Visual project management method and system based on task tracking

Abstract

The invention provides a visual project management method and system based on task tracking, and relates to the technical field of project management, and the method comprises the steps: collecting project risk information through employing a context awareness technology, generating a risk propagation network, and calculating a risk diffusion coefficient; performing hierarchical analysis on project requirements based on the risk diffusion coefficient to obtain a task decomposition scheme, constructing a task transmission chain based on the task decomposition scheme and generating a visual task relation graph, and generating a task state transition rule and an alternative task path based on an early warning state of a task node, meanwhile, constructing a task elastic scheduling mechanism and storing the task elastic scheduling mechanism into a rule database; and identifying a preorder task node and a post-order task node on the task transmission chain, sending the task reconstruction instruction to the preorder task node and the post-order task node to realize dynamic task adjustment, updating the risk propagation network according to an adjustment result, and optimizing the risk diffusion coefficient.

Description

Technical Field

[0001] The present invention relates to project management technology, and in particular to a visual project management method and system based on task tracking. Background Art

[0002] With the increasing complexity of project management, traditional project management methods face many challenges in terms of task allocation, risk control and resource scheduling. In the prior art, the project management system has the following main deficiencies: the existing project management methods often adopt a static task decomposition method, lacking the ability to dynamically perceive and evaluate project risks in real time. During the project execution process, the inability to identify and respond to various risk factors in a timely manner can easily lead to task execution deviating from the expected goals. The traditional task transfer mechanism usually schedules tasks based on a preset execution order, and fails to fully consider the correlation between task nodes and the risk transfer characteristics. When the project environment changes, the fixed task transfer mode is difficult to adjust flexibly, affecting the efficiency of project execution. Due to the inability to accurately assess the risk status of task nodes, it is difficult to take effective preventive measures before the risk spreads, which can easily cause a chain reaction of project risks. In addition, the current task adjustment method mainly relies on manual experience judgment and lacks a scientific decision support mechanism. When task reorganization is required, it is difficult to quickly determine the optimal adjustment plan, affecting the overall execution effect of the project.

[0003] Therefore, there is an urgent need for a project management method that can achieve risk perception, task tracking and dynamic adjustment to improve the scientificity and effectiveness of project management. Summary of the invention

[0004] The embodiment of the present invention provides a visual project management method and system based on task tracking, which can solve the problems in the prior art.

[0005] According to a first aspect of the embodiments of the present invention,

[0006] Provides a visual project management method based on task tracking, including:

[0007] Use situational awareness technology to collect project risk information, which includes market risk data, technical risk data, resource risk data, and management risk data. Input the project risk information into a pre-trained risk assessment model for multi-dimensional cross-analysis, generate a risk propagation network, and calculate the risk diffusion coefficient. Based on the risk diffusion coefficient, perform a hierarchical analysis of project requirements to obtain a task decomposition plan. Based on the task decomposition plan, build a task transfer chain and generate a visual task relationship diagram. The visual task relationship diagram displays the execution order and correlation between task nodes.

[0008] An adaptive task execution network is constructed based on the task transfer chain, the relationship between task nodes is identified in the adaptive task execution network, and task constraints including risk thresholds, resource thresholds and time thresholds are established based on the risk propagation network. The risk transfer probability between task nodes is calculated in combination with the task constraints, and an early warning threshold is set according to the degree of association of the task nodes. The early warning state of the task node is determined based on the comparison result of the risk transfer probability and the early warning threshold, and task state migration rules and alternative task paths are generated based on the early warning state of the task node. At the same time, a task elastic scheduling mechanism is constructed and stored in a rule database;

[0009] Task execution data is collected and the risk diffusion coefficient is monitored. When the task execution data triggers the task constraint condition, the state migration rule and the alternative task path are extracted from the rule database. The state conversion direction of the task node is determined based on the state migration rule, and the optimal path is selected from the alternative task paths to reorganize the task. The task reconstruction instruction is generated, the predecessor task node and the successor task node on the task transfer chain are identified, and the task reconstruction instruction is sent to the predecessor task node and the successor task node to realize dynamic adjustment of the task. The risk propagation network is updated according to the adjustment result and the risk diffusion coefficient is optimized.

[0010] In an optional embodiment,

[0011] The project risk information is input into a pre-trained risk assessment model for multi-dimensional cross-analysis, a risk propagation network is generated and a risk diffusion coefficient is calculated. Based on the risk diffusion coefficient, the project requirements are hierarchically analyzed to obtain a task decomposition solution including:

[0012] Preprocessing the project risk data, extracting feature sequences from time series data through a sliding window, converting category data into numerical data through encoding, and normalizing the preprocessed data to obtain standardized risk data;

[0013] Constructing a multi-layer attention risk assessment model, the multi-layer attention risk assessment model includes an input layer, an intermediate layer and an output layer, wherein the input layer is provided with parallel long short-term memory units, each long short-term memory unit processes a type of standardized risk data respectively, the intermediate layer adopts a multi-head self-attention mechanism to realize interactive modeling of different risk dimensions to obtain attention scores, and the output layer constructs a risk propagation network based on the attention scores;

[0014] Setting risk nodes in the risk propagation network, determining edge weights between the risk nodes based on the attention scores, calculating risk diffusion coefficients using a random walk algorithm, constructing a judgment matrix based on the risk diffusion coefficients, and calculating weight vectors using an eigenvector method, combining the weight vectors with a fuzzy relationship matrix to obtain a demand priority vector, and stratifying project requirements based on the judgment matrix and the demand priority vector to obtain a demand hierarchy;

[0015] Based on the requirement hierarchy, the critical path method is used to decompose tasks, the priority relationship between tasks is determined according to the judgment matrix, and the task criticality is calculated in combination with the earliest task start time, the latest task start time and the risk diffusion coefficient. Under the condition of satisfying resource constraints and risk threshold constraints, the task decomposition plan is obtained by minimizing the sum of task criticalities.

[0016] In an optional embodiment,

[0017] Based on the requirement hierarchy, the critical path method is used to decompose tasks, the priority relationship between tasks is determined according to the judgment matrix, and the task criticality is calculated in combination with the earliest task start time, the latest task start time and the risk diffusion coefficient. Under the condition of satisfying resource constraints and risk threshold constraints, the task decomposition scheme is obtained by minimizing the sum of task criticalities, including:

[0018] A task dependency network is constructed based on a judgment matrix, wherein the task dependency network includes dependency relationships between task nodes, and the dependency strength between task nodes is calculated by a dependency function, wherein the dependency function integrates judgment matrix element values, task semantic similarity calculated based on a word vector model, and a temporal correlation coefficient based on time window overlap;

[0019] Based on the task dependency network, a dynamic programming algorithm is used to calculate the earliest start time and the latest start time of the task node, the earliest start time is recursively calculated based on the predecessor task node set, and the latest start time is reversely recursively calculated based on the successor task node set, and the task dependency network is topologically sorted to determine the task execution priority;

[0020] A time margin factor is calculated based on the earliest start time and the latest start time, a risk assessment factor is calculated based on the risk diffusion coefficient, a structural importance factor is calculated based on the topological characteristics of the task nodes in the task dependency network, and the time margin factor, the risk assessment factor and the structural importance factor are integrated in an adaptive weight combination manner to obtain the task criticality;

[0021] Construct a multi-objective task decomposition optimization model, with minimization of task criticality and maximization of resource utilization as optimization goals, and set resource constraints, risk threshold constraints, and task dependency constraints. The resource constraints include dynamic usage constraints on human resources, equipment resources, and material resources.

[0022] A multi-objective optimization algorithm is used to solve the task decomposition optimization model. The task sequence is represented by real number coding. An adaptive crossover operator and a local search mutation operator are designed to generate a new population. The adaptive crossover operator adjusts the crossover strategy according to the population diversity, and the local search mutation operator performs neighborhood search under dependency constraints. A comprehensive fitness function including task criticality, resource utilization and constraint violation is constructed, and the elite retention strategy and Pareto non-dominated sorting are used to obtain the task decomposition scheme.

[0023] In an optional embodiment,

[0024] The warning status of the task node is determined based on the comparison result between the risk transfer probability and the warning threshold. The task status migration rules and alternative task paths are generated based on the warning status of the task node, including:

[0025] Obtain the risk transfer probability and warning threshold of the task node, compare the risk transfer probability with the lowest threshold and the highest threshold of the warning threshold, divide the warning state of the task node into normal state, warning state and alarm state according to the comparison result, calculate the warning level under the warning state and the alarm level under the alarm state, and record the state feature vector under each warning state respectively, wherein the state feature vector includes risk index, resource occupation index and time progress index;

[0026] A state transition rule set is constructed based on the warning state, and the state transition rule set includes a migration rule from a normal state to a warning state that is triggered when the risk transfer probability is greater than a minimum threshold and the risk increment is greater than the risk tolerance; a migration rule from a warning state to an alarm state that is triggered when the risk transfer probability is greater than a maximum threshold or the alarm level is greater than a critical alarm level; a recovery rule from an alarm state to an alarm state that is triggered when the risk transfer probability is less than a maximum threshold and the potential impact is less than an impact threshold; a recovery rule from an alarm state to a normal state that is triggered when the risk transfer probability is less than a minimum threshold and the observation time is greater than the observation period;

[0027] The task nodes are constructed as a node set, the state migration rules are constructed as an edge set, a task migration network is constructed based on the node set and the edge set, the matching degree of state feature data between nodes in the task migration network is calculated to obtain state similarity, the resource consumption of state transition between nodes is calculated to obtain migration cost, and the migration weight between nodes is calculated according to the risk transfer probability, state similarity and migration cost;

[0028] Multiple candidate migration paths are generated according to the node set, edge set and migration weight. Each candidate migration path includes a node migration sequence and a corresponding state transition sequence. The cumulative value of the risk transfer probability of adjacent nodes in the node migration sequence is calculated to obtain the path risk. The number of state transition sequences that satisfy the state migration rules is calculated to obtain the path feasibility. The candidate migration path with the smallest weighted combination value of path risk and path feasibility is used as the alternative task path.

