be managed to obtain the corresponding dynamic subgraphs;

[0013] Step 3: Perform image reasoning based on the obtained dynamic subgraph to obtain the corresponding expected project information, and determine whether there are project risks in the project to be managed. If there are risks, generate corresponding risk warning information.

[0014] Step 4: Construct a set of backup response strategies based on risk warning information, and optimize strategy selection based on the dependency relationships between data nodes in the dynamic subgraph to obtain the corresponding optimal response strategy. Monitor the execution process of the corresponding optimal response strategy in real time and provide feedback based on the monitoring results.

[0015] Furthermore, the process of obtaining project optimization information includes:

[0016] Data entities are extracted from the obtained project management information to obtain the corresponding source data entities, which include entity names and word vectors.

[0017] Construct an ontology database, compare the obtained source data entities with the project management entity ontology in the corresponding ontology database, and obtain the corresponding entity similarity measure.

[0018] The obtained entity similarity measure is compared with a pre-set similarity threshold, and entity combinations are constructed based on the comparison results;

[0019] Based on the obtained entity combinations, project management information is mapped to a semantic and data format consistent with the ontology database; and entity attribute information and relationship information are extracted simultaneously.

[0020] Once the mapping is complete, construct the corresponding project Gantt chart based on the mapped project management information;

[0021] Based on the project Gantt chart, obtain the dependencies of project tasks at the corresponding time nodes, and determine the number of dependency edges; construct a corresponding dynamic window based on the number of dependency edges;

[0022] Obtain the local project management information corresponding to the dynamic window and input it into the pre-selected variational autoencoder to obtain the corresponding reconstruction project information; and perform anomaly assessment on the task data of the project task at the corresponding time node based on it to obtain the corresponding data score; evaluate whether there is anomaly in the task data at the corresponding time node based on the data score, and if so, correct the data.

[0023] Repeat the above process until all abnormal data points are corrected and the corresponding project optimization data is obtained.

[0024] Furthermore, the process of constructing the project knowledge graph includes:

[0025] Based on the project optimization data corresponding to the project set, the project set is divided into several sub-projects according to functional modules, business areas or organizational structures; the data entities corresponding to each sub-project are obtained and recorded as project entities; the communication records corresponding to each sub-project are obtained based on the project optimization information.

[0026] Construct data nodes, map the project entities to the corresponding data nodes, and store the entity attribute information of the corresponding project entities in the corresponding data nodes;

[0027] Based on the task relationships and communication records within the obtained project Gantt chart, connect each data node;

[0028] Acquire project text data and extract its text features to obtain the corresponding text feature vectors. Simultaneously, convert the time information in the corresponding project Gantt chart into time vectors and store the corresponding text feature vectors and time vectors into the corresponding data nodes to obtain the corresponding project sub-maps.

[0029] By combining the project subgraphs corresponding to different sub-projects, a corresponding project knowledge graph can be obtained.

[0030] Furthermore, the process of obtaining the dynamic subgraph includes:

[0031] Obtain real-time project optimization information, and based on this, obtain the current project stage and set of real-time tasks being executed for the project to be managed.

[0032] Based on the project stage and real-time task set, the corresponding data nodes are matched in the corresponding project knowledge graph and used as the starting point of the path. Based on this, all other data nodes and edges that are directly or indirectly related to the project tasks are recursively extracted to obtain the corresponding connected subgraph. The time attributes corresponding to each project task in the real-time task set are obtained, and the extracted data nodes are arranged in time sequence based on these attributes, transforming the static project subgraph into a dynamic subgraph with time sequence characteristics.

[0033] Furthermore, the process of image reasoning based on the obtained dynamic subgraph includes:

[0034] The obtained dynamic subgraph is input into a pre-built graph neural network inference model to obtain the corresponding project expectation information, which includes task indicator prediction data and potential node risk information.

[0035] Furthermore, based on the project knowledge graph, the expected information of the corresponding projects is structured and organized to obtain the corresponding project prediction information. The project prediction information includes the predicted values ​​of each key indicator, confidence intervals, risk point descriptions, risk levels, impact ranges, and corresponding time information.

[0036] Furthermore, the backbone network of the graph neural network inference model is a hierarchical neural network, which is composed of graph neural networks and temporal neural networks. The basic framework of the hierarchical neural network includes an input layer, a convolutional layer, a prediction layer, a detection layer, and an output layer.

