Project risk prediction avoidance method and system fusing deep learning

Abstract

The present application belongs to the field of project risk prediction, and particularly relates to a project risk prediction avoidance method and system fusing deep learning, comprising: acquiring project reports and overall personnel configuration data, using a relation extraction decomposition model to extract project decomposition configuration data, secondly, constructing a project implementation node chain according to graph theory, thirdly, obtaining a set of correlated and non-correlated single sub-project from single sub-project implementation planning data in the node chain through a correlation algorithm, and then constructing a configuration risk model based on a particle swarm algorithm-optimized Bayesian algorithm, inputting the above-mentioned sub-project set to obtain single and correlated configuration risk probabilities, and finally constructing a personnel configuration adjustment factor according to the two risk probabilities, configuring the adjustment factor to the node chain connection relationship, and dynamically adjusting the initial configuration of personnel of each single sub-project until all nodes satisfy a preset risk probability threshold, so as to effectively cope with risks caused by personnel configuration in project implementation.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of project risk prediction, and particularly relates to a project risk prediction avoidance method and system fusing deep learning. BACKGROUND

[0002] With the continuous improvement of project scale and complexity, accurate prediction and timely avoidance of project risks have become key factors for project success. A project covers multiple stages from start to end, and each stage is composed of a series of specific steps, which are interrelated and interdependent. In the project implementation process, not only the smooth implementation of each step needs to be ensured, but also the connection between different steps and the reasonable matching of personnel need to be considered. Any problem in any link may cause risks. The existing technology is based on experience to judge the matching degree of personnel and project steps, which has great limitations. Different projects have unique characteristics, and past experience may not be accurately applicable to the current project. In addition, in the joint implementation process of multiple steps of the project, the existing method pays insufficient attention to the collaborative risks between steps. The interface and dependency relationship between different steps are complex, which may cause a chain reaction and lead to risk expansion.

[0003] A kind of innovation and entrepreneurship project auxiliary management system based on artificial intelligence is disclosed in Chinese patent with authorization announcement No.CN117764536B, comprising: login management module, project information input module, data acquisition module, decision information acquisition module, storage module, analysis module;Login management module and project information input module manage user registration, login and project information;Data acquisition module obtains reference data of existing project;Decision information acquisition module obtains policy, technical literature, market information and arranges to obtain auxiliary decision data;The prediction unit of analysis module is based on project reference data, uses BP neural network algorithm to predict the initial cost and return rate of project, and auxiliary unit provides assistance for user decision based on project reference data and auxiliary decision data.

[0004] A kind of full-process management system of power wisdom engineering project is disclosed in Chinese patent with authorization announcement No.CN118735473B, comprising project establishment module, database, progress management module, cost management module, contract management module, procurement management module and report and data display module;Through highly integrated modular design, overall control and management of project are realized.

[0005] The above existing technologies have the following problems: most of the existing technologies predict and handle from the overall risk, which is difficult to solve the problems of project delay or project implementation risk rate increase caused by the lack of personnel and cost configuration adjustment in the implementation process of project internal, therefore, the application provides a project risk prediction avoidance method and system fusing deep learning. SUMMARY

[0006] In view of the deficiencies of the prior art, the present application proposes a project risk prediction avoidance method and system fusing deep learning, which comprises: obtaining project report data and overall personnel configuration data, using a relation extraction decomposition model to extract project decomposition configuration data, then constructing a project implementation node chain according to graph theory, thirdly, obtaining a set of associated and a set of non-associated single sub-projects from the node chain by a correlation algorithm, then constructing a configuration risk model based on a particle swarm algorithm-optimized Bayesian algorithm, inputting the above-mentioned sets of sub-projects to obtain single and associated configuration risk probabilities, and finally constructing a personnel configuration adjustment factor according to the two kinds of risk probabilities and configuring it to the connection relationship between nodes in the node chain to dynamically adjust the initial configuration of personnel for each single sub-project until all nodes meet a preset risk probability threshold, thereby effectively dealing with the risks caused by personnel configuration in project implementation.

[0007] To achieve the above object, the present application provides the following technical scheme:

[0008] The project risk prediction avoidance method fusing deep learning comprises:

[0009] S1, obtaining project report data and corresponding overall personnel and cost configuration data, and extracting entities such as single sub-project implementation planning, planned start time, planned end time and personnel initial configuration of the project as a whole by using a configuration relation extraction decomposition model to obtain project decomposition configuration data;

[0010] S2, constructing a project implementation key node chain by using a graph theory algorithm and a bidirectional traversal algorithm according to the project decomposition configuration data;

[0011] S3, obtaining a set of associated single sub-projects and a set of non-associated single sub-projects by using a correlation algorithm according to the project decomposition configuration data saved by the corresponding nodes of the project implementation node chain;

[0012] S4, constructing a configuration risk model based on a particle swarm algorithm-optimized Bayesian algorithm, inputting the set of associated single sub-projects and the set of non-associated single sub-projects into the configuration risk model to obtain single configuration risk probability and associated configuration risk probability;

[0013] S5, constructing a configuration adjustment factor corresponding to each node according to the single configuration risk probability and the associated configuration risk probability, configuring the configuration adjustment factor corresponding to each node to the connection relationship between two nodes in the project implementation node chain, and dynamically adjusting the initial configuration of personnel and cost for each single sub-project until all nodes meet the single configuration risk probability threshold and the associated configuration risk probability threshold.