[0029] In an optional embodiment,

[0030] Calculating the cumulative value of the risk transfer probability of adjacent nodes in the node migration sequence to obtain the path risk, calculating the number of the state transition sequence that satisfies the state migration rule to obtain the path feasibility, and taking the candidate migration path with the smallest weighted combination value of the path risk and the path feasibility as the candidate task path includes:

[0031] Based on the node migration sequence, the historical risk event frequency and impact degree of adjacent task nodes are obtained, the risk transfer probability between nodes is calculated, the distance attenuation factor is calculated based on the network distance between the task nodes, and the actual risk transfer value between nodes is obtained by combining the risk transfer probability and the distance attenuation factor;

[0032] Determine the node importance weight according to the criticality of the task node in the task execution process, input the node importance weight and the actual risk transfer value between nodes into the pre-built risk accumulation model, and calculate the path risk of the candidate migration path;

[0033] Constructing a state transition compliance evaluation function based on a state transition rule set, inputting the state transition sequence into the state transition compliance evaluation function, and calculating the ratio of the number of state transitions that satisfy the state transition rule to the total length of the state sequence to obtain path feasibility;

[0034] The path risk and path feasibility are weighted and combined using the balance factor to construct a path quality assessment model and calculate the quality assessment index of the candidate migration path.

[0035] The risk transfer probability and state migration rules are randomly perturbed to establish a Monte Carlo simulation model, the candidate migration path is input into the Monte Carlo simulation model to verify the stability of the quality evaluation index, and the candidate migration path with the smallest quality evaluation index is selected as the alternative task path.

[0036] In an optional embodiment,

[0037] Extracting state transition rules and alternative task paths from a rule database, determining the state transition direction of a task node based on the state transition rules, and selecting the optimal path from the alternative task paths for task reorganization, generating task reorganization instructions including:

[0038] Acquire a state transition rule from a rule database, construct a state transition decision tree based on the state transition rule set, calculate a state transition direction through the state transition decision tree, and combine the state transition direction with the current state to form a target state;

[0039] Constructing a feature extraction model and a utility evaluation model, inputting the target state, task execution data and risk diffusion coefficient into the feature extraction model to extract state features, inputting the candidate path set into the feature extraction model to extract action features, inputting the state features and the action features into the utility evaluation model to fuse them to obtain a state-action representation, and calculating the utility evaluation value of the state-action representation through the utility evaluation model;

[0040] Calculating the path conversion cost, resource consumption and risk accumulation value of each path in the candidate path set, calculating the path weight value based on the path conversion cost, the resource consumption and the risk accumulation value, and sampling the candidate path set according to the path weight value to obtain a training sample;

[0041] Input the training samples into the feature extraction model and the utility evaluation model for training, calculate the loss function value and update the model parameters, and use the updated model to recalculate the utility evaluation value of the candidate path set to obtain an updated utility evaluation value;

[0042] Calculating probability distribution parameters based on the updated utility evaluation value, sampling using the probability distribution parameters to obtain sampling values, and selecting a path with a maximum sampling value as a reconstructed path;

[0043] A reconstruction instruction is generated according to the reconstruction path, and the reconstruction instruction includes a state migration instruction, a resource allocation instruction and an execution timing instruction.

[0044] In an optional embodiment,

[0045] Identifying a preceding task node and a succeeding task node on a task transfer chain, and sending the task reconstruction instruction to the preceding task node and the succeeding task node to realize dynamic task adjustment includes:

[0046] Constructing a directed communication network of a task delivery chain, setting task execution nodes as network nodes, setting task delivery relationships between the network nodes as network connection relationships, and generating an adjacency matrix of the task delivery chain based on the network nodes and the network connection relationships;

[0047] Using a depth-first search algorithm, obtaining a set of preceding task nodes of the node to be reconstructed based on a transposed matrix of the adjacency matrix, and obtaining a set of succeeding task nodes of the node to be reconstructed based on the adjacency matrix;

[0048] Performing semantic analysis on the task reconstruction instruction of the node to be reconstructed, obtaining the instruction execution dependency, and decomposing the task reconstruction instruction into a pre-order instruction subset and a post-order instruction subset based on the instruction execution dependency;

[0049] Encapsulating the preceding instruction subset into a preceding task adjustment message and encapsulating the following instruction subset into a following task adjustment message, wherein the task adjustment message includes a message header, an instruction body and verification information;

[0050] Sending a predecessor task adjustment message to the predecessor task node set through a directed communication network, and receiving the predecessor execution status information returned by the predecessor task node set; when the predecessor execution status information indicates that the task adjustment is completed, sending a subsequent task adjustment message to the subsequent task node set, and receiving the subsequent execution status information returned by the subsequent task node set;

[0051] receiving status monitoring data packets sent by the preceding task node set and the succeeding task node set according to a preset period, constructing a node status change model based on the status monitoring data packets, updating the risk propagation network and optimizing the risk diffusion coefficient;

[0052] When the node state change model detects a state abnormality, the task adjustment process is suspended, the node state is restored to the last stable state, and the task adjustment plan is regenerated.

[0053] According to a second aspect of the embodiments of the present invention,

[0054] Provides a visual project management system based on task tracking, including:

[0055] The first unit is used to collect project risk information using situational awareness technology, wherein the project risk information includes market risk data, technical risk data, resource risk data, and management risk data, input the project risk information into a pre-trained risk assessment model for multi-dimensional cross-analysis, generate a risk propagation network and calculate the risk diffusion coefficient, perform hierarchical analysis on project requirements based on the risk diffusion coefficient to obtain a task decomposition plan, construct a task transfer chain based on the task decomposition plan, and generate a visual task relationship diagram, wherein the visual task relationship diagram displays the execution order and correlation degree between task nodes;

[0056] The second unit is used to build an adaptive task execution network based on the task transfer chain, identify the relationship between task nodes in the adaptive task execution network, and establish task constraints including risk thresholds, resource thresholds and time thresholds based on the risk propagation network. The risk transfer probability between task nodes is calculated in combination with the task constraints, and an early warning threshold is set according to the degree of association of the task nodes. The early warning state of the task node is judged based on the comparison result of the risk transfer probability and the early warning threshold, and a task state migration rule and an alternative task path are generated based on the early warning state of the task node. At the same time, a task elastic scheduling mechanism is built and stored in a rule database;

[0057] The third unit is used to collect task execution data and monitor the risk diffusion coefficient. When the task execution data triggers the task constraint condition, the state migration rule and the alternative task path are extracted from the rule database, the state conversion direction of the task node is judged based on the state migration rule, and the optimal path is selected from the alternative task paths to reorganize the task, generate task reconstruction instructions, identify the predecessor task node and the successor task node on the task transfer chain, send the task reconstruction instructions to the predecessor task node and the successor task node to realize dynamic adjustment of the task, update the risk propagation network according to the adjustment result, and optimize the risk diffusion coefficient.

[0058] According to a third aspect of the embodiments of the present invention,

[0059] An electronic device is provided, comprising:

[0060] processor;

[0061] a memory for storing processor-executable instructions;

[0062] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0063] According to a fourth aspect of the embodiments of the present invention,

[0064] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0065] In this embodiment, multi-dimensional risk information is collected through situational awareness technology, and a risk assessment model is used for cross-analysis to generate a risk propagation network and a risk diffusion coefficient. This data-driven risk assessment method can identify and predict project risks more comprehensively and accurately, avoiding the subjectivity and lag of traditional risk management methods. At the same time, based on the risk diffusion coefficient, the project requirements are layered and analyzed, and the scope and degree of risk impact can be quantified, providing a scientific basis for formulating targeted risk response strategies. An adaptive task execution network is constructed, and multi-dimensional task constraints are established in combination with the risk propagation network. By calculating the risk transmission probability between task nodes and judging the warning status of task nodes according to the warning threshold, potential risks can be predicted in advance and the corresponding task adjustment mechanism can be triggered. The establishment of task state migration rules and alternative task paths, as well as the application of task elastic scheduling mechanism, make the project execution process more flexible, and can dynamically adjust tasks according to actual conditions, effectively respond to risks and changes, and avoid project delays or failures. The execution order and degree of association between task nodes are displayed through a visual task relationship diagram, which realizes the automation of dynamic task adjustment, reduces manual intervention and communication costs, and improves project management efficiency. At the same time, by selecting the optimal path to reorganize tasks, we can optimize resource allocation, improve resource utilization and reduce project costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 A flowchart of a visual project management method based on task tracking according to an embodiment of the present invention;

[0067] Figure 2 It is a structural diagram of a visual project management system based on task tracking according to an embodiment of the present invention. DETAILED DESCRIPTION

[0068] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0069] The technical solution of the present invention is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0070] Figure 1 FIG. 1 is a flow chart of a visual project management method based on task tracking according to an embodiment of the present invention. Figure 1 As shown, the method includes:

[0071] S101. Use situational awareness technology to collect project risk information, which includes market risk data, technical risk data, resource risk data, and management risk data. Input the project risk information into a pre-trained risk assessment model for multi-dimensional cross-analysis, generate a risk propagation network and calculate the risk diffusion coefficient, perform a hierarchical analysis of project requirements based on the risk diffusion coefficient to obtain a task decomposition plan, build a task delivery chain based on the task decomposition plan and generate a visual task relationship diagram, which displays the execution order and correlation between task nodes.