[0037] The input layer is used to receive the input dynamic sub-graph and perform image preprocessing on it;

[0038] The convolutional layer is used to perform graph convolution operations on the dynamic subgraphs after image preprocessing to obtain the corresponding node feature matrix;

[0039] The prediction layer is used to read the historical task parameter sequence of each data node in the dynamic subgraph, perform data prediction based on the historical task parameter sequence and node feature matrix, obtain the prediction state of the data node at several future time nodes, perform state parsing on the corresponding prediction state to obtain the corresponding feature vector, and obtain the prediction values ​​of each key indicator of the corresponding project task of the corresponding data node based on the feature vector; at the same time, the Monte Carlo dropout method is used to predict the confidence interval corresponding to the key indicator.

[0040] The detection layer performs anomaly scoring on the corresponding data nodes and their corresponding edges based on the feature vectors in the prediction layer, and determines whether the corresponding data node is an abnormal node. If it is an abnormal node, the risk level is classified based on the anomaly score to obtain the corresponding node risk level.

[0041] The output layer is used to receive the predicted values ​​of key indicators and the anomaly scores and risk levels of each data node, and map them to a pre-set data space to obtain the corresponding project prediction information.

[0042] Furthermore, the process of determining whether the projects to be managed have any abnormal risks based on the project prediction information includes:

[0043] Based on the obtained project forecast information, the deviation of the key indicator parameters corresponding to each project task in the corresponding project forecast information and the corresponding confidence interval is calculated to obtain the parameter deviation of each project task. Based on this, a comprehensive anomaly assessment is conducted on the projects to be managed to obtain the corresponding project risk score.

[0044] Set risk thresholds, compare the obtained project risk scores with the corresponding risk thresholds, and determine whether the project risk is under control. If it is under control, the project will be continuously monitored; if it is not under control, the corresponding risk warning information will be generated.

[0045] Furthermore, the process of assembling a set of backup response strategies includes:

[0046] When a risk warning is identified, the data nodes with risks within the dynamic subgraph are obtained and marked as risk nodes.

[0047] Based on the project prediction information, obtain the risk types corresponding to the risk nodes and the trigger probability of each risk type;

[0048] Based on the project prediction information, the risk feature vector of the project task corresponding to the corresponding risk node is obtained. Then, in combination with the task parameters and risk feature vector of the project task, several similar historical project task cases are obtained from the historical project database. The risk response measures of the corresponding historical project task cases are used as reference response strategies.

[0049] Obtain the task differences between the current project tasks and historical project tasks, and adjust the corresponding reference response strategies based on them to obtain a set of corresponding backup response strategies.

[0050] Furthermore, the process of obtaining the optimal response strategy includes:

[0051] Based on the dynamic subgraph, the dependency relationship between the corresponding risk node and other data nodes is obtained, and the influence strength of other data nodes on the corresponding risk node under the corresponding risk type is determined.

[0052] Obtain the impact strength of other data nodes on the corresponding risk nodes, and construct the corresponding project selection optimization model by combining the obtained risk types and the set of alternative response strategies; and define the objective function of the corresponding project selection optimization model based on the pre-set constraints.

[0053] Obtain the task execution scenario and construct several expected task execution scenarios based on the uncertainties in the task execution scenario; obtain the strategy stability index corresponding to the backup response strategy based on the expected task execution scenario;

[0054] Obtain the management preferences of relevant staff for the required projects and the strategy stability index under various expected task execution scenarios, and solve the corresponding objective function using the particle swarm optimization algorithm to obtain the corresponding optimal response strategy.

[0055] Furthermore, the process of real-time monitoring of the execution of the corresponding optimal response strategy and providing feedback based on the monitoring results includes:

[0056] A real-time monitoring mechanism is established to dynamically track the process of implementing the optimal response strategy. Key indicators of the corresponding project tasks are collected in real time during the execution of the optimal response strategy and compared with the pre-set expected goals. If the expected goals are not met, the project management information of the corresponding managed project is collected in real time and preprocessed. The dynamic subgraph is then updated based on the data, and the optimal response strategy being executed is optimized based on the updated dynamic subgraph. At the same time, the strategy execution results are fed back to the project knowledge graph to update the corresponding data nodes. Beneficial effects

[0057] The technical effects and advantages of the project management big data dynamic optimization method of the present invention are as follows:

[0058] 1. By obtaining dynamic subgraphs through real-time tasks, the project's real-time status and task relationships are clearly displayed; this enables managers to grasp the project from a global perspective, facilitates smoother communication and collaboration among team members, effectively avoids task connection problems, achieves reasonable resource allocation, improves project execution efficiency, and ensures the orderly progress of the project.