[0014] Specifically, the steps of constructing the project implementation key node chain comprise:

[0015] S201, constructing a project implementation node chain input sequence ([X ij ] i,j∈N , Y i , [T is , T ie ] i∈N , v i , c i ) according to project decomposition configuration data, wherein [X ij ] i,j∈N represents a connection relationship set of all sub-projects after project decomposition, X ij represents a connection relationship between the i-th sub-project and the j-th sub-project, N represents the total number of sub-projects after decomposition, Y i represents a single sub-project implementation plan corresponding to the i-th sub-project, T is and T ie represent a planned start time and a planned end time corresponding to the i-th sub-project in turn, v i represents an initial personnel configuration corresponding to the i-th sub-project, and c i represents an initial cost configuration corresponding to the i-th sub-project.

[0016] S202, constructing a corresponding node in the project implementation node chain according to Y i , and constructing a relationship connection between the i-th node and the j-th node in the project implementation node chain according to X ij .

[0017] S203, obtaining a project implementation node chain through a graph theory algorithm according to the constructed node and the constructed connection relationship set, saving a single sub-project implementation plan, a planned start time, a planned end time, an initial personnel configuration, and an initial cost configuration corresponding to the sub-project into a corresponding node, and marking an implementation order of a corresponding node in the project implementation node chain according to a connection relationship order.

[0018] Specifically, the steps of S3 include:

[0019] S301, calculating a similarity degree and a corresponding similarity matrix of nodes corresponding to the i-th sub-project and the j-th sub-project through a similarity algorithm according to a connection relationship corresponding to the sub-project saved in the project implementation node chain, a project implementation content coincidence degree, a project implementation time coincidence degree, and a personnel configuration coincidence degree;

[0020] S302, constructing an association factor of nodes corresponding to the i-th sub-project and the j-th sub-project according to the similarity degree and the corresponding similarity matrix of the nodes;

[0021] S303, input the data corresponding to all nodes and the association factor corresponding to the i-th and j-th nodes into the clustering algorithm, and obtain the association Mahalanobis distance by multiplying the association factor between the corresponding nodes by the Mahalanobis distance between the corresponding two nodes in the clustering algorithm;

[0022] S304, set an association distance threshold value, when the association Mahalanobis distance between the corresponding node and the remaining nodes is at least one less than the association distance threshold value, the corresponding node is a single sub-project corresponding node in the association single sub-project set, when the association Mahalanobis distance between the corresponding node and the remaining nodes is all greater than or equal to the association distance threshold value, the corresponding node is a single sub-project corresponding node in the non-association single sub-project set, and the association Mahalanobis distance between the corresponding two nodes belonging to the association single sub-project set is built into the corresponding relationship connection, to obtain the relevant relationship connection.

[0023] Specifically, the steps of constructing the project implementation key node chain further include:

[0024] S211, according to the association single sub-project set, the non-association single sub-project set and the corresponding implementation sequence mark, the earliest start time and the earliest end time and the latest start time and the latest end time of each node are obtained by a bidirectional traversal algorithm;

[0025] S212, according to the earliest start time and the earliest end time of each node, the float time s i of each node is obtained.

[0026] S213, set s i = 0 corresponding node as the node on the current time key implementation path, and build the corresponding key implementation path into the project implementation node chain to obtain the project implementation key node chain.

[0027] Specifically, the steps of configuring the risk model include:

[0028] S401, according to the personnel initial configuration state, the cost initial configuration state of each node corresponding sub-project, the relevant association between the current node and the remaining nodes, whether the current node is a node on the key implementation path and the evaluation complexity of the corresponding sub-project, a first input sub-sequence of the particle swarm algorithm is constructed, and by using the Bayesian algorithm and the support vector machine, the initial single configuration risk probability of each node and the association configuration risk probability of the corresponding association node in the association single sub-project set are obtained. The configuration adjustment factor fitting the configuration of the corresponding node sub-project risk probability, personnel configuration and cost configuration;

[0029] S402, constructing a second input sub-sequence of the particle swarm algorithm according to the personnel configuration state and the cost configuration state of each node corresponding sub-project, the correlation of the current node and the remaining nodes, whether the current node is a node on the critical implementation path, the evaluation complexity of each sub-project, the initial single configuration risk probability of each node, and the correlation single sub-project corresponding correlation node configuration risk probability;

[0030] S403, constructing a corresponding fitness function and constraint condition according to the project duration length corresponding to all sub-projects of the node on the current implementation path, the proficiency and average completion rate of the configuration personnel to the current sub-project, the configuration adjustment factor fitted by the risk probability and personnel configuration and cost configuration of the corresponding sub-project, the single configuration risk probability and the correlation configuration risk probability of each node, and the key implementation path node importance weighting coefficient.

[0031] Specifically, the step of constructing the configuration risk model further comprises:

[0032] S404, setting a key configuration risk probability threshold and a global risk probability threshold, inputting the key configuration risk probability threshold, the global risk probability threshold, the first input sub-sequence and the second input sub-sequence of the particle swarm algorithm, the fitness function and the constraint condition into the particle swarm algorithm for training, and obtaining the personnel configuration and the cost configuration of the nodes on all critical implementation paths and the nodes on non-critical implementation paths;

[0033] S405, inputting the personnel configuration and the cost configuration of the nodes on all critical implementation paths and the nodes on non-critical implementation paths into the Bayesian algorithm to obtain the single configuration risk probability and the correlation configuration risk probability corresponding to the nodes on the critical implementation path and the single configuration risk probability and the correlation configuration risk probability corresponding to the nodes on the non-critical implementation path.