[0072] S102. Construct an adaptive task execution network based on the task transfer chain, identify the relationship between task nodes in the adaptive task execution network, and establish task constraints including risk thresholds, resource thresholds and time thresholds based on the risk propagation network. Calculate the risk transfer probability between task nodes in combination with the task constraints, set a warning threshold according to the degree of association of the task nodes, judge the warning status of the task node based on the comparison result of the risk transfer probability and the warning threshold, generate task status migration rules and alternative task paths based on the warning status of the task node, and construct a task elastic scheduling mechanism and store it in a rule database.

[0073] S103. Collect task execution data and monitor the risk diffusion coefficient. When the task execution data triggers the task constraint condition, extract the state migration rule and the alternative task path from the rule database, judge the state conversion direction of the task node based on the state migration rule, and select the optimal path from the alternative task path to reorganize the task, generate task reconstruction instructions, identify the predecessor task node and the successor task node on the task transfer chain, send the task reconstruction instructions to the predecessor task node and the successor task node to realize dynamic adjustment of the task, update the risk propagation network according to the adjustment result and optimize the risk diffusion coefficient.

[0074] Among them, situational awareness technology collects real-time information on market dynamics, technology updates, resource availability, team management efficiency, etc. related to the project through various data collection methods such as sensor networks, Internet of Things devices, and mobile terminals. This information will be structured to form a project risk information set including market risk data, technology risk data, resource risk data, and management risk data.

[0075] Project risk information is input into a pre-trained risk assessment model. The model can be built based on machine learning algorithms (such as support vector machines, neural networks, etc.) or a rule engine based on expert experience. The model will perform multi-dimensional cross-analysis on the input risk information, such as analyzing the correlation between market risk and technical risk, the impact of resource risk on project progress, and the impact of management risk on team collaboration efficiency. The analysis results will be presented in the form of a risk propagation network. At the same time, the model will also calculate the risk diffusion coefficient, which is used to quantify the scope and speed of risk propagation in the project. For example, the higher the coefficient, the wider the scope and faster the risk propagation.

[0076] Based on the risk diffusion coefficient, the project requirements are analyzed in layers and a task decomposition plan is formulated. According to the size of the risk diffusion coefficient, the project requirements are decomposed into subtasks of different levels. Requirements with high risk coefficients will be decomposed into more fine-grained subtasks to control risks more finely. According to the task decomposition plan, a task transfer chain is constructed and a visual task relationship diagram is generated. The task transfer chain describes the execution order and dependency relationship between task nodes. For example, task B can only be started after task A is completed. The visual task relationship diagram graphically displays the execution order and degree of association between task nodes, such as using arrows to represent the dependency relationship between tasks and using different colors to represent the risk level of tasks.

[0077] Combined with the task constraints, calculate the risk transfer probability between task nodes. For example, if the risk of task A is high and task B can only be started after task A is completed, the risk of task A will be transferred to task B, and the risk probability of task B needs to be calculated. Set the warning threshold according to the degree of correlation between task nodes. For example, if the correlation between task A and task B is very high, set a lower warning threshold to detect potential risks early.

[0078] Compare the risk transfer probability with the warning threshold to determine the warning status of the task node. If the risk transfer probability exceeds the warning threshold, the task node is in the warning state and corresponding measures need to be taken. Based on the warning state of the task node, generate task state migration rules and alternative task paths. For example, if task A is in the warning state, the state of task A can be migrated from "in progress" to "paused", and an alternative task path can be found to complete the goal of task A. Store these rules and paths in the rule database.

[0079] Task execution data includes task completion progress, resource consumption, actual risk occurrence, etc. When task execution data triggers task constraints, such as when the risk probability of a task exceeds the warning threshold, the corresponding task state transition rules and alternative task paths are extracted from the rule database.

[0080] In this embodiment, by real-time monitoring of the risk diffusion coefficient and task execution data, potential risks can be discovered in time and corresponding measures can be taken to avoid risk expansion and improve risk management efficiency. Through the adaptive task execution network and task elastic scheduling mechanism, the task execution path can be dynamically adjusted according to the actual situation, the flexibility of project execution can be improved, and changes and uncertainties can be better coped with. By optimizing the task decomposition plan and task execution path, project risks can be reduced, resource utilization efficiency can be improved, and ultimately the success rate of the project can be improved.

[0081] In an optional implementation, the project risk information is input into a pre-trained risk assessment model for multi-dimensional cross-analysis, a risk propagation network is generated and a risk diffusion coefficient is calculated, and a hierarchical analysis of project requirements based on the risk diffusion coefficient is performed to obtain a task decomposition solution including:

[0082] Preprocessing the project risk data, extracting feature sequences from time series data through a sliding window, converting category data into numerical data through encoding, and normalizing the preprocessed data to obtain standardized risk data;

[0083] Constructing a multi-layer attention risk assessment model, the multi-layer attention risk assessment model includes an input layer, an intermediate layer and an output layer, wherein the input layer is provided with parallel long short-term memory units, each long short-term memory unit processes a type of standardized risk data respectively, the intermediate layer adopts a multi-head self-attention mechanism to realize interactive modeling of different risk dimensions to obtain attention scores, and the output layer constructs a risk propagation network based on the attention scores;

[0084] Setting risk nodes in the risk propagation network, determining edge weights between the risk nodes based on the attention scores, calculating risk diffusion coefficients using a random walk algorithm, constructing a judgment matrix based on the risk diffusion coefficients, and calculating weight vectors using an eigenvector method, combining the weight vectors with a fuzzy relationship matrix to obtain a demand priority vector, and stratifying project requirements based on the judgment matrix and the demand priority vector to obtain a demand hierarchy;

[0085] Based on the requirement hierarchy, the critical path method is used to decompose tasks, the priority relationship between tasks is determined according to the judgment matrix, and the task criticality is calculated in combination with the earliest task start time, the latest task start time and the risk diffusion coefficient. Under the condition of satisfying resource constraints and risk threshold constraints, the task decomposition plan is obtained by minimizing the sum of task criticalities.

[0086] Exemplarily, first, the project risk data is preprocessed. Assume that the collected project risk data contains information such as the probability of risk occurrence, the loss caused by the risk, the time of risk occurrence, and the risk category. For time series data such as the time of risk occurrence, the sliding window method is used to extract the feature sequence. For example, taking three months as a time window, the average value of the risk occurrence probability in the past three months is used as the risk probability feature of the current time. For categorical data such as risk category, it is converted into numerical data using the unique hot encoding method. For example, assuming that the risk category is divided into three categories: technical risk, management risk, and environmental risk, then technical risk can be encoded as [1, 0, 0], management risk can be encoded as [0, 1, 0], and environmental risk can be encoded as [0, 0, 1]. Finally, the preprocessed data is normalized, for example, the Min-Max normalization method is used to scale all feature data to between 0 and 1 to obtain standardized risk data.

[0087] Next, a multi-layer attention risk assessment model is constructed. The model consists of an input layer, an intermediate layer, and an output layer. The input layer is equipped with parallel long short-term memory units (LSTMs), and each LSTM unit processes a type of standardized risk data, for example, one LSTM unit processes risk probability data, and another LSTM unit processes risk loss data. The intermediate layer uses a multi-head self-attention mechanism to achieve interactive modeling between different risk dimensions. Specifically, the output of each LSTM unit is used as the input of the multi-head self-attention, and the correlation between different risk features is captured by calculating the attention scores between different risk dimensions. The output layer constructs a risk propagation network based on the attention scores.

[0088] In the risk propagation network, risk nodes are set, each node represents a risk event. The edge weights between risk nodes are determined based on the attention score calculated by the middle layer. The higher the attention score, the stronger the correlation between the two risk events, and the larger the corresponding edge weight. Then, the random walk algorithm is used to calculate the risk diffusion coefficient. The risk diffusion coefficient indicates the scope and speed of risk propagation in the network.

[0089] Construct a judgment matrix based on the risk diffusion coefficient. The judgment matrix is ​​used to represent the relative importance of different requirements. For example, if the risk diffusion coefficient of requirement A is greater than that of requirement B, then requirement A is considered more important than requirement B, and the value of A relative to B in the judgment matrix will be greater than 1. The eigenvector method is used to calculate the weight vector, which represents the relative importance of each requirement. The weight vector is combined with the fuzzy relationship matrix to obtain the requirement priority vector. The fuzzy relationship matrix represents the interdependence between requirements.

[0090] Based on the judgment matrix and the demand priority vector, the project requirements are layered to obtain the demand hierarchy. For example, the requirements can be divided into three levels: high priority requirements, medium priority requirements, and low priority requirements.

[0091] Based on the requirements hierarchy, the critical path method is used to decompose tasks. The priority relationship between tasks is determined according to the judgment matrix. The task criticality is calculated by combining the earliest task start time, the latest task start time and the risk diffusion coefficient. The task criticality indicates the degree of impact of the task on the overall progress of the project. Under the condition of satisfying the resource constraints and risk threshold constraints, the task decomposition plan is obtained by minimizing the sum of task criticalities. For example, assuming that the total project budget is 1 million yuan and the risk threshold is 0.5, the final task decomposition plan needs to meet the budget constraint and risk constraint.