[0059] 2. By performing image reasoning on dynamic subgraphs, it accurately predicts task indicators and potential risks, and organizes detailed project prediction information in a structured manner; by calculating project risk scores and comparing them with thresholds, it promptly and accurately determines the risk status and issues early warnings; enabling managers to plan risk response strategies in advance, transforming passive response into proactive prevention and control, significantly reducing the probability of risk occurrence, minimizing losses caused by risks, and ensuring the achievement of project goals. Attached Figure Description

[0060] Figure 1 is a schematic diagram of a project management big data dynamic optimization method according to the present invention;

[0061] Figure 2 is a schematic diagram of a project management big data dynamic optimization system according to the present invention. Embodiments of the present invention

[0062] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1

[0063] Please refer to Figure 1. The project management big data dynamic optimization method described in this embodiment includes:

[0064] Step 1: Obtain project management information and perform data preprocessing on the corresponding project management information based on the pre-built ontology database to obtain the corresponding project optimization information;

[0065] Step 2: Construct a project knowledge graph and perform graph traversal based on the real-time task set of the project to be managed to obtain the corresponding dynamic subgraphs;

[0066] Step 3: Perform image reasoning based on the obtained dynamic subgraph to obtain the corresponding expected project information, and determine whether there are project risks in the project to be managed. If there are risks, generate corresponding risk warning information.

[0067] Step 4: Construct a set of backup response strategies based on risk warning information, and optimize strategy selection based on the dependency relationships between data nodes in the dynamic subgraph to obtain the corresponding optimal response strategy. Monitor the execution process of the corresponding optimal response strategy in real time and provide feedback based on the monitoring results.

[0068] It should be further explained that, in the specific implementation process, the process of obtaining project management information includes:

[0069] A data acquisition unit is set up to collect multi-source data on the target management project to obtain corresponding project management information. The project management information includes internal and external information. The internal information includes project parameters such as project risks, response strategies, costs, schedules, and quality. The external data includes information such as market environment, industry trends, and policies and regulations. The data acquisition unit is a distributed data acquisition terminal that connects to multiple data sources through an API interface and acquires information based on the API interface.

[0070] It should be further explained that, in the specific implementation process, the process of obtaining project optimization information includes:

[0071] The obtained project management information is used to extract data entities such as tasks, risks, resources, and personnel to obtain corresponding source data entities, which include entity names and word vectors.

[0072] An ontology database is constructed, which includes a project domain ontology, a project management entity ontology, and a project risk ontology. This database is used to define the conceptual framework and relationship rules between entities in the project management domain. The construction process of the ontology database is existing technology and will not be elaborated upon in this invention.

[0073] The obtained source data entities are compared with the project management entity ontology in the corresponding ontology database to obtain the corresponding entity similarity measure. In the formula, e1 and e2 represent the source data entity and the project entity ontology in the ontology database, respectively. and Both represent entity names; Used to calculate the edit distance between e1 and e2, whereby the edit distance refers to the minimum number of editing operations required to transform one string into another, including insertion, deletion, and replacement of characters; and These represent the word vectors of the entities; Matching scores based on domain rules, whereby domain rules are pre-defined by professionals in the field based on industry standards; , and Represents the weight parameters, and ;

[0074] The obtained entity similarity measure is compared with a pre-set similarity threshold. If the entity similarity measure is not less than the similarity threshold, the corresponding source data entity and the corresponding project entity ontology are classified into the same category to obtain the corresponding entity combination.

[0075] Furthermore, based on the obtained entity combination, the project management information collected from multiple sources is mapped to a semantic and data format consistent with the ontology database; and entity attribute information and relationship information are extracted simultaneously; for example, if the source data entity and the project management entity ontology in the entity combination are "project delay" and "schedule delay" respectively, then they are uniformly mapped to "schedule delay".

[0076] Once the mapping is complete, a corresponding project Gantt chart is constructed based on the project management information after the mapping is completed. The project Gantt chart includes the time information of each project task in the corresponding project set and the dependencies between tasks.

[0077] Taking a specific project task as an example, the dependencies of the project task at the corresponding time node are obtained based on the project Gantt chart, and the number of dependency edges is determined accordingly. A corresponding dynamic window is then constructed based on the number of dependency edges. The formula for constructing the dynamic window is as follows: In the formula, , and This represents trainable parameters obtained through regression fitting of historical project data; Indicates the maximum window limit. and These represent the number of dependency edges for a given task and the total number of tasks, respectively. and These represent the resource volatility of the corresponding task and the number of historical anomalies for the corresponding task, respectively.