[0034] Specifically, the step of constructing the configuration risk model further comprises:

[0035] S406, when the personnel configuration and the cost configuration satisfying the key configuration risk probability threshold corresponding to the nodes on all critical implementation paths are searched, continue to search downward, and when the personnel configuration and the cost configuration satisfying the global risk probability threshold corresponding to the nodes on all critical implementation paths and the nodes on non-critical implementation paths are searched, feedback the personnel configuration and the cost configuration satisfying the global risk probability threshold to the personnel and cost configuration contained in each node on the project implementation key node chain for adjustment;

[0036] S407、When only personnel configuration and cost configuration meeting the key configuration risk probability threshold of the node on all key implementation paths can be searched, the corresponding configuration is fed back to the node corresponding to the key implementation path in the project implementation key node chain, and the personnel configuration and cost configuration corresponding to the minimum single configuration risk probability and the associated configuration risk probability are searched for the nodes corresponding to the non-key implementation path, and the personnel configuration and cost configuration corresponding to the minimum single configuration risk probability and the associated configuration risk probability are fed back to the nodes corresponding to the non-key implementation path.

[0037] Specifically, the configuration adjustment factor corresponding to each node includes a single node adjustment factor and an associated adjustment factor;

[0038] The single node adjustment factor is obtained by fitting the corresponding node sub-project historical single configuration risk probability and the corresponding personnel configuration and cost configuration by support vector machine;

[0039] The associated adjustment factor is obtained by fitting the associated relationship between all predecessor nodes and the current node, the single configuration risk probability corresponding to all predecessor nodes and the corresponding personnel and cost configuration, and the single configuration risk probability corresponding to the current node and the corresponding personnel and cost configuration by Bayesian algorithm and support vector machine under the premise that the single configuration risk probability corresponding to the predecessor nodes of the current node meets the key configuration risk probability threshold.

[0040] The project risk prediction avoidance system integrating deep learning includes a data decomposition module, a node chain construction module, and an association analysis module.

[0041] The data decomposition module includes a data acquisition unit and a data decomposition unit.

[0042] The data acquisition unit is configured to acquire project report data and corresponding overall personnel and cost configuration data; and the data decomposition unit is configured to extract sub-project connection relationship, single sub-project implementation plan, planned start time, planned end time, and personnel initial configuration entity of the project overall by a configuration relationship extraction and decomposition model, to obtain project decomposition configuration data.

[0043] The node chain construction module is configured to construct a project implementation key node chain by graph theory algorithm and bidirectional traversal algorithm according to the project decomposition configuration data.

[0044] The association analysis module is configured to obtain an associated single sub-project set and a non-associated single sub-project set by a correlation algorithm according to the project decomposition configuration data in the project implementation node chain.

[0045] Specifically, the risk prediction avoidance system further includes a risk prediction module and an optimization module.

[0046] A risk prediction module is configured to construct a configuration risk model based on a Bayesian algorithm optimized by a particle swarm algorithm, input the set of single sub-projects with relevance and the set of single sub-projects without relevance into the configuration risk model, and obtain single configuration risk probability and associated configuration risk probability;

[0047] An optimization module is configured to construct a configuration adjustment factor corresponding to each node according to the single configuration risk probability and the associated configuration risk probability, configure the configuration adjustment factor corresponding to each node to a connection relationship between two nodes in the project implementation node chain, and dynamically adjust the initial configuration of personnel and cost of each single sub-project until all nodes satisfy the single configuration risk probability threshold and the associated configuration risk probability threshold.

[0048] Compared with the prior art, the present application has the following beneficial effects:

[0049] The present application is aimed at the deficiencies of the prior art, and the configuration relationship extraction decomposition model is used to accurately extract key information such as sub-project connection relationship, implementation plan, time and initial configuration of personnel and cost from project reports and personnel configuration data, and to deeply understand project details. Secondly, the graph theory algorithm and the bidirectional traversal algorithm are used to construct a project implementation key node chain, and the project structure is presented in an intuitive chain structure. This structure not only clearly shows the logical relationship between each sub-project, but also can mine potential paths and key links in the project through bidirectional traversal, which helps to quickly locate potential risk points of key sub-projects, makes the personnel configuration adjustment more targeted, thirdly, the correlation algorithm is used to distinguish between the set of single sub-projects with relevance and the set of single sub-projects without relevance, further refine the understanding of the project sub-structure, and provide the possibility for evaluating the configuration risk of different types of sub-projects, breaking the limitations of the traditional method of generally processing risks, fourthly, the configuration risk model is constructed based on the configuration risk model, and the single and associated configuration risk probability is accurately calculated, and the configuration adjustment factor is constructed according to the risk probability, and the initial configuration of personnel and cost of each sub-project is dynamically adjusted until the risk probability threshold is satisfied, the accuracy of risk prediction is improved, the problems of project delay or risk rate increase caused by the lack of personnel and cost configuration adjustment are effectively solved, and the project is ensured to proceed smoothly. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 A process flow chart of the project risk prediction avoidance method of embodiment 1 of the present application is shown.

[0051] Figure 2 A simplified diagram of the predecessor and successor nodes of the present application is shown.

[0052] Figure 3 A module diagram of the project risk prediction avoidance system of embodiment 2 of the present application is shown. DETAILED DESCRIPTION

[0053] Embodiment 1

[0054] The prior art is mostly from the overall risk for prediction and corresponding processing, it is difficult to target the project inside in the implementation process due to the lack of personnel and cost configuration adjustment caused by the project delay or the project implementation risk rate increase problem, for example, in the outsourcing software platform design process, each sub-platform or subsystem is subcontracted to different team or individual to do, but due to the importance of the corresponding sub-platform or subsystem is different, the initial personnel or cost configuration can not complete some important sub-platform or subsystem on time, increase the risk of project delay;

[0055] Please refer to Figure 1 , the present application provides an embodiment: a project risk prediction avoidance method based on deep learning, applied to IT personnel dispatch outsourcing service, steps include:

[0056] S1, obtain project report data and corresponding overall personnel and cost configuration data, and extract the connection relationship of sub-projects, single sub-project implementation plan, planned start time, planned end time and personnel initial configuration entity of the project overall through the configuration relationship extraction decomposition model, to obtain project decomposition configuration data;

[0057] Further, the project report data in the embodiment includes the specific process steps of the outsourced software platform design;

[0058] S2, according to the project decomposition configuration data, the project implementation key node chain is constructed through the graph theory algorithm and the bidirectional traversal algorithm;

[0059] S3, according to the project decomposition configuration data saved in the corresponding node of the project implementation node chain, the correlation algorithm is used to obtain the associated single sub-project set and the non-associated single sub-project set;

[0060] Further, the steps of constructing the project implementation key node chain in the embodiment include:

[0061] S201, according to the project decomposition configuration data, the project implementation node chain input sequence ([X ij ] i,j∈N , Y i , [T is , T ie ] i∈N , v i , c i ) is constructed, wherein [X ij ] i,j∈N represents the connection relationship set of all sub-projects after decomposition, X ij represents the connection relationship between the i-th sub-project and the j-th sub-project, N represents the total number of sub-projects after decomposition, Y i represents the single sub-project implementation plan corresponding to the i-th sub-project, T is and Tie denotes the planned start time and the planned end time corresponding to the i-th sub-project, v i denotes the initial personnel configuration corresponding to the i-th sub-project, c i denotes the initial cost configuration corresponding to the i-th sub-project;

[0062] S202, according to Y i constructs the corresponding node in the project implementation node chain, and according to X ij constructs the relationship connection between the i-th node and the j-th node in the project implementation node chain;

[0063] S203, according to the constructed node and the constructed connection relationship set, obtaining the project implementation node chain through a graph theory algorithm, and saving the single sub-project implementation plan, the planned start time, the planned end time, the initial personnel configuration and the initial cost configuration corresponding to the sub-project into the corresponding node, and marking the implementation order of the corresponding node in the project implementation node chain according to the connection relationship order.

[0064] Further, the steps of S3 in the embodiment include:

[0065] S301, according to the connection relationship, the project implementation content coincidence degree, the project implementation time coincidence degree and the personnel configuration coincidence degree corresponding to the sub-projects saved in the project implementation node chain, calculating the similarity between the i-th sub-project and the j-th sub-project corresponding node and the corresponding similarity matrix through a similarity algorithm;

[0066] S302, according to the similarity between the i-th sub-project and the j-th sub-project corresponding node and the corresponding similarity matrix, constructing the association factor of the i-th sub-project and the j-th sub-project corresponding node;

[0067] S303, inputting all the data saved by the corresponding nodes and the association factor corresponding to the i-th and j-th nodes into a clustering algorithm, and multiplying the association factor between the corresponding nodes by the Mahalanobis distance between the corresponding two nodes in the clustering algorithm to obtain the associated Mahalanobis distance;

[0068] S304, setting an association distance threshold, when the associated Mahalanobis distance between the corresponding node and the remaining nodes is at least one less than the association distance threshold, the corresponding node is the sub-project corresponding node in the association single sub-project set, when the associated Mahalanobis distance between the corresponding node and the remaining nodes is all greater than or equal to the association distance threshold, the corresponding node is the sub-project corresponding node in the non-association single sub-project set, and the associated Mahalanobis distance between the corresponding two nodes belonging to the association single sub-project set is built into the corresponding relationship connection to obtain the relevant relationship connection.

[0069] Further, the steps of constructing the project implementation key node chain in the embodiment further include:

[0070] S211, obtaining the earliest start time and the earliest end time and the latest start time and the latest end time of each node according to the relevance single sub-project set, the non-relevance single sub-project set and the corresponding implementation sequence mark through a bidirectional traversal algorithm;

[0071] S212, obtaining the float time s of each node according to the earliest start time and the earliest end time of each node i ;

[0072] Further, the float time s i of the embodiment includes:

[0073] S2121, obtaining the earliest start time and the earliest end time of each node through a forward traversal sub-algorithm in the bidirectional traversal algorithm, specifically:

[0074]

[0075] Wherein ES(Y i ) represents the earliest start time corresponding to the i-th node, EF(Y i ) represents the earliest end time corresponding to the i-th node, pre(i) represents a set of all predecessor nodes of the i-th node, that is, a set composed of nodes whose implementation sequence is in front of the i-th node and which have an association relationship with the i-th node; ES(Y k ) represents the earliest start time corresponding to the k-th node in front of the i-th node, d ki represents the time duration from the k-th node to the i-th node, EF(Y i ) represents the earliest end time corresponding to the i-th node, d i represents the sub-project plan duration length corresponding to the i-th node, that is, obtained by the plan start time and the plan end time;

[0076] S2122, obtaining the latest start time and the latest end time of each node through a backward traversal sub-algorithm in the bidirectional traversal algorithm, specifically:

[0077]

[0078] Wherein LS(Y i ) represents the latest start time corresponding to the i-th node, LF(Y i ) represents the latest end time corresponding to the i-th node, succ(i) represents a set of all successor nodes of the i-th node, that is, a set composed of nodes whose implementation sequence is behind the i-th node and which have an association relationship with the i-th node; LF(Y h ) represents the latest end time corresponding to the h-th node behind the i-th node, dhi denotes the time duration from the i-th node to the h-th node; for predecessors and successors, see Figure 2 where 1 denotes the node corresponding to the initial sub-project, a, b, c correspond to the a-th, b-th, and c-th predecessor nodes associated with the i-th node, and similarly, u, v correspond to the successor nodes associated with the i-th node. This diagram is a simplified diagram for reference only.