[0092] Assume a software development project and identify three risks: requirement changes, technical difficulties, and staff turnover. Through data collection and preprocessing, standardized risk data is obtained. After calculation by the multi-layer attention risk assessment model, the risk propagation network and risk diffusion coefficient are obtained. Assume that the risk diffusion coefficient of requirement changes is 0.8, technical difficulties is 0.6, and staff turnover is 0.5. Construct a judgment matrix based on the risk diffusion coefficient and calculate the requirement priority vector. Finally, decompose the project requirements into multiple subtasks, and arrange the task execution order according to the task criticality and resource constraints.

[0093] In this embodiment, the multi-layer attention risk assessment model can capture the correlation between different risk dimensions, so as to more accurately assess project risks. By decomposing tasks based on risk diffusion coefficients and demand priorities, limited resources can be allocated to high-risk and high-priority tasks first, thereby improving resource utilization efficiency. By performing layered analysis and task decomposition of project requirements, the overall progress and risk status of the project can be clearly understood, thereby better controlling project progress and costs.

[0094] In an optional implementation, based on the requirement hierarchy, the critical path method is used to decompose tasks, the priority relationship between tasks is determined according to the judgment matrix, and the task criticality is calculated in combination with the earliest task start time, the latest task start time and the risk diffusion coefficient. Under the condition of satisfying resource constraints and risk threshold constraints, the task decomposition scheme is obtained by minimizing the sum of task criticalities, including:

[0095] A task dependency network is constructed based on a judgment matrix, wherein the task dependency network includes dependency relationships between task nodes, and the dependency strength between task nodes is calculated by a dependency function, wherein the dependency function integrates judgment matrix element values, task semantic similarity calculated based on a word vector model, and a temporal correlation coefficient based on time window overlap;

[0096] Based on the task dependency network, a dynamic programming algorithm is used to calculate the earliest start time and the latest start time of the task node, the earliest start time is recursively calculated based on the predecessor task node set, and the latest start time is reversely recursively calculated based on the successor task node set, and the task dependency network is topologically sorted to determine the task execution priority;

[0097] A time margin factor is calculated based on the earliest start time and the latest start time, a risk assessment factor is calculated based on the risk diffusion coefficient, a structural importance factor is calculated based on the topological characteristics of the task nodes in the task dependency network, and the time margin factor, the risk assessment factor and the structural importance factor are integrated in an adaptive weight combination manner to obtain the task criticality;

[0098] Construct a multi-objective task decomposition optimization model, with minimization of task criticality and maximization of resource utilization as optimization goals, and set resource constraints, risk threshold constraints, and task dependency constraints. The resource constraints include dynamic usage constraints on human resources, equipment resources, and material resources.

[0099] A multi-objective optimization algorithm is used to solve the task decomposition optimization model. The task sequence is represented by real number coding. An adaptive crossover operator and a local search mutation operator are designed to generate a new population. The adaptive crossover operator adjusts the crossover strategy according to the population diversity, and the local search mutation operator performs neighborhood search under dependency constraints. A comprehensive fitness function including task criticality, resource utilization and constraint violation is constructed, and the elite retention strategy and Pareto non-dominated sorting are used to obtain the task decomposition scheme.

[0100] For example, first, the requirements are decomposed hierarchically to form a tree-structured requirements hierarchy diagram. For example, the requirements of a software development project can be decomposed into modules such as user interface, database, and background logic, and each module can be further decomposed into more fine-grained functional points. In this process, referring to Maslow's hierarchy of needs theory, the requirements are classified according to importance and urgency, for example, the core needs of users are placed at a higher level, while the secondary needs are placed at a lower level.

[0101] Next, construct a task judgment matrix. Compare the tasks in the requirements hierarchy diagram in pairs by expert scoring or group discussion to determine the priority relationship between them. For example, if task A is more important than task B, the cells corresponding to A and B in the matrix are assigned a value of 3, indicating that the importance of A is significantly higher than that of B. Otherwise, the value is 1 / 3. If the two tasks are equally important, the value is 1.

[0102] Then, a task dependency network is constructed based on the judgment matrix. The task dependency network is a directed graph, where nodes represent tasks and edges represent dependencies between tasks. The dependency function takes into account three factors: the judgment matrix element value, task semantic similarity, and temporal correlation coefficient. For example, if the completion of task A is a prerequisite for the start of task B, the weight of the edge from A to B will be higher. Task semantic similarity is calculated using a word vector model. For example, the task description is converted into a vector using the Word2Vec model, and the cosine similarity between the vectors is calculated. The temporal correlation coefficient is calculated based on the overlap of the time windows. For example, if the planned execution time periods of two tasks overlap a lot, their temporal correlation coefficients are also high.

[0103] On this basis, a dynamic programming algorithm is used to calculate the earliest start time and the latest start time of each task. The earliest start time refers to the earliest time after all predecessor tasks of the task are completed, and the latest start time refers to the latest time that the task can start without affecting the overall completion time of the project. For example, if Task A depends on Task B and Task C, then the earliest start time of Task A is the later of the latest completion times of Task B and Task C.

[0104] Next, calculate the task criticality. Task criticality is obtained by weighted summation of the time margin factor, risk assessment factor, and structural importance factor. The time margin factor is equal to the latest start time minus the earliest start time. The smaller the time margin, the more critical the task. The risk assessment factor is calculated based on the risk diffusion coefficient. The risk diffusion coefficient indicates the degree of impact of task risk on other tasks. The larger the risk diffusion coefficient, the higher the risk assessment factor. The structural importance factor is calculated based on the topological characteristics of the task in the dependency network. For example, if a task is a predecessor task of many other tasks, its structural importance factor will be higher. The weights are adjusted adaptively according to the specific situation of the project.

[0105] Construct a multi-objective task decomposition optimization model. The optimization goal is to minimize the sum of task criticality and maximize resource utilization. Resource constraints include dynamic usage constraints of human resources, equipment resources, and material resources. Risk threshold constraints limit the overall risk level of the project. Task dependency constraints ensure the order of task execution. For example, a project needs to be completed within 10 days, and a maximum of 3 human resources can be invested per day.

[0106] A multi-objective optimization algorithm is used to solve the model. The task sequence is represented by real number coding. For example, [1, 3, 2] means that task 1 is executed first, then task 3, and finally task 2. An adaptive crossover operator and a local search mutation operator are designed to generate a new task sequence. The adaptive crossover operator adjusts the crossover strategy according to the population diversity. For example, if the population diversity is low, the crossover probability is increased. The local search mutation operator performs neighborhood search under dependency constraints. For example, the order of two adjacent tasks is swapped. The fitness function comprehensively considers task criticality, resource utilization, and constraint violation. The elite retention strategy and Pareto non-dominated sorting are used to obtain the final task decomposition solution.

[0107] For example, a project contains three tasks A, B and C. Task A depends on task B, and task C is independent of A and B. Assume that the duration of tasks A, B and C is 2 days, 3 days and 1 day respectively, the human resource requirements are 2, 1 and 3 people respectively, and the risk diffusion coefficients are 0.8, 0.5 and 0.2 respectively. Through the above method, an optimal task decomposition plan can be obtained, for example, task B is executed first, then task A, and finally task C, with a total duration of 5 days, satisfying the resource constraint conditions and risk threshold constraint conditions, and the total task criticality is minimized.

[0108] In this embodiment, by decomposing and prioritizing tasks, the project schedule can be arranged more effectively, task conflicts and delays can be reduced, and the efficiency of project completion can be improved. By considering the risk diffusion coefficient and risk threshold constraints, project risks can be effectively identified and controlled, risk diffusion and escalation can be avoided, and the possibility of project failure can be reduced. By considering resource constraints and the goal of maximizing resource utilization, resource allocation can be optimized, resource utilization efficiency can be improved, and project costs can be reduced.

[0109] In an optional implementation, judging the warning state of the task node based on the comparison result of the risk transfer probability and the warning threshold, and generating the task state migration rule and the alternative task path based on the warning state of the task node include:

[0110] Obtain the risk transfer probability and warning threshold of the task node, compare the risk transfer probability with the lowest threshold and the highest threshold of the warning threshold, divide the warning state of the task node into normal state, warning state and alarm state according to the comparison result, calculate the warning level under the warning state and the alarm level under the alarm state, and record the state feature vector under each warning state respectively, wherein the state feature vector includes risk index, resource occupation index and time progress index;

[0111] A state transition rule set is constructed based on the warning state, and the state transition rule set includes a migration rule from a normal state to a warning state that is triggered when the risk transfer probability is greater than a minimum threshold and the risk increment is greater than the risk tolerance; a migration rule from a warning state to an alarm state that is triggered when the risk transfer probability is greater than a maximum threshold or the alarm level is greater than a critical alarm level; a recovery rule from an alarm state to an alarm state that is triggered when the risk transfer probability is less than a maximum threshold and the potential impact is less than an impact threshold; a recovery rule from an alarm state to a normal state that is triggered when the risk transfer probability is less than a minimum threshold and the observation time is greater than the observation period;

[0112] The task nodes are constructed as a node set, the state migration rules are constructed as an edge set, a task migration network is constructed based on the node set and the edge set, the matching degree of state feature data between nodes in the task migration network is calculated to obtain state similarity, the resource consumption of state transition between nodes is calculated to obtain migration cost, and the migration weight between nodes is calculated according to the risk transfer probability, state similarity and migration cost;

[0113] Multiple candidate migration paths are generated according to the node set, edge set and migration weight. Each candidate migration path includes a node migration sequence and a corresponding state transition sequence. The cumulative value of the risk transfer probability of adjacent nodes in the node migration sequence is calculated to obtain the path risk. The number of state transition sequences that satisfy the state migration rules is calculated to obtain the path feasibility. The candidate migration path with the smallest weighted combination value of path risk and path feasibility is used as the alternative task path.