[0078] Then, the local project management information corresponding to the dynamic window is obtained and input into the pre-selected variational autoencoder to obtain the corresponding refactoring project information; and based on this, anomaly evaluation is performed on the task data of the project tasks at the corresponding time node to obtain the corresponding data score. In the formula, This represents the weight parameter, and it is the default value. x and x^ represent the local project management data and refactored project data in the dynamic window, respectively; This represents the KL divergence operation, used to measure the difference between two probability distributions. Represents the standard normal distribution. The probability distribution of latent variable z under the original data; where the working principle of the variational autoencoder is to map the input local item optimization data into the latent space through the built-in encoder to obtain the corresponding latent variables and probability distribution; then, the latent variables are reconstructed through the decoder to obtain the corresponding reconstructed item information;

[0079] An anomaly threshold is set. If the anomaly score is less than the threshold, no further action is taken. If the anomaly score is not less than the threshold, it indicates that the task data of the project at the corresponding time point is abnormal. In this case, the abnormal data point is optimized based on pre-defined business constraints, and the original abnormal data point is replaced by the optimized one. The optimized abnormal data point must satisfy both the business constraints and the anomaly score being less than the threshold. The business constraints include resource constraints, business cost constraints, and task time constraints. The optimization process of the abnormal data point refers to searching for the best fit value of the corresponding abnormal data point within the corresponding latent variables based on the business constraints.

[0080] Repeat the above process until all abnormal data points are corrected and the corresponding project optimization data is obtained.

[0081] It should be further explained that, in the specific implementation process, the construction process of the project knowledge graph includes:

[0082] Based on the project optimization data corresponding to the project set, the project set is divided into several sub-projects according to functional modules, business domains, or organizational structures; for example, a software development project can be divided into sub-projects such as requirements analysis, design, coding, testing, and deployment; each sub-project can correspond to multiple project tasks;

[0083] Obtain the data entities corresponding to each sub-project and record them as project entities;

[0084] Based on project optimization information, communication records corresponding to each sub-project are obtained, and these communication records reflect the collaborative relationships and task allocation among personnel.

[0085] Construct data nodes and map the project entities to the corresponding data nodes. At the same time, obtain the entity attribute information corresponding to the project entity based on the ontology database and store it in the corresponding data nodes.

[0086] Furthermore, based on the task relationships and communication records within the obtained project Gantt chart, each data node is linked together.

[0087] Acquire project textual data and extract its textual features to obtain corresponding textual feature vectors. Simultaneously, convert the time information within the corresponding project Gantt chart into time vectors. Store the corresponding textual feature vectors and time vectors into the corresponding data nodes to obtain the corresponding project sub-maps, thereby enhancing the information representation capability of the nodes.

[0088] By combining the project subgraphs corresponding to different sub-projects, a corresponding project knowledge graph is obtained; the project knowledge graph consists of several data nodes and edges connecting the data nodes.

[0089] It should be further explained that, in the specific implementation process, the acquisition process of the dynamic subgraph includes:

[0090] Obtain real-time project optimization information, and based on this, obtain the current project stage and set of real-time tasks being executed for the project to be managed.

[0091] Furthermore, based on the project stage and real-time task set, corresponding data nodes are matched within the corresponding project knowledge graph and used as the starting point of the path. Based on this, all other data nodes and edges directly or indirectly related to the project tasks are recursively extracted to obtain the corresponding connected subgraph. The time attributes corresponding to each project task in the real-time task set are obtained, and the extracted data nodes are arranged in a time sequence based on these attributes, transforming the static project subgraph into a dynamic subgraph with time sequence characteristics. The time sequence arrangement process uses a topological sorting algorithm.

[0092] It should be further explained that, in the specific implementation process, the image reasoning process based on the obtained dynamic subgraph includes:

[0093] The obtained dynamic subgraph is input into a pre-built graph neural network inference model to obtain the corresponding project expectation information, which includes task indicator prediction data and potential node risk information.

[0094] Furthermore, based on the project knowledge graph, the expected information of the corresponding projects is structured and organized to obtain the corresponding project prediction information. The project prediction information includes the predicted values ​​of each key indicator, confidence intervals, risk point descriptions, risk levels, impact ranges, and corresponding time information.

[0095] The construction process of the graph neural network inference model includes:

[0096] The backbone network of the graph neural network inference model is a hierarchical neural network, which is composed of graph neural networks and temporal neural networks. The basic framework of the hierarchical neural network includes an input layer, a convolutional layer, a prediction layer, a detection layer, and an output layer.

[0097] The input layer is used to receive the input dynamic subgraph and perform image preprocessing on it to meet the input requirements of the hierarchical neural network.