[0079] S2123, obtain the float time s according to the following formula i :

[0080] s i = LS(Y i ) - ES(Y i ) = LF(Y i ) - EF(Y i ).

[0081] S213, set s i = 0 corresponds to the node on the critical implementation path at the current time, and embed the corresponding critical implementation path into the project implementation node chain to obtain the project implementation key node chain.

[0082] This process constructs a precise project implementation key node chain through systematic project decomposition, graph theory algorithm, and bidirectional traversal technology, significantly improving the efficiency and accuracy of project management. Specifically, first, detailed sub-project connection relationships, implementation plans, time arrangements, and resource configurations are obtained through entity extraction of project report data, ensuring the accuracy and completeness of the basic data. Second, graph theory algorithm is used to construct the node chain, and the earliest and latest times of each node are calculated through bidirectional traversal, effectively identifying the critical implementation path and float time, providing a scientific basis for resource optimization. Third, similarity algorithm and clustering analysis are introduced to distinguish between associated and non-associated single sub-project sets based on correlation factors and Mahalanobis distance, enhancing the understanding of the complex relationships within the project. Finally, the critical implementation path is embedded into the project implementation node chain, realizing visual project progress tracking, which facilitates real-time monitoring and decision-making by management. This method not only accurately plans project time but also optimizes resource allocation, strengthens risk management, and promotes team collaboration, providing a solid guarantee for the successful implementation of the project.

[0083] S4, construct a configuration risk model based on the particle swarm algorithm optimized Bayesian algorithm, and input the associated and non-associated single sub-project sets into the configuration risk model to obtain the single configuration risk probability and the associated configuration risk probability;

[0084] S5, according to the single configuration risk probability and the associated configuration risk probability, the configuration adjustment factor corresponding to each node is constructed, and the configuration adjustment factor corresponding to each node is configured to the connection relationship between two nodes in the project implementation node chain, the initial configuration of personnel and cost of each single subproject is dynamically adjusted until all nodes meet the single configuration risk probability threshold and the associated configuration risk probability threshold.

[0085] Further, the configuration adjustment factor corresponding to each node in the embodiment includes a single node adjustment factor and an associated adjustment factor;

[0086] The single node adjustment factor is obtained by fitting the corresponding node subproject historical single configuration risk probability and the corresponding subproject personnel configuration and cost configuration by support vector machine;

[0087] The associated adjustment factor is obtained by fitting the associated relationship between all predecessor nodes and the current node, the single configuration risk probability corresponding to all predecessor nodes and the corresponding personnel and cost configuration, and the single configuration risk probability corresponding to the current node and the corresponding personnel and cost configuration by Bayesian algorithm and support vector machine, on the premise that the single configuration risk probability of the predecessor node pointing to the current node meets the key configuration risk probability threshold.

[0088] Further, the steps of constructing the configuration risk model in the embodiment include:

[0089] S401, according to the initial personnel configuration state, the initial cost configuration state, the relevant association between the current node and the remaining nodes, whether the current node is a node on the key implementation path and the evaluation complexity of the corresponding subproject of each node corresponding to the subproject in history, the first input subsequence of particle swarm algorithm is constructed, and the initial single configuration risk probability of each node and the associated single subproject corresponding to the associated configuration risk probability of the node are obtained by Bayesian algorithm and support vector machine, and the configuration adjustment factor fitted by the personnel configuration, cost configuration of the corresponding node subproject risk probability and the corresponding node subproject risk probability;

[0090] Further, in the embodiment, the initial personnel configuration state of each node corresponding to the subproject in the historical data includes detailed information such as the number of personnel, skill level distribution, which is expressed in the form of vector in a quantitative way; The initial cost configuration state is collected, such as budget allocation, cost composition, etc., which is converted into a numerical vector; The evaluation complexity of the corresponding subproject is evaluated, for example, according to the technical difficulty of the project task, the breadth of the involved field and other factors, which is expressed in the form of complexity score;

[0091] S402, constructing a second input sub-sequence of the particle swarm algorithm according to a personnel configuration state of each node corresponding sub-project, a cost configuration state, a correlation between the current node and the remaining nodes, whether the current node is a node on the critical implementation path, an evaluation complexity of each sub-project, an initial single configuration risk probability of each node, and an associated configuration risk probability of the corresponding associated node in the associated single sub-project set;

[0092] S403, constructing a corresponding fitness function and constraint condition according to a project duration length corresponding to all sub-projects of the node on the current implementation path, a proficiency and average completion rate of the configuration personnel on the current sub-project, a configuration adjustment factor fitted according to the risk probability of the corresponding sub-project and the personnel configuration and the cost configuration, a single configuration risk probability of each node, an associated configuration risk probability, and a key implementation path node importance weighting coefficient;

[0093] Further, in the present embodiment, the proficiency of the configuration personnel on the current sub-project is obtained by evaluating the proficiency of the configuration personnel on the current sub-project, for example, by quantitatively evaluating the past project experience and skill mastery of the personnel; secondly, the average completion rate of the corresponding configuration personnel is obtained according to the average rate of the configuration personnel in the same type of sub-project processing.