[0114] Exemplarily, first, the risk transfer probability and warning threshold of each task node are obtained. The risk transfer probability indicates the possibility of risk transfer from one node to another, and the warning threshold defines the boundary of different warning states. For example, the risk transfer probability of task node A is 0.6, the minimum threshold of the warning threshold is 0.3, and the maximum threshold is 0.8.

[0115] Next, the risk transfer probability is compared with the warning threshold to determine the warning state of the task node. If the risk transfer probability is less than the minimum threshold, the task node is in a normal state; if the risk transfer probability is greater than or equal to the minimum threshold and less than the maximum threshold, the task node is in a warning state; if the risk transfer probability is greater than or equal to the maximum threshold, the task node is in an alarm state. In the above example, the risk transfer probability of task node A, 0.6, is between the minimum threshold of 0.3 and the maximum threshold of 0.8, so task node A is in a warning state.

[0116] For task nodes in the warning state, calculate their warning level. The warning level can be calculated based on the difference between the risk transfer probability and the lowest threshold. The larger the difference, the higher the warning level. For example, the warning level can be defined as the difference between the risk transfer probability and the lowest threshold divided by the lowest threshold. Assuming that the risk transfer probability of task node B is 0.4 and the lowest threshold is 0.3, the warning level of task node B is (0.4-0.3) / 0.3=0.33. For task nodes in the alarm state, calculate their alarm level. The alarm level can be calculated based on the difference between the risk transfer probability and the highest threshold. The larger the difference, the higher the alarm level. For example, the alarm level can be defined as the difference between the risk transfer probability and the highest threshold divided by (1-highest threshold). Assuming that the risk transfer probability of task node C is 0.9 and the highest threshold is 0.8, the alarm level of task node C is (0.9-0.8) / (1-0.8)=0.5.

[0117] At the same time, the state feature vector of each warning state is recorded. The state feature vector contains risk index, resource occupancy index and time progress index. For example, the state feature vector of task node A can be expressed as [risk index: 0.6, resource occupancy index: 0.8, time progress index: 0.5].

[0118] A set of state transition rules is constructed based on the warning state. The state transition rules define the transition conditions between different warning states. For example, when the risk transfer probability is greater than the minimum threshold and the risk increment is greater than the risk tolerance, the migration rule from the normal state to the warning state is triggered; when the risk transfer probability is greater than the maximum threshold or the warning level is greater than the critical warning level, the migration rule from the warning state to the alarm state is triggered; when the risk transfer probability is less than the maximum threshold and the potential impact is less than the impact threshold, the recovery rule from the alarm state to the warning state is triggered; when the risk transfer probability is less than the minimum threshold and the observation time is greater than the observation period, the recovery rule from the alarm state to the normal state is triggered.

[0119] The task nodes are constructed as node sets, the state transition rules are constructed as edge sets, and the task transition network is constructed based on the node sets and edge sets. For example, task nodes A, B, and C constitute a node set, and the state transition rules constitute an edge set, which together constitute a task transition network.

[0120] The matching degree of state feature data between nodes in the task migration network is calculated to obtain state similarity, the resource consumption of state transition between nodes is calculated to obtain migration cost, and the migration weight between nodes is calculated based on risk transfer probability, state similarity and migration cost. For example, if the state similarity between nodes A and B is 0.8, the migration cost is 0.2, and the risk transfer probability is 0.6, then the migration weight from node A to node B can be calculated by the weighted combination of these three values.

[0121] Generate multiple candidate migration paths based on the node set, edge set and migration weight. Each candidate migration path contains a node migration sequence and a corresponding state transition sequence. For example, a candidate migration path can be expressed as: A->B->C, and the corresponding state transition sequence is: normal->warning->alarm.

[0122] Calculate the path risk and path feasibility of each candidate migration path. The path risk can be calculated by the cumulative value of the risk transfer probability of adjacent nodes in the node migration sequence. The path feasibility can be calculated by the number of state transition sequences that satisfy the state transition rules. For example, assuming that in the candidate path A->B->C, the risk transfer probability of A->B is 0.6, and the risk transfer probability of B->C is 0.7, then the risk of the path is 0.6+0.7=1.3. Assuming that the state transition sequence of the path satisfies 3 state transition rules, the feasibility of the path is 3.

[0123] The candidate migration path with the smallest weighted combination value of path risk and path feasibility is selected as the alternative task path.

[0124] In this embodiment, by combining the risk transfer probability, warning threshold and state feature vector, the warning state of the task node can be judged more accurately to avoid misjudgment and missed judgment. By generating alternative task paths, resources can be planned in advance to avoid resource waste and improve resource utilization efficiency. By selecting the alternative task path with the best risk and feasibility, potential risks can be effectively avoided and the success rate of task execution can be improved.

[0125] In an optional implementation, calculating the cumulative value of the risk transfer probability of adjacent nodes in the node migration sequence to obtain the path risk, calculating the number of the state transition sequence that satisfies the state migration rule to obtain the path feasibility, and taking the candidate migration path with the smallest weighted combination value of the path risk and the path feasibility as the candidate task path includes:

[0126] Based on the node migration sequence, the historical risk event frequency and impact degree of adjacent task nodes are obtained, the risk transfer probability between nodes is calculated, the distance attenuation factor is calculated based on the network distance between the task nodes, and the actual risk transfer value between nodes is obtained by combining the risk transfer probability and the distance attenuation factor;

[0127] Determine the node importance weight according to the criticality of the task node in the task execution process, input the node importance weight and the actual risk transfer value between nodes into the pre-built risk accumulation model, and calculate the path risk of the candidate migration path;

[0128] Constructing a state transition compliance evaluation function based on a state transition rule set, inputting the state transition sequence into the state transition compliance evaluation function, and calculating the ratio of the number of state transitions that satisfy the state transition rule to the total length of the state sequence to obtain path feasibility;

[0129] The path risk and path feasibility are weighted and combined using the balance factor to construct a path quality assessment model and calculate the quality assessment index of the candidate migration path.

[0130] The risk transfer probability and state migration rules are randomly perturbed to establish a Monte Carlo simulation model, the candidate migration path is input into the Monte Carlo simulation model to verify the stability of the quality evaluation index, and the candidate migration path with the smallest quality evaluation index is selected as the alternative task path.

[0131] Exemplarily, first, it is necessary to model the risk transfer between task nodes. Obtain the number of historical risk events that occurred in adjacent task nodes and the degree of impact caused by these events. For example, the historical records from task node A to task node B show that three risk events occurred in the past year, and the impact of each event was assessed as 0.8, 0.5, and 0.2, respectively, with a numerical range between 0 and 1, and the larger the value, the greater the impact. Based on these historical data, calculate the probability of risk transfer from node A to node B.

[0132] Next, consider the network distance between task nodes. Assuming that the network distance from node A to node B is 10, the distance attenuation factor is calculated based on the distance. The longer the distance, the smaller the attenuation factor. For example, an exponential attenuation function can be used here to obtain a distance attenuation factor of 0.7. Multiply the risk transfer probability by the distance attenuation factor to obtain the actual risk transfer value from node A to node B.

[0133] Then, the node importance weight is determined according to the criticality of the task node in the task execution process. For example, node A is the starting node of the entire task and is very critical, so its importance weight is set to 0.9; node B is an intermediate node and its importance weight is set to 0.6. The node importance weight and the actual risk transfer value between nodes are input into the pre-built risk accumulation model, which can be a simple weighted summation model, to calculate the path risk of the candidate migration path. Assuming that a candidate path contains nodes A, B, and C, the actual risk transfer values ​​from A to B and from B to C are multiplied by the importance weights of the corresponding nodes, and then the results are added to obtain the path risk of the candidate path, for example, the calculated result is 0.65.

[0134] Next, the feasibility of the candidate migration path needs to be evaluated. A state transition compliance evaluation function is constructed based on a predefined set of state transition rules. For example, the rule set can be defined as: the task state must be converted from "pending" to "in progress" and then to "completed". The state transition sequence of the candidate path is input into the state transition compliance evaluation function. For example, a state transition sequence is "pending" -> "in progress" -> "completed" -> "closed", where "pending" -> "in progress" -> "completed" complies with the predefined rules. Calculate the ratio of the number of state transitions that meet the state transition rules to the total length of the state sequence to obtain the path feasibility. For example, the path feasibility of this example is 3 / 4=0.75.

[0135] The path quality evaluation model is constructed by weighting the path risk and path feasibility using the balance factor. For example, if the balance factor is set to 0.5, the path quality evaluation index = 0.5*path risk + 0.5*(1-path feasibility). According to the above case, the path quality evaluation index = 0.5*0.65+0.5*(1-0.75)=0.45.