[0098] The convolutional layer is used to perform graph convolution operations on the dynamic subgraphs after image preprocessing to obtain the corresponding node feature matrix. The mathematical formula for the graph convolution operation is as follows: In the formula, and They represent the first Convolutional layers and the first The node feature matrix corresponding to each convolutional layer is used to represent the feature information of the corresponding data node at the corresponding level; D represents a diagonal matrix, where the matrix elements on the main diagonal of the diagonal matrix are the number of edges connected to the corresponding data node; This represents a self-loop matrix, which means that in addition to the original node connection information, the nodes' own connections (i.e., self-loops) are added on the main diagonal. This helps to preserve the information of each node and prevents oversmoothing to some extent. and softmax represent the weight matrix and activation function, respectively;

[0099] The prediction layer is used to read the historical task parameter sequence of each data node in the dynamic subgraph, perform data prediction based on the historical task parameter sequence and node feature matrix, obtain the prediction state of the data node at several future time nodes, perform state parsing on the corresponding prediction state to obtain the corresponding feature vector, and obtain the prediction values ​​of each key indicator of the corresponding project task of the corresponding data node based on the feature vector; at the same time, the Monte Carlo dropout method is used to predict the confidence interval corresponding to the key indicator.

[0100] The detection layer performs anomaly scoring on the corresponding data nodes and their corresponding edges based on the feature vectors within the prediction layer, and determines whether the corresponding data node is an anomalous node. If it is an anomalous node, a risk level classification is performed based on the anomaly score to obtain the corresponding node risk level. The formula for anomaly scoring of data nodes is as follows: In the formula, This represents the anomaly score of the i-th data node within the dynamic subgraph; and These represent the actual feature vector and the predicted feature vector of the corresponding data node, respectively;

[0101] The formula for anomaly scoring of the nodes and edges of the corresponding data nodes is as follows: In the formula, This represents the anomaly score of the edge corresponding to the j-th node and the i-th data node. This represents a vector concatenation operation; This represents the actual feature vector corresponding to the data node connected by the j-th edge; This represents a pre-selected feature mapping function;

[0102] The output layer is used to receive the predicted values ​​of key indicators and the anomaly scores and risk levels of each data node, and map them to a pre-set data space to obtain the corresponding project prediction information.

[0103] Define the loss function of the graph neural network inference model, and construct a training dataset based on historical project optimization information. Iteratively train the corresponding network inference model based on the training dataset until the corresponding loss function gradually converges; that is, the model training is completed.

[0104] It should be further explained that, in the specific implementation process, the process of determining whether there are abnormal risks in the projects to be managed based on the project prediction information includes:

[0105] Based on the obtained project forecast information, the deviations of key indicator parameters corresponding to each project task within the forecast information and their corresponding confidence intervals are calculated to obtain the parameter deviation degree of each project task. Based on this, a comprehensive anomaly assessment is conducted on the projects to be managed to obtain the corresponding project risk score. In the formula, and Indicates risk weight; , These represent the indicator weights and parameter deviations of the key indicator corresponding to the i-th data node, respectively. This represents the index weight of the j-th edge connected to the i-th data node. This indicates the risk level corresponding to the data node; I0 and i represent the total number and index of data nodes within the dynamic subgraph, respectively.

[0106] Set risk thresholds, compare the obtained project risk scores with the corresponding risk thresholds, and determine whether the project risk is under control. If it is under control, the project will be continuously monitored; if it is not under control, the corresponding risk warning information will be generated.

[0107] It should be further explained that, in the specific implementation process, the process of constructing a set of backup response strategies includes:

[0108] When a risk warning is identified, the data nodes with risks within the dynamic subgraph are obtained and marked as risk nodes.

[0109] Based on the risk point description, risk level, and impact range within the project prediction information, obtain the risk type corresponding to the risk node and the trigger probability of each risk type.

[0110] Furthermore, based on the project prediction information, risk feature vectors of project tasks corresponding to the corresponding risk nodes are obtained. Then, in combination with the task parameters and risk feature vectors of the project tasks, several similar historical project task cases are obtained from the historical project database. The risk response measures of the corresponding historical project task cases are used as reference response strategies. The historical project database stores project management information, risk feature vectors, and risk response measures.

[0111] Furthermore, the task differences between the current project task and the historical project task are obtained, and the corresponding reference response strategies are adjusted based on these differences to obtain a set of alternative response strategies. The adjustment process of the reference response strategies is as follows: the adjustment process of the corresponding reference response strategies is modeled as a Markov decision process, the corresponding task differences are the state space, and the strategy adjustment decision is the action space; the corresponding reference response strategies are optimized based on the deep deterministic policy gradient algorithm to obtain a set of alternative response strategies.

