Engineering project risk management method and system

Abstract

The invention relates to the technical field of project management, and discloses an engineering project risk management method and system, and the method comprises the steps: carrying out the feature extraction and dimension division of engineering project monitoring data, and obtaining a risk feature data set; performing cross influence analysis of quality risk-cost risk and progress risk-resource risk on the risk feature data set by adopting a bidirectional correlation calculation mechanism to obtain a risk correlation intensity matrix; based on the risk association intensity matrix and the historical risk evolution data, comprehensively evaluating the risk occurrence probability, the influence degree and the duration to obtain a risk level evaluation vector; performing hierarchical disassembly on the risk level evaluation vector to generate a risk decomposition hierarchical tree; performing multi-level risk classification on the risk decomposition hierarchical tree, and constructing a multi-level risk classification matrix; and establishing a risk response decision table according to the multi-level risk classification matrix, and generating a risk management control strategy, thereby improving the accuracy of engineering project risk management.

Description

Technical Field

[0001] The present application relates to the technical field of project management, and in particular to a method and system for risk management of engineering projects. Background Art

[0002] Traditional engineering project risk management methods mainly rely on experience judgment and single-dimensional risk assessment, which makes it difficult to accurately grasp the mutual influence between multiple dimensions such as quality risk, cost risk, schedule risk and resource risk.

[0003] The current risk assessment methods do not adequately analyze the interactions between risk factors, making it difficult to effectively identify and quantify the strength of the correlation between risk factors. In addition, the cross-domain risk factors that influence each other easily introduce interference information, resulting in deviations in the assessed risk level, affecting the pertinence and effectiveness of risk prevention and control measures. In actual engineering project management, due to the lack of a systematic risk classification and grading method, it is impossible to accurately identify and dynamically manage risks at different stages and levels. Summary of the invention

[0004] The present application provides a method and system for engineering project risk management, thereby improving the accuracy of engineering project risk management.

[0005] The first aspect of the present application provides a method for risk management of an engineering project, the method comprising: Extract features and divide dimensions of engineering project monitoring data to obtain risk feature data sets; A bidirectional correlation calculation mechanism is used to perform a cross-impact analysis of quality risk-cost risk and schedule risk-resource risk on the risk feature data set to obtain a risk correlation intensity matrix; Based on the risk correlation intensity matrix and historical risk evolution data, a comprehensive assessment is performed on the probability of risk occurrence, degree of impact and duration to obtain a risk level assessment vector; Decomposing the risk level assessment vector hierarchically to generate a risk decomposition hierarchy tree; Performing multi-level risk classification on the risk decomposition hierarchy tree to construct a multi-level risk classification matrix; A risk response decision table is established based on the multi-level risk classification matrix, and a risk management control strategy is generated.

[0006] A second aspect of the present application provides a project risk management system, the project risk management system comprising: The feature extraction module is used to extract features and divide dimensions of the project monitoring data to obtain a risk feature data set; A cross-impact analysis module, used to perform a cross-impact analysis of quality risk-cost risk and schedule risk-resource risk on the risk feature data set using a bidirectional correlation calculation mechanism to obtain a risk correlation intensity matrix; A comprehensive assessment module, for comprehensively assessing the probability of risk occurrence, degree of impact and duration based on the risk correlation strength matrix and historical risk evolution data, to obtain a risk level assessment vector; A hierarchical decomposition module, used for performing hierarchical decomposition on the risk level assessment vector to generate a risk decomposition hierarchical tree; A multi-level risk classification module is used to perform multi-level risk classification on the risk decomposition hierarchy tree and construct a multi-level risk classification matrix; The generation module is used to establish a risk response decision table according to the multi-level risk classification matrix and generate a risk management control strategy.

[0007] The third aspect of the present application provides an electronic device, comprising: a memory and at least one processor, wherein the memory stores instructions; the at least one processor calls the instructions in the memory so that the electronic device executes the above-mentioned engineering project risk management method.

[0008] A fourth aspect of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the above-mentioned engineering project risk management method.

[0009] Compared with the prior art, the present application has the following beneficial effects: through multi-dimensional risk feature extraction and bidirectional correlation analysis, the cross-impact calculation between quality risk-cost risk and schedule risk-resource risk is realized, and the correlation strength between different risk dimensions is accurately quantified. The hierarchical risk decomposition and linear regression model are used to accurately calculate the cumulative effect of risk, establish a complete risk decomposition hierarchy tree, and reveal the transmission law of risk in different project stages. Based on the multi-level risk classification model, the feature similarity calculation and weight allocation of risk nodes are performed, and a multi-level risk classification matrix is ​​constructed to make the risk classification more accurate and reasonable. Through the calculation of the dynamic risk indicator matrix and the real-time risk state sequence, a complete risk response decision-making mechanism is established, so that risk prevention and control measures can be dynamically adjusted according to actual conditions. Through the application of association rule mining and decision tree model, a systematic risk management control strategy is formed, which enhances the scientificity and operability of risk prevention and control. The prevention and control strategy is optimized by fuzzy reasoning operation, and a risk management system with a dual index structure is established, making the risk management process more systematic and standardized. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0011] The structures, proportions, sizes, etc. illustrated in the drawings of this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0012] Figure 1 It is a flowchart of a method for project risk management provided by an embodiment of the present invention; Figure 2 is a schematic block diagram of the structure of an engineering project risk management system provided by