[0136] Finally, randomly perturb the risk transfer probability and state transition rules to establish a Monte Carlo simulation model. For example, add or subtract a small random number to the risk transfer probability, and randomly modify certain conditions of the state transition rule. Input the candidate migration path into the Monte Carlo simulation model, run the simulation multiple times, and verify the stability of the quality assessment index. For example, perform 1,000 simulations to observe the distribution of the quality assessment index. Select the candidate migration path with the smallest quality assessment index as the candidate task path.

[0137] In this embodiment, by considering the risk transfer probability and the distance attenuation factor, the path risk is more accurately evaluated to avoid selecting high-risk paths. By considering the state transition rules, the feasibility of the path is ensured, and task execution interruption or failure due to illegal state transition is avoided. Through Monte Carlo simulation, the stability of the path quality evaluation index is verified, so that the selected path can still maintain good performance in the face of uncertain factors.

[0138] In an optional implementation, extracting state transition rules and alternative task paths from a rule database, determining the state transition direction of a task node based on the state transition rules, and selecting an optimal path from the alternative task paths for task reorganization, generating a task reorganization instruction includes:

[0139] Acquire a state transition rule from a rule database, construct a state transition decision tree based on the state transition rule set, calculate a state transition direction through the state transition decision tree, and combine the state transition direction with the current state to form a target state;

[0140] Constructing a feature extraction model and a utility evaluation model, inputting the target state, task execution data and risk diffusion coefficient into the feature extraction model to extract state features, inputting the candidate path set into the feature extraction model to extract action features, inputting the state features and the action features into the utility evaluation model to fuse them to obtain a state-action representation, and calculating the utility evaluation value of the state-action representation through the utility evaluation model;

[0141] Calculating the path conversion cost, resource consumption and risk accumulation value of each path in the candidate path set, calculating the path weight value based on the path conversion cost, the resource consumption and the risk accumulation value, and sampling the candidate path set according to the path weight value to obtain a training sample;

[0142] Input the training samples into the feature extraction model and the utility evaluation model for training, calculate the loss function value and update the model parameters, and use the updated model to recalculate the utility evaluation value of the candidate path set to obtain an updated utility evaluation value;

[0143] Calculating probability distribution parameters based on the updated utility evaluation value, sampling using the probability distribution parameters to obtain sampling values, and selecting a path with a maximum sampling value as a reconstructed path;

[0144] A reconstruction instruction is generated according to the reconstruction path, and the reconstruction instruction includes a state migration instruction, a resource allocation instruction and an execution timing instruction.

[0145] In order to extract state transition rules from the rule database and reconstruct tasks, we first need to build a rule database. The database contains various task states and the rules for transitions between states. For example, state A can be transformed into state B if condition C is met and resource D is consumed. Specific examples include: the task state "pending" can be transformed into "processing" if condition "resources available" is met and resource "manpower" is consumed.

[0146] Next, the state transition rules are extracted from the rule database, and a state transition decision tree is constructed based on these rules. Each node of the decision tree represents a task state, and the branches represent the transition conditions and resource consumption between states. For example: the branch from the "pending" state to the "processing" state has the condition "resources available" and the resource consumption "manpower". By traversing the decision tree, the state transition direction of the task node can be determined. For example: the current state is "pending", if "resources available", the target state is "processing".

[0147] Then, construct a feature extraction model and a utility evaluation model. The feature extraction model is used to extract state features and action features. For example, state features may include the current task state, remaining resources, etc.; action features may include the length of the candidate path, risk level, etc. The utility evaluation model is used to fuse state features and action features, and calculate the utility evaluation value of the state action representation. For example, inputting the state feature "remaining resources" and the action feature "resource consumption" into the utility evaluation model can obtain the utility evaluation value of the state action. Assuming that the current state is "pending", the remaining resources are 10, the resource consumption of candidate path 1 is 2, and the resource consumption of candidate path 2 is 5, the utility evaluation value of path 1 may be higher than that of path 2.

[0148] Next, calculate the path conversion cost, resource consumption, and risk accumulation value of each path in the candidate path set. The path conversion cost can be defined as the number of steps required for state transition, the resource consumption can be defined as the sum of all resources required to complete the path, and the risk accumulation value can be defined as the sum of the risk values ​​of all nodes on the path. For example: the conversion cost of path 1 is 3, the resource consumption is 6, and the risk accumulation value is 1; the conversion cost of path 2 is 2, the resource consumption is 8, and the risk accumulation value is 2. Based on these values, calculate the path weight value. For example, the path weight value can be obtained by weighting the conversion cost, resource consumption, and risk accumulation value according to the preset weight coefficient. Assume that the weight value of path 1 is 0.8 and the weight value of path 2 is 0.6.

[0149] Then, the candidate path set is sampled according to the path weight value to obtain the training sample. For example, weighted random sampling can be performed according to the weight value, and the path with a higher weight value has a greater probability of being extracted. The training sample is input into the feature extraction model and the utility evaluation model for training, and the model parameters are updated. For example, an optimization algorithm such as gradient descent can be used to minimize the loss function and update the model parameters. The utility evaluation value of the candidate path set is recalculated using the updated model.

[0150] Calculate the probability distribution parameters based on the updated utility evaluation value. For example, the utility evaluation value can be converted into a probability value using a softmax function. Sampling is performed using the probability distribution parameters to obtain a sampling value. For example, random sampling can be performed according to the probability distribution. Select the path with the maximum sampling value as the reconstruction path. For example, assuming that the sampling value of path 1 is 0.7 and the sampling value of path 2 is 0.3, path 1 is selected as the reconstruction path.

[0151] Finally, the reconstruction instructions are generated according to the reconstruction path, including state migration instructions, resource allocation instructions, and execution timing instructions. For example, if the reconstruction path is "pending" -> "processing" -> "completed", the generated reconstruction instructions may include: migrate the task status from "pending" to "processing", allocate resources "manpower", and execute at "2024-01-01 10:00:00"; migrate the task status from "processing" to "completed", reclaim resources "manpower", and execute at "2024-01-01 12:00:00".

[0152] In this embodiment, by reconstructing the task path, the steps and resources required to complete the task can be reduced, thereby improving the efficiency of task execution. By considering the risk diffusion coefficient, a path with lower risk can be selected, thereby reducing the risk during task execution. By optimizing resource allocation instructions, existing resources can be used more effectively to avoid resource waste.

[0153] In an optional implementation, identifying a preceding task node and a succeeding task node on a task delivery chain, and sending the task reconstruction instruction to the preceding task node and the succeeding task node to implement dynamic task adjustment includes:

[0154] Constructing a directed communication network of a task delivery chain, setting task execution nodes as network nodes, setting task delivery relationships between the network nodes as network connection relationships, and generating an adjacency matrix of the task delivery chain based on the network nodes and the network connection relationships;

[0155] Using a depth-first search algorithm, obtaining a set of preceding task nodes of the node to be reconstructed based on a transposed matrix of the adjacency matrix, and obtaining a set of succeeding task nodes of the node to be reconstructed based on the adjacency matrix;

[0156] Performing semantic analysis on the task reconstruction instruction of the node to be reconstructed, obtaining the instruction execution dependency, and decomposing the task reconstruction instruction into a pre-order instruction subset and a post-order instruction subset based on the instruction execution dependency;

[0157] Encapsulating the preceding instruction subset into a preceding task adjustment message and encapsulating the following instruction subset into a following task adjustment message, wherein the task adjustment message includes a message header, an instruction body and verification information;

[0158] Sending a predecessor task adjustment message to the predecessor task node set through a directed communication network, and receiving the predecessor execution status information returned by the predecessor task node set; when the predecessor execution status information indicates that the task adjustment is completed, sending a subsequent task adjustment message to the subsequent task node set, and receiving the subsequent execution status information returned by the subsequent task node set;

[0159] receiving status monitoring data packets sent by the preceding task node set and the succeeding task node set according to a preset period, constructing a node status change model based on the status monitoring data packets, updating the risk propagation network and optimizing the risk diffusion coefficient;

[0160] When the node state change model detects a state abnormality, the task adjustment process is suspended, the node state is restored to the last stable state, and the task adjustment plan is regenerated.

[0161] Exemplarily, first, a directed communication network of the task transfer chain is constructed. Each task execution node is abstracted as a node in the network, and the task transfer relationship between the nodes is represented as a directed connection in the network. For example, suppose there are three task nodes A, B, and C. After A is completed, the result is passed to B, and after B is completed, the result is passed to C. Then there is a directed connection between A and B, and between B and C. Based on the node and connection relationship, an adjacency matrix of the task transfer chain is generated. In this example, the adjacency matrix can be represented as a 3x3 matrix, in which the first row and second column element is 1, indicating that there is a connection from A to B; the second row and third column element is 1, indicating that there is a connection from B to C; the other elements are 0, indicating that there is no connection.

[0162] Next, the depth-first search algorithm is used to identify the pre-order and post-order task node sets of the node to be reconstructed. Taking node B as an example, using the transposed matrix of the adjacency matrix to perform a depth-first search, the pre-order node set {A} of B can be found. Using the original adjacency matrix to perform a depth-first search, the post-order node set {C} of B can be found.

[0163] Then, semantic parsing is performed on the task reconfiguration instructions of the node to be reconfigured. For example, suppose the task reconfiguration instruction of node B is "modify the value of parameter X, and then re-execute the task". Through semantic parsing, it can be identified that the instruction contains two dependencies: first modify the value of parameter X, and then re-execute the task. Based on the parsing results, the task reconfiguration instruction is decomposed into a pre-order instruction subset {modify the value of parameter X} and a post-order instruction subset {re-execute the task}.