[0112] It should be further explained that, in the specific implementation process, the process of obtaining the optimal response strategy includes:

[0113] Taking a specific risk node as an example, the dependencies between the risk node and other data nodes are obtained based on the dynamic subgraph, and the influence strength of other data nodes on the corresponding risk node under its corresponding risk type is determined; wherein, the formula for obtaining the influence strength is:

[0114] In the formula, This indicates the intensity of the impact of risk type 'a' on other dependent projects during the corresponding project cycle for the risk type 'a' of the project task. This indicates that the corresponding project cycle is divided into T project sub-cycles, where t0 represents the time index within the project sub-cycle. This represents the other project tasks being executed at time t0, excluding the project tasks corresponding to the risk nodes; k represents the index of the number of other project tasks being executed. This indicates the strength of the dependency between the risk node's corresponding project task and other project tasks; it is determined by the direct and indirect dependencies between projects. and Let represent the success probability and failure probability of other executed project task k, respectively; and These represent whether the failure or success of project task k has any impact. If there is an impact, the value is 1; if there is no impact, the value is 0. , , and These settings are all determined by the relevant staff based on the actual circumstances of the projects to be managed.

[0115] Furthermore, the impact strength of other data nodes on the corresponding risk nodes is obtained, and a corresponding project selection optimization model is constructed by combining the obtained risk types and the set of alternative response strategies; and the objective function of the corresponding project selection optimization model is defined based on the pre-set constraints; the constraints include dependency relationships, resource constraints, time constraints and cost constraints.

[0116] Furthermore, the task execution scenarios corresponding to the project tasks at the respective risk nodes are obtained, along with the uncertainties under these scenarios, including task duration and resource availability. Based on these uncertainties, several possible task execution scenarios are constructed, and probability weights are assigned to each scenario. Then, the corresponding backup response strategies are simulated under the respective task execution scenarios to obtain strategy stability indices. These indices measure the volatility of the backup response strategies under different task execution scenarios; higher volatility indicates greater sensitivity of the backup response strategies to the corresponding task execution scenarios. The simulation process utilizes digital twin technology simulation software.

[0117] Obtain the management preferences of relevant staff for the projects to be managed and the strategy stability index under various task execution scenarios, and solve the corresponding objective function by combining the particle swarm optimization algorithm to obtain the corresponding optimal response strategy; and apply it to the projects to be managed.

[0118] A real-time monitoring mechanism is established to dynamically track the process of implementing the optimal response strategy. Key indicators of relevant project tasks during the execution of the optimal response strategy are collected in real time and compared with pre-set expected goals. If the expected goals are not met, project management information of the relevant managed project is collected in real time, preprocessed, and fed back to the dynamic subgraph within the step for updating. Based on this, the optimal response strategy being executed is optimized. At the same time, the strategy execution results are fed back to the project knowledge graph to update the attribute values ​​and relationship strength of relevant nodes. Effective experiences are also added to the historical project experience database, forming a closed-loop optimization mechanism for continuously optimizing project management methods and improving the management efficiency of future projects.

[0119] This invention acquires project management information through a distributed data acquisition unit, preprocesses it using an ontology database to obtain project optimization information, then constructs a project knowledge graph and obtains dynamic subgraphs based on real-time tasks. Next, it uses a graph neural network inference model to perform image reasoning on the dynamic subgraphs to determine if the project has risks and generate early warnings. Based on these early warnings, it constructs a set of backup response strategies, optimizes them to obtain the optimal response strategy, and monitors it in real time. This allows for early risk prediction and prevention; optimized strategy selection improves the effectiveness of response plans; and real-time monitoring and feedback form a closed-loop optimization, continuously improving project management methods and efficiency. Example 2

[0120] Please refer to Figure 2. For details not described in this embodiment, please refer to the description in Embodiment 1. A project management big data dynamic optimization system is provided, including:

[0121] The information acquisition module acquires project management information and performs data preprocessing on the corresponding project management information based on a pre-built ontology database to obtain corresponding project optimization information.

[0122] The graph construction module is used to build a project knowledge graph and perform graph traversal on the project knowledge graph based on the real-time task set of the project to be managed, to obtain the corresponding dynamic subgraphs.

[0123] The risk assessment module is used to perform image reasoning based on the obtained dynamic subgraphs to obtain the corresponding expected project information, and to determine whether there are project risks in the project to be managed. If there are risks, corresponding risk warning information is generated.

[0124] The strategy construction module builds a set of backup response strategies based on risk warning information, and optimizes strategy selection based on the dependencies between data nodes within the dynamic subgraph to obtain the corresponding optimal response strategy.

[0125] The monitoring and feedback module is used to monitor the execution process of the corresponding optimal response strategy in real time and provide information feedback based on the monitoring results.