[0094] S404, setting a key configuration risk probability threshold and a global risk probability threshold, inputting the key configuration risk probability threshold, the global risk probability threshold, the first input sub-sequence and the second input sub-sequence of the particle swarm algorithm, the fitness function and the constraint condition into the particle swarm algorithm for training, and obtaining the personnel configuration and the cost configuration of all nodes on the critical implementation path and the non-critical implementation path;

[0095] S405, inputting the personnel configuration and the cost configuration of all nodes on the critical implementation path and the non-critical implementation path into the Bayesian algorithm to obtain the single configuration risk probability and the associated configuration risk probability corresponding to the nodes on the critical implementation path and the single configuration risk probability and the associated configuration risk probability corresponding to the nodes on the non-critical implementation path;

[0096] S406, when the personnel configuration and the cost configuration satisfying the key configuration risk probability threshold of all nodes on the critical implementation path are searched, continue to search downward, when the personnel configuration and the cost configuration satisfying the global risk probability threshold of all nodes on the critical implementation path and the non-critical implementation path are searched, feedback the personnel configuration and the cost configuration satisfying the global risk probability threshold to the personnel and cost configuration adjustment of each node included in the project implementation key node chain;

[0097] Further, the personnel and cost adjustment configuration in the embodiment is based on the original overall personnel and cost and the initial configuration of the personnel and corresponding cost of each sub-project, and the members and cost are re-divided and configured according to the corresponding nodes of the key implementation path and the corresponding nodes of the non-key implementation path, on the basis of which the implementation risk of the corresponding sub-project of the corresponding node of the key implementation path is minimized, the key configuration risk probability threshold is met, and all sub-projects corresponding to the personnel configuration and cost configuration at the current time are searched to meet the global risk probability threshold.

[0098] S407, when only the personnel configuration and cost configuration meeting the key configuration risk probability threshold of the node on all key implementation paths can be searched, the corresponding configuration is fed back to the node corresponding to the key implementation path in the key node chain of the project implementation, and the personnel configuration and cost configuration corresponding to the minimum single configuration risk probability and the associated configuration risk probability are searched for the corresponding node of the non-key implementation path, and the personnel configuration and cost configuration corresponding to the minimum single configuration risk probability and the associated configuration risk probability are fed back to the corresponding node of the non-key implementation path.

[0099] The process builds a particle swarm algorithm input sequence through historical data, and combines a Bayesian algorithm and a support vector machine (SVM), so that the single configuration risk probability and the associated configuration risk probability of each node can be accurately calculated. This statistical learning-based method not only considers the state of the current node, but also comprehensively considers the correlation of the node with other nodes, so that the risk assessment is more comprehensive and accurate. Secondly, according to the calculated risk probability, a configuration adjustment factor corresponding to each node is constructed and applied to the connection relationship between two nodes in the project implementation node chain, so that dynamic adjustment of the initial configuration of personnel and cost is realized. This method ensures that the nodes on the key implementation path can obtain necessary resource support in priority, and optimizes the resource configuration on the non-key implementation path without affecting the overall progress, thereby improving the overall utilization rate of resources. Thirdly, by introducing multi-dimensional factors such as the node importance weighting coefficient of the key implementation path, the proficiency of the configuration personnel and the average completion rate, a scientific and reasonable fitness function and constraint condition are constructed. This enables the particle swarm algorithm to maximize the reduction of risk probability while meeting various constraints, so as to find the optimal or approximate optimal resource configuration scheme. Fourthly, the key configuration risk probability threshold and the global risk probability threshold are set to ensure that the nodes on the key implementation path first reach a low risk level, and then the global risk of the entire project is gradually optimized. This hierarchical risk control strategy effectively reduces the possibility of project failure and enhances the robustness and reliability of the project. Finally, when a configuration that meets the key configuration risk probability threshold is searched, the search is continued to meet the global risk probability threshold; if only the key implementation path threshold can be met, the non-key implementation path is adjusted to minimize the risk. This process forms a closed-loop feedback system that iteratively optimizes resource configuration until the ideal scheme is found. In addition, the optimized configuration is fed back to the key node chain in the project implementation in real time, so as to ensure the consistency of project progress and plan.

[0100] Embodiment 2

[0101] Please refer to Figure 3 Another embodiment provided by the present application is a project risk prediction avoidance system based on deep learning, which comprises a data decomposition module, a node chain construction module, an association analysis module, a risk prediction module and an optimization module.

[0102] The data decomposition module is used for decomposition of the overall project. The data decomposition module comprises a data acquisition unit and a data decomposition unit.

[0103] The data acquisition unit is used for acquiring project report data and corresponding overall personnel and cost configuration data. The data decomposition unit is used for extracting a project decomposition configuration data by using a configuration relationship extraction decomposition model to extract a connection relationship between sub-projects, a single sub-project implementation plan, a planned start time, a planned end time and a personnel initial configuration entity of the overall project.

[0104] a node chain construction module configured to construct a project implementation key node chain according to project breakdown configuration data through a graph theory algorithm and a bidirectional traversal algorithm;

[0105] a correlation analysis module configured to obtain a correlation single subproject set and a non-correlation single subproject set through a correlation algorithm according to the project breakdown configuration data in the project implementation node chain;

[0106] a risk prediction module configured to construct a configuration risk model based on a Bayesian algorithm optimized by a particle swarm algorithm, and input the correlation single subproject set and the non-correlation single subproject set into the configuration risk model to obtain a single configuration risk probability and a correlation configuration risk probability;

[0107] an optimization module configured to construct a configuration adjustment factor corresponding to each node according to the single configuration risk probability and the correlation configuration risk probability, and configure the configuration adjustment factor corresponding to each node to a connection relationship between two nodes in the project implementation node chain to dynamically adjust an initial configuration of personnel and cost of each single subproject until all nodes satisfy a single configuration risk probability threshold and a correlation configuration risk probability threshold.

[0108] The embodiments of the present application are described above with reference to the drawings; however, the present application is not limited to the specific embodiments described above, which are merely illustrative rather than restrictive, and any person of ordinary skill in the art can make changes, modifications, replacements and variations to the above-described embodiments without departing from the purpose of the present application and the scope protected by the claims, and these are all within the protection scope of the present application.