[0164] The pre-order instruction subset and the post-order instruction subset are encapsulated into the pre-order task adjustment message and the post-order task adjustment message respectively. Each message contains a message header, an instruction body and verification information. For example, the instruction body of the pre-order task adjustment message is "modify the value of parameter X", and the instruction body of the post-order task adjustment message is "re-execute the task". The message header contains the target node information, and the verification information is used to ensure the integrity of the message.

[0165] The predecessor task adjustment message is sent to the predecessor task node set {A} through the directed communication network, and the returned predecessor execution status information is received. For example, node A returns the status information "parameter X has been modified". When all predecessor nodes return the status information of task adjustment completion, the successor task adjustment message is sent to the successor node set {C}, and the returned successor execution status information is received. For example, node C returns the status information "task has been re-executed".

[0166] The system receives status monitoring data packets sent by the preceding and succeeding task nodes at a preset period (e.g., once per second). The data packets contain the current status information of the node, such as CPU utilization, memory occupancy, etc. Based on these data, a node status change model is constructed. For example, a time series analysis method can be used to construct a node status change model to predict the future status of the node. At the same time, the risk propagation network is updated and the risk diffusion coefficient is optimized. For example, according to the results predicted by the node status change model, the connection weights between nodes in the risk propagation network are adjusted to more accurately assess the possibility of risk propagation.

[0167] If the node state change model detects a state anomaly, such as a sudden spike in the CPU utilization of a node, the task adjustment process is suspended and the node state is restored to the previous stable state. For example, parameter X of node B is rolled back to the value before the modification. Then, the task adjustment plan is regenerated, for example, tasks are decomposed into smaller granularities or the order of task execution is adjusted.

[0168] In this embodiment, by dynamically adjusting tasks, the task execution process can be optimized according to actual conditions, and unnecessary waiting time can be reduced, thereby improving the overall task execution efficiency. For example, when a task node fails, the task can be dynamically adjusted to allocate subsequent tasks to other available nodes to avoid blocking the entire task chain. By monitoring the node status and handling abnormal situations in a timely manner, the spread of risks can be effectively prevented and the stability and reliability of the system can be improved. By dynamically adjusting tasks, resources can be dynamically allocated according to the actual needs of the task, avoiding waste of resources and improving resource utilization. Figure 2 FIG. 1 is a schematic diagram of a visual project management system based on task tracking according to an embodiment of the present invention. Figure 2 As shown, the system comprises:

[0169] The first unit is used to collect project risk information using situational awareness technology, wherein the project risk information includes market risk data, technical risk data, resource risk data, and management risk data, input the project risk information into a pre-trained risk assessment model for multi-dimensional cross-analysis, generate a risk propagation network and calculate the risk diffusion coefficient, perform hierarchical analysis on project requirements based on the risk diffusion coefficient to obtain a task decomposition plan, construct a task transfer chain based on the task decomposition plan, and generate a visual task relationship diagram, wherein the visual task relationship diagram displays the execution order and correlation degree between task nodes;

[0170] The second unit is used to build an adaptive task execution network based on the task transfer chain, identify the relationship between task nodes in the adaptive task execution network, and establish task constraints including risk thresholds, resource thresholds and time thresholds based on the risk propagation network. The risk transfer probability between task nodes is calculated in combination with the task constraints, and an early warning threshold is set according to the degree of association of the task nodes. The early warning state of the task node is judged based on the comparison result of the risk transfer probability and the early warning threshold, and a task state migration rule and an alternative task path are generated based on the early warning state of the task node. At the same time, a task elastic scheduling mechanism is built and stored in a rule database;

[0171] The third unit is used to collect task execution data and monitor the risk diffusion coefficient. When the task execution data triggers the task constraint condition, the state migration rule and the alternative task path are extracted from the rule database, the state conversion direction of the task node is judged based on the state migration rule, and the optimal path is selected from the alternative task paths to reorganize the task, generate task reconstruction instructions, identify the predecessor task node and the successor task node on the task transfer chain, send the task reconstruction instructions to the predecessor task node and the successor task node to realize dynamic adjustment of the task, update the risk propagation network according to the adjustment result, and optimize the risk diffusion coefficient.

[0172] According to a third aspect of the embodiments of the present invention,

[0173] An electronic device is provided, comprising:

[0174] processor;

[0175] a memory for storing processor-executable instructions;

[0176] The processor is configured to call the instructions stored in the memory to execute the aforementioned method.

[0177] According to a fourth aspect of the embodiments of the present invention,

[0178] A computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the aforementioned method is implemented.

[0179] The present invention may be a method, an apparatus, a system and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for executing various aspects of the present invention.

[0180] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention 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 invention.

Claims

1. A visual project management method based on task tracking, characterized in that: include: Use situational awareness technology to collect project risk information, which includes market risk data, technical risk data, resource risk data, and management risk data. Input the project risk information into a pre-trained risk assessment model for multi-dimensional cross-analysis, generate a risk propagation network, and calculate the risk diffusion coefficient. Based on the risk diffusion coefficient, perform a hierarchical analysis of project requirements to obtain a task decomposition plan. Based on the task decomposition plan, build a task transfer chain and generate a visual task relationship diagram. The visual task relationship diagram displays the execution order and correlation between task nodes. An adaptive task execution network is constructed based on the task transfer chain, the relationship between task nodes is identified in the adaptive task execution network, and task constraints including risk thresholds, resource thresholds and time thresholds are established based on the risk propagation network. The risk transfer probability between task nodes is calculated in combination with the task constraints, and an early warning threshold is set according to the degree of association of the task nodes. The early warning state of the task node is determined based on the comparison result of the risk transfer probability and the early warning threshold, and task state migration rules and alternative task paths are generated based on the early warning state of the task node. At the same time, a task elastic scheduling mechanism is constructed and stored in a rule database; Task execution data is collected and the risk diffusion coefficient is monitored. When the task execution data triggers the task constraint condition, the state migration rule and the alternative task path are extracted from the rule database. The state conversion direction of the task node is determined based on the state migration rule, and the optimal path is selected from the alternative task paths to reorganize the task. The task reconstruction instruction is generated, the predecessor task node and the successor task node on the task transfer chain are identified, and the task reconstruction instruction is sent to the predecessor task node and the successor task node to realize dynamic adjustment of the task. The risk propagation network is updated according to the adjustment result and the risk diffusion coefficient is optimized.

2. The method according to claim 1, characterized in that The project risk information is input into a pre-trained risk assessment model for multi-dimensional cross-analysis, a risk propagation network is generated and a risk diffusion coefficient is calculated. Based on the risk diffusion coefficient, the project requirements are hierarchically analyzed to obtain a task decomposition solution including: Preprocessing the project risk data, extracting feature sequences from time series data through a sliding window, converting category data into numerical data through encoding, and normalizing the preprocessed data to obtain standardized risk data; Constructing a multi-layer attention risk assessment model, the multi-layer attention risk assessment model includes an input layer, an intermediate layer and an output layer, wherein the input layer is provided with parallel long short-term memory units, each long short-term memory unit processes a type of standardized risk data respectively, the intermediate layer adopts a multi-head self-attention mechanism to realize interactive modeling of different risk dimensions to obtain attention scores, and the output layer constructs a risk propagation network based on the attention scores; Setting risk nodes in the risk propagation network, determining edge weights between the risk nodes based on the attention scores, calculating risk diffusion coefficients using a random walk algorithm, constructing a judgment matrix based on the risk diffusion coefficients, and calculating weight vectors using an eigenvector method, combining the weight vectors with a fuzzy relationship matrix to obtain a demand priority vector, and stratifying project requirements based on the judgment matrix and the demand priority vector to obtain a demand hierarchy; Based on the requirement hierarchy, the critical path method is used to decompose tasks, the priority relationship between tasks is determined according to the judgment matrix, and the task criticality is calculated in combination with the earliest task start time, the latest task start time and the risk diffusion coefficient. Under the condition of satisfying resource constraints and risk threshold constraints, the task decomposition plan is obtained by minimizing the sum of task criticalities.

3. The method according to claim 2, characterized in that Based on the requirement hierarchy, the critical path method is used to decompose tasks, the priority relationship between tasks is determined according to the judgment matrix, and the task criticality is calculated in combination with the earliest task start time, the latest task start time and the risk diffusion coefficient. Under the condition of satisfying resource constraints and risk threshold constraints, the task decomposition scheme is obtained by minimizing the sum of task criticalities, including: A task dependency network is constructed based on a judgment matrix, wherein the task dependency network includes dependency relationships between task nodes, and the dependency strength between task nodes is calculated by a dependency function, wherein the dependency function integrates judgment matrix element values, task semantic similarity calculated based on a word vector model, and a temporal correlation coefficient based on time window overlap; Based on the task dependency network, a dynamic programming algorithm is used to calculate the earliest start time and the latest start time of the task node, the earliest start time is recursively calculated based on the predecessor task node set, and the latest start time is reversely recursively calculated based on the successor task node set, and the task dependency network is topologically sorted to determine the task execution priority; A time margin factor is calculated based on the earliest start time and the latest start time, a risk assessment factor is calculated based on the risk diffusion coefficient, a structural importance factor is calculated based on the topological characteristics of the task nodes in the task dependency network, and the time margin factor, the risk assessment factor and the structural importance factor are integrated in an adaptive weight combination manner to obtain the task criticality; Construct a multi-objective task decomposition optimization model, with minimization of task criticality and maximization of resource utilization as optimization goals, and set resource constraints, risk threshold constraints, and task dependency constraints. The resource constraints include dynamic usage constraints on human resources, equipment resources, and material resources. A multi-objective optimization algorithm is used to solve the task decomposition optimization model. The task sequence is represented by real number coding. An adaptive crossover operator and a local search mutation operator are designed to generate a new population. The adaptive crossover operator adjusts the crossover strategy according to the population diversity, and the local search mutation operator performs neighborhood search under dependency constraints. A comprehensive fitness function including task criticality, resource utilization and constraint violation is constructed, and the elite retention strategy and Pareto non-dominated sorting are used to obtain the task decomposition scheme.

4. The method according to claim 1, characterized in that: The warning status of the task node is determined based on the comparison result between the risk transfer probability and the warning threshold. The task status migration rules and alternative task paths are generated based on the warning status of the task node, including: Obtain the risk transfer probability and warning threshold of the task node, compare the risk transfer probability with the lowest threshold and the highest threshold of the warning threshold, divide the warning state of the task node into normal state, warning state and alarm state according to the comparison result, calculate the warning level under the warning state and the alarm level under the alarm state, and record the state feature vector under each warning state respectively, wherein the state feature vector includes risk index, resource occupation index and time progress index; A state transition rule set is constructed based on the warning state, and the state transition rule set includes a migration rule from a normal state to a warning state that is triggered when the risk transfer probability is greater than a minimum threshold and the risk increment is greater than the risk tolerance; a migration rule from a warning state to an alarm state that is triggered when the risk transfer probability is greater than a maximum threshold or the alarm level is greater than a critical alarm level; a recovery rule from an alarm state to an alarm state that is triggered when the risk transfer probability is less than a maximum threshold and the potential impact is less than an impact threshold; a recovery rule from an alarm state to a normal state that is triggered when the risk transfer probability is less than a minimum threshold and the observation time is greater than the observation period; The task nodes are constructed as a node set, the state migration rules are constructed as an edge set, a task migration network is constructed based on the node set and the edge set, the matching degree of state feature data between nodes in the task migration network is calculated to obtain state similarity, the resource consumption of state transition between nodes is calculated to obtain migration cost, and the migration weight between nodes is calculated according to the risk transfer probability, state similarity and migration cost; Multiple candidate migration paths are generated according to the node set, edge set and migration weight. Each candidate migration path includes a node migration sequence and a corresponding state transition sequence. The cumulative value of the risk transfer probability of adjacent nodes in the node migration sequence is calculated to obtain the path risk. The number of state transition sequences that satisfy the state migration rules is calculated to obtain the path feasibility. The candidate migration path with the smallest weighted combination value of path risk and path feasibility is used as the alternative task path.

5. The method according to claim 4, characterized in that Calculating the cumulative value of the risk transfer probability of adjacent nodes in the node migration sequence to obtain the path risk, calculating the number of the state transition sequence that satisfies the state migration rule to obtain the path feasibility, and taking the candidate migration path with the smallest weighted combination value of the path risk and the path feasibility as the candidate task path includes: Based on the node migration sequence, the historical risk event frequency and impact degree of adjacent task nodes are obtained, the risk transfer probability between nodes is calculated, the distance attenuation factor is calculated based on the network distance between the task nodes, and the actual risk transfer value between nodes is obtained by combining the risk transfer probability and the distance attenuation factor; Determine the node importance weight according to the criticality of the task node in the task execution process, input the node importance weight and the actual risk transfer value between nodes into the pre-built risk accumulation model, and calculate the path risk of the candidate migration path; Constructing a state transition compliance evaluation function based on a state transition rule set, inputting the state transition sequence into the state transition compliance evaluation function, and calculating the ratio of the number of state transitions satisfying the state transition rule to the total length of the state sequence to obtain path feasibility; The path risk and path feasibility are weighted and combined using the balance factor to construct a path quality assessment model and calculate the quality assessment index of the candidate migration path. The risk transfer probability and state migration rules are randomly perturbed to establish a Monte Carlo simulation model, the candidate migration path is input into the Monte Carlo simulation model to verify the stability of the quality evaluation index, and the candidate migration path with the smallest quality evaluation index is selected as the alternative task path.

6. The method according to claim 1, characterized in that Extracting state transition rules and alternative task paths from a rule database, determining the state transition direction of a task node based on the state transition rules, and selecting the optimal path from the alternative task paths for task reorganization, generating task reorganization instructions including: Acquire a state transition rule from a rule database, construct a state transition decision tree based on the state transition rule set, calculate a state transition direction through the state transition decision tree, and combine the state transition direction with the current state to form a target state; Constructing a feature extraction model and a utility evaluation model, inputting the target state, task execution data and risk diffusion coefficient into the feature extraction model to extract state features, inputting the candidate path set into the feature extraction model to extract action features, inputting the state features and the action features into the utility evaluation model to fuse them to obtain a state-action representation, and calculating the utility evaluation value of the state-action representation through the utility evaluation model; Calculating the path conversion cost, resource consumption and risk accumulation value of each path in the candidate path set, calculating the path weight value based on the path conversion cost, the resource consumption and the risk accumulation value, and sampling the candidate path set according to the path weight value to obtain a training sample; Input the training samples into the feature extraction model and the utility evaluation model for training, calculate the loss function value and update the model parameters, and use the updated model to recalculate the utility evaluation value of the candidate path set to obtain an updated utility evaluation value; Calculating probability distribution parameters based on the updated utility evaluation value, sampling using the probability distribution parameters to obtain sampling values, and selecting a path with a maximum sampling value as a reconstructed path; A reconstruction instruction is generated according to the reconstruction path, and the reconstruction instruction includes a state migration instruction, a resource allocation instruction and an execution timing instruction.

7. The method according to claim 1, characterized in that Identifying a preceding task node and a succeeding task node on a task transfer chain, and sending the task reconstruction instruction to the preceding task node and the succeeding task node to realize dynamic task adjustment includes: Constructing a directed communication network of a task delivery chain, setting task execution nodes as network nodes, setting task delivery relationships between the network nodes as network connection relationships, and generating an adjacency matrix of the task delivery chain based on the network nodes and the network connection relationships; Using a depth-first search algorithm, obtaining a set of preceding task nodes of the node to be reconstructed based on a transposed matrix of the adjacency matrix, and obtaining a set of succeeding task nodes of the node to be reconstructed based on the adjacency matrix; Performing semantic analysis on the task reconstruction instruction of the node to be reconstructed, obtaining the instruction execution dependency, and decomposing the task reconstruction instruction into a pre-order instruction subset and a post-order instruction subset based on the instruction execution dependency; Encapsulating the preceding instruction subset into a preceding task adjustment message and encapsulating the following instruction subset into a following task adjustment message, wherein the task adjustment message includes a message header, an instruction body and verification information; Sending a predecessor task adjustment message to the predecessor task node set through a directed communication network, and receiving the predecessor execution status information returned by the predecessor task node set; when the predecessor execution status information indicates that the task adjustment is completed, sending a subsequent task adjustment message to the subsequent task node set, and receiving the subsequent execution status information returned by the subsequent task node set; receiving status monitoring data packets sent by the preceding task node set and the succeeding task node set according to a preset period, constructing a node status change model based on the status monitoring data packets, updating the risk propagation network and optimizing the risk diffusion coefficient; When the node state change model detects a state abnormality, the task adjustment process is suspended, the node state is restored to the last stable state, and the task adjustment plan is regenerated.

8. A visual project management system based on task tracking, used to implement the method described in any one of claims 1 to 7, characterized in that: include: The first unit is used to collect project risk information using situational awareness technology, wherein the project risk information includes market risk data, technical risk data, resource risk data, and management risk data, input the project risk information into a pre-trained risk assessment model for multi-dimensional cross-analysis, generate a risk propagation network and calculate the risk diffusion coefficient, perform hierarchical analysis on project requirements based on the risk diffusion coefficient to obtain a task decomposition plan, construct a task transfer chain based on the task decomposition plan, and generate a visual task relationship diagram, wherein the visual task relationship diagram displays the execution order and correlation degree between task nodes; The second unit is used to build an adaptive task execution network based on the task transfer chain, identify the relationship between task nodes in the adaptive task execution network, and establish task constraints including risk thresholds, resource thresholds and time thresholds based on the risk propagation network. The risk transfer probability between task nodes is calculated in combination with the task constraints, and an early warning threshold is set according to the degree of association of the task nodes. The early warning state of the task node is judged based on the comparison result of the risk transfer probability and the early warning threshold, and a task state migration rule and an alternative task path are generated based on the early warning state of the task node. At the same time, a task elastic scheduling mechanism is built and stored in a rule database; The third unit is used to collect task execution data and monitor the risk diffusion coefficient. When the task execution data triggers the task constraint condition, the state migration rule and the alternative task path are extracted from the rule database, the state conversion direction of the task node is judged based on the state migration rule, and the optimal path is selected from the alternative task paths to reorganize the task, generate task reconstruction instructions, identify the predecessor task node and the successor task node on the task transfer chain, send the task reconstruction instructions to the predecessor task node and the successor task node to realize dynamic adjustment of the task, update the risk propagation network according to the adjustment result, and optimize the risk diffusion coefficient.

9. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; The processor is configured to call the instructions stored in the memory to execute the method described in any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that: When the computer program instructions are executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images