[0126] The modules are connected via wired and / or wireless means to enable data transmission between them. Example 3

[0127] This embodiment discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the above-described project management big data dynamic optimization method.

[0128] Since the electronic device described in this embodiment is used to implement the project management big data dynamic optimization method described in this application embodiment, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the project management big data dynamic optimization method described in this application embodiment. Therefore, how the electronic device implements the method in this application embodiment will not be described in detail here. Any electronic device used by those skilled in the art to implement the project management big data dynamic optimization method described in this application embodiment falls within the scope of protection of this application.

[0129] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0130] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A dynamic optimization method for project management based on big data, characterized in that, include: Step 1: Obtain project management information and perform data preprocessing on the corresponding project management information based on the pre-built ontology database to obtain the corresponding project optimization information; Step 2: Construct a project knowledge graph and perform graph traversal based on the real-time task set of the project to be managed to obtain the corresponding dynamic subgraphs; Step 3: Perform image reasoning based on the obtained dynamic subgraph to obtain the corresponding expected project information, and determine whether there are project risks in the project to be managed. If there are risks, generate corresponding risk warning information. Step 4: Construct a set of backup response strategies based on risk warning information, and optimize strategy selection based on the dependency relationships between data nodes in the dynamic subgraph to obtain the corresponding optimal response strategy. Monitor the execution process of the corresponding optimal response strategy in real time and provide feedback based on the monitoring results.

2. The project management big data optimization method according to claim 1, characterized in that, The process of obtaining project optimization information includes: Data entities are extracted from the obtained project management information to obtain the corresponding source data entities, which include entity names and word vectors. Construct an ontology database, compare the obtained source data entities with the project management entity ontology in the corresponding ontology database, and obtain the corresponding entity similarity measure. The obtained entity similarity measure is compared with a pre-set similarity threshold, and entity combinations are constructed based on the comparison results; Based on the obtained entity combinations, project management information is mapped to a semantic and data format consistent with the ontology database; and entity attribute information and relationship information are extracted simultaneously. Once the mapping is complete, construct the corresponding project Gantt chart based on the mapped project management information; Based on the project Gantt chart, obtain the dependencies of project tasks at the corresponding time nodes, and determine the number of dependency edges; construct a corresponding dynamic window based on the number of dependency edges; Obtain the local project management information corresponding to the dynamic window and input it into the pre-selected variational autoencoder to obtain the corresponding reconstruction project information; and perform anomaly assessment on the task data of the project task at the corresponding time node based on it to obtain the corresponding data score; evaluate whether there is anomaly in the task data at the corresponding time node based on the data score, and if so, correct the data. Repeat the above process until all abnormal data points are corrected and the corresponding project optimization data is obtained.

3. The project management big data optimization method according to claim 2, characterized in that, The process of constructing the project knowledge graph includes: Based on the project optimization data corresponding to the project set, the project set is divided into several sub-projects according to functional modules, business areas or organizational structures; the data entities corresponding to each sub-project are obtained and recorded as project entities; the communication records corresponding to each sub-project are obtained based on the project optimization information. Construct data nodes, map the project entities to the corresponding data nodes, and store the entity attribute information of the corresponding project entities in the corresponding data nodes; Based on the task relationships and communication records within the obtained project Gantt chart, connect each data node; Acquire project text data and extract its text features to obtain the corresponding text feature vectors. Simultaneously, convert the time information in the corresponding project Gantt chart into time vectors and store the corresponding text feature vectors and time vectors into the corresponding data nodes to obtain the corresponding project sub-maps. By combining the project subgraphs corresponding to different sub-projects, a corresponding project knowledge graph can be obtained.

4. The project management big data optimization method according to claim 3, characterized in that, The process of obtaining a dynamic subgraph includes: Obtain real-time project optimization information, and based on this, obtain the current project stage and set of real-time tasks being executed for the project to be managed. Based on the project stage and real-time task set, the corresponding data nodes are matched in the corresponding project knowledge graph and used as the starting point of the path. Based on this, all other data nodes and edges that are directly or indirectly related to the project tasks are recursively extracted to obtain the corresponding connected subgraph. The time attributes corresponding to each project task in the real-time task set are obtained, and the extracted data nodes are arranged in time sequence based on these attributes, transforming the static project subgraph into a dynamic subgraph with time sequence characteristics.

5. The project management big data optimization method according to claim 4, characterized in that, The process of image reasoning based on the obtained dynamic subgraph includes: The obtained dynamic subgraph is input into a pre-built graph neural network inference model to obtain the corresponding project expectation information, which includes task indicator prediction data and potential node risk information. Furthermore, based on the project knowledge graph, the expected information of the corresponding projects is structured and organized to obtain the corresponding project prediction information. The project prediction information includes the predicted values ​​of each key indicator, confidence intervals, risk point descriptions, risk levels, impact ranges, and corresponding time information.

6. The project management big data optimization method according to claim 5, characterized in that, The backbone network of the graph neural network inference model is a hierarchical neural network, which is composed of graph neural networks and temporal neural networks. The basic framework of the hierarchical neural network includes an input layer, a convolutional layer, a prediction layer, a detection layer, and an output layer. The input layer is used to receive the input dynamic sub-graph and perform image preprocessing on it; The convolutional layer is used to perform graph convolution operations on the dynamic subgraphs after image preprocessing to obtain the corresponding node feature matrix; The prediction layer is used to read the historical task parameter sequence of each data node in the dynamic subgraph, perform data prediction based on the historical task parameter sequence and node feature matrix, obtain the prediction state of the data node at several future time nodes, perform state parsing on the corresponding prediction state to obtain the corresponding feature vector, and obtain the prediction values ​​of each key indicator of the corresponding project task of the corresponding data node based on the feature vector; at the same time, the Monte Carlo dropout method is used to predict the confidence interval corresponding to the key indicator. The detection layer performs anomaly scoring on the corresponding data nodes and their corresponding edges based on the feature vectors in the prediction layer, and determines whether the corresponding data node is an abnormal node. If it is an abnormal node, the risk level is classified based on the anomaly score to obtain the corresponding node risk level. The output layer is used to receive the predicted values ​​of key indicators and the anomaly scores and risk levels of each data node, and map them to a pre-set data space to obtain the corresponding project prediction information.

7. The project management big data optimization method according to claim 6, characterized in that, The process of determining whether there are abnormal risks in the projects to be managed based on the project prediction information includes: Based on the obtained project forecast information, the deviation of the key indicator parameters corresponding to each project task in the corresponding project forecast information and the corresponding confidence interval is calculated to obtain the parameter deviation of each project task. Based on this, a comprehensive anomaly assessment is conducted on the projects to be managed to obtain the corresponding project risk score. Set risk thresholds, compare the obtained project risk scores with the corresponding risk thresholds, and determine whether the project risk is under control. If it is under control, the project will be continuously monitored; if it is not under control, the corresponding risk warning information will be generated.

8. The project management big data optimization method according to claim 7, characterized in that, The process of compiling a set of backup response strategies includes: When a risk warning is identified, the data nodes with risks within the dynamic subgraph are obtained and marked as risk nodes. Based on the project prediction information, obtain the risk types corresponding to the risk nodes and the trigger probability of each risk type; Based on the project prediction information, the risk feature vector of the project task corresponding to the corresponding risk node is obtained. Then, in combination with the task parameters and risk feature vector of the project task, several similar historical project task cases are obtained from the historical project database. The risk response measures of the corresponding historical project task cases are used as reference response strategies. Obtain the task differences between the current project tasks and historical project tasks, and adjust the corresponding reference response strategies based on them to obtain a set of corresponding backup response strategies.

9. The project management big data optimization method according to claim 8, characterized in that, The process of obtaining the optimal response strategy includes: Based on the dynamic subgraph, the dependency relationship between the corresponding risk node and other data nodes is obtained, and the influence strength of other data nodes on the corresponding risk node under the corresponding risk type is determined. Obtain the impact strength of other data nodes on the corresponding risk nodes, and construct the corresponding project selection optimization model by combining the obtained risk types and the set of alternative response strategies; and define the objective function of the corresponding project selection optimization model based on the pre-set constraints. Obtain the task execution scenario and construct several expected task execution scenarios based on the uncertainties in the task execution scenario; obtain the strategy stability index corresponding to the backup response strategy based on the expected task execution scenario; Obtain the management preferences of relevant staff for the required projects and the strategy stability index under various expected task execution scenarios, and solve the corresponding objective function using the particle swarm optimization algorithm to obtain the corresponding optimal response strategy.

10. The project management big data optimization method according to claim 9, characterized in that, The process of real-time monitoring of the execution of the corresponding optimal response strategy and providing feedback based on the monitoring results includes: A real-time monitoring mechanism is established to dynamically track the process of implementing the optimal response strategy. Key indicators of the corresponding project tasks are collected in real time during the execution of the optimal response strategy and compared with the pre-set expected goals. If the expected goals are not met, the project management information of the corresponding managed project is collected in real time and preprocessed. The dynamic subgraph is then updated based on the data, and the optimal response strategy being executed is optimized based on the updated dynamic subgraph. At the same time, the strategy execution results are fed back to the project knowledge graph to update the corresponding data nodes.