[0109] If the technical solution of the present disclosure involves personal information, the product applying the technical solution of the present disclosure has explicitly informed the personal information processing rules before processing the personal information and obtained the personal independent consent. If the technical solution of the present disclosure involves sensitive personal information, the product applying the technical solution of the present disclosure has obtained the personal independent consent before processing the sensitive personal information and at the same time meets the requirement of "explicit consent". For example, at the personal information collection device such as camera, a clear and prominent sign is set to inform that the personal information collection range has been entered and the personal information will be collected, and if the individual voluntarily enters the collection range, it is deemed to agree to collect the personal information; or on the device for processing personal information, the personal information processing rules are informed through obvious signs / information, and the personal authorization is obtained through pop-up information or asking the individual to upload the personal information by himself / herself; wherein, the personal information processing rules can include personal information processor, personal information processing purpose, processing method and personal information type, etc.

Claims

1. A project risk prediction and avoidance method integrating deep learning, characterized in that, include: S1. Obtain project report data and corresponding overall personnel and cost configuration data. Use the configuration relationship extraction and decomposition model to perform initial configuration entity extraction on the overall project and obtain project decomposition configuration data. S2. Based on the project decomposition and configuration data, knowledge is extracted using graph theory algorithms and bidirectional traversal algorithms to construct a chain of key nodes for project implementation. S3. Based on the project decomposition configuration data stored in the corresponding nodes of the project implementation node chain, obtain the related single sub-project set and the non-related single sub-project set through the correlation algorithm. S4. Construct a configuration risk model using a Bayesian algorithm optimized by particle swarm optimization, and input related single sub-item sets and unrelated single sub-item sets into the configuration risk model to obtain the probability of single configuration risk and the probability of related configuration risk. S5. Construct configuration adjustment factors for each node based on the single configuration risk probability and the associated configuration risk probability, and configure the configuration adjustment factors for each node to the connection relationship between each pair of nodes in the project implementation node chain. Dynamically adjust the initial configuration of personnel and costs for each individual sub-project until all nodes meet the single configuration risk probability threshold and the associated configuration risk probability threshold.

2. The project risk prediction and avoidance method integrating deep learning as described in claim 1, characterized in that, The project decomposition configuration data includes sub-project connection relationships, single sub-project implementation plans, planned start times, planned end times, and personnel; the steps for constructing the project implementation key node chain include: S201. Construct the project implementation node chain input sequence ([X) based on the project decomposition and configuration data. ij ] i,j∈N Y i [T] is T ie ] i∈N v i c i ), where [X ij ] i,j∈N X represents the set of connections between all sub-projects after project decomposition. ij Y represents the connection between the i-th sub-item and the j-th sub-item, N represents the total number of sub-items after decomposition, and Y represents the connection between the i-th sub-item and the j-th sub-item. i T represents the implementation plan for a single sub-project corresponding to the i-th sub-project. is and T ie v represents the planned start time and planned end time corresponding to the i-th sub-project, respectively. i c represents the initial personnel configuration corresponding to the i-th sub-project. i This represents the initial cost configuration corresponding to the i-th sub-item; S202, According to Y i Construct the corresponding nodes in the project implementation node chain, and based on X ij Construct a connection between the i-th node and the j-th node in the project implementation node chain; S203. Based on the constructed nodes and the constructed set of connections, obtain the project implementation node chain through graph theory algorithm, and save the single sub-project implementation plan, planned start time, planned end time, initial personnel configuration and initial cost configuration of the corresponding sub-project into the corresponding node, and mark the implementation order of the corresponding nodes in the project implementation node chain according to the connection order.

3. The project risk prediction and avoidance method integrating deep learning as described in claim 2, characterized in that, The steps in S3 include: S301. Based on the connection relationship, overlap of project implementation content, overlap of project implementation time, and overlap of personnel configuration stored in the project implementation node chain, the similarity between the nodes corresponding to the i-th sub-project and the j-th sub-project and the corresponding similarity matrix are calculated using a similarity algorithm. S302. Based on the similarity between the corresponding nodes of the i-th sub-item and the j-th sub-item and the corresponding similarity matrix, construct the association factor between the corresponding nodes of the i-th sub-item and the j-th sub-item. S303. Input the data stored for all nodes and the association factors between the i-th and j-th nodes into the clustering algorithm, and multiply the association factors between the corresponding nodes by the Mahalanobis distance between the corresponding two nodes in the clustering algorithm to obtain the association Mahalanobis distance. S304. Set the association distance threshold. When at least one of the association Mahalanobis distances between the corresponding node and the remaining nodes is less than the association distance threshold, the corresponding node is the corresponding node of the sub-item in the associated single sub-item set. When the association Mahalanobis distances between the corresponding node and the remaining nodes are both greater than or equal to the association distance threshold, the corresponding node is the corresponding node of the sub-item in the non-associated single sub-item set. The association Mahalanobis distances between the two corresponding nodes in the associated single sub-item set are then embedded into the corresponding relational connection to obtain the related relational connection.

4. The project risk prediction and avoidance method integrating deep learning as described in claim 3, characterized in that, The steps for constructing the critical node chain in the project implementation also include: S211. Based on the related single sub-items set, the non-related single sub-items set and the corresponding implementation order mark, obtain the earliest start time, the earliest end time and the latest start time and the latest end time of each node through a bidirectional traversal algorithm. S212. Based on the earliest start time and earliest end time of each node, obtain the floating time s of each node. i ; S213, Set s i The node corresponding to 0 is a node on the critical implementation path at the current moment, and the corresponding critical implementation path is built into the project implementation node chain to obtain the project implementation critical node chain.

5. The project risk prediction and avoidance method integrating deep learning as described in claim 4, characterized in that, The steps for constructing the configuration risk model include: S401. Based on the initial personnel configuration status, initial cost configuration status, correlation between the current node and the remaining nodes, whether the current node is a node on the critical implementation path, and the evaluation complexity of the corresponding sub-project for each historical node, construct the first input sub-sequence of the particle swarm algorithm. Through Bayesian algorithm and support vector machine, obtain the initial single configuration risk probability and correlation of each node, the correlation configuration risk probability of the corresponding related nodes in the single sub-project set, and the configuration adjustment factor that fits the personnel configuration and cost configuration. S402. Construct the second input subsequence of the particle swarm algorithm based on the personnel configuration status, cost configuration status, correlation between the current node and the remaining nodes, whether the current node is a node on the critical implementation path, the evaluation complexity of each sub-project, the initial single configuration risk probability of each node, and the correlation configuration risk probability of the associated nodes in the single sub-project set. S403. Based on the project duration of all sub-projects corresponding to the nodes on the current implementation path, the proficiency and average completion rate of the configured personnel in the current sub-project, the risk probability of the corresponding sub-project and the personnel configuration, the configuration adjustment factor fitted to the cost configuration, the single configuration risk probability and associated configuration risk probability of each node and the weighted coefficient of the importance of key implementation path nodes, construct the corresponding fitness function and constraints.

6. The project risk prediction and avoidance method integrating deep learning as described in claim 5, characterized in that, The steps for constructing the configuration risk model also include: S404. Set the critical configuration risk probability threshold and the global risk probability threshold. Input the critical configuration risk probability threshold, the global risk probability threshold, the first and second input subsequences of the particle swarm algorithm, the fitness function and the constraints into the particle swarm algorithm for training to obtain the personnel configuration and cost configuration of nodes on all critical implementation paths and nodes on non-critical implementation paths. S405. Input the personnel and cost configurations of all nodes on the critical implementation path and nodes on the non-critical implementation path into the Bayesian algorithm to obtain the single configuration risk probability and associated configuration risk probability of nodes on the critical implementation path and nodes on the non-critical implementation path.

7. The project risk prediction and avoidance method integrating deep learning as described in claim 6, characterized in that, The steps for constructing the configuration risk model also include: S406. When the search finds personnel and cost configurations that satisfy the risk probability thresholds of the key configurations corresponding to the nodes on all key implementation paths, continue searching downwards. When the search finds personnel and cost configurations that satisfy the global risk probability thresholds of the nodes corresponding to all key implementation paths and the nodes corresponding to non-key implementation paths, feed back the personnel and cost configurations that satisfy the global risk probability thresholds to the personnel and cost configurations contained in each node on the key node chain of project implementation for adjustment. S407. When only personnel and cost configurations that meet the risk probability thresholds of critical configurations corresponding to nodes on all critical implementation paths can be found, the corresponding configurations are fed back to the nodes corresponding to critical implementation paths within the project implementation critical node chain. At the same time, for nodes corresponding to non-critical implementation paths, the personnel and cost configurations corresponding to minimizing the single configuration risk probability and associated configuration risk probability are searched, and the personnel and cost configurations corresponding to minimizing the single configuration risk probability and associated configuration risk probability are fed back to the nodes corresponding to non-critical implementation paths.

8. The project risk prediction and avoidance method integrating deep learning as described in claim 7, characterized in that, The configuration adjustment factor corresponding to each node includes a single node adjustment factor and an associated adjustment factor. The single-node adjustment factor is obtained by fitting a support vector machine with the historical single configuration risk probability of the corresponding node sub-project and the personnel and cost configuration of the corresponding sub-project. The correlation adjustment factor is obtained by fitting the single configuration risk probability of all predecessor nodes to the current node, the single configuration risk probability of all predecessor nodes and the current node, the corresponding personnel and cost configurations, and the single configuration risk probability and corresponding personnel and cost configurations of the current node, based on the correlation relationship between all predecessor nodes and the current node, under the premise that the single configuration risk probability of all predecessor nodes and the current node meets the critical configuration risk probability threshold.

9. A project risk prediction and avoidance system integrating deep learning, used to implement the project risk prediction and avoidance method integrating deep learning as described in any one of claims 1-8, characterized in that, include: Data decomposition module, node chain construction module, and correlation analysis module; The data decomposition module includes a data acquisition unit and a data decomposition unit; The data acquisition unit is used to acquire project report data and corresponding overall personnel and cost configuration data; the data decomposition unit is used to extract sub-project connection relationships, single sub-project implementation plans, planned start time, planned end time and initial personnel configuration entities from the overall project through a configuration relationship extraction decomposition model to obtain project decomposition configuration data. The node chain construction module is used to construct a key node chain for project implementation based on the project decomposition configuration data, using graph theory algorithms and bidirectional traversal algorithms. The correlation analysis module is used to obtain correlated single sub-project sets and non-correlated single sub-project sets based on the project decomposition configuration data in the project implementation node chain through a correlation algorithm.

10. The project risk prediction and avoidance system integrating deep learning as described in claim 9, characterized in that, The risk prediction and avoidance system also includes a risk prediction module and an optimization module; The risk prediction module is used to construct a configuration risk model based on a Bayesian algorithm optimized by particle swarm optimization, and input related single sub-item sets and unrelated single sub-item sets into the configuration risk model to obtain the probability of single configuration risk and the probability of related configuration risk. The optimization module is used to construct a configuration adjustment factor for each node based on the single configuration risk probability and the associated configuration risk probability, and to configure the configuration adjustment factor for each node to the connection relationship between each pair of nodes in the project implementation node chain, so as to dynamically adjust the initial configuration of personnel and cost for each single sub-project until all nodes meet the single configuration risk probability threshold and the associated configuration risk probability threshold.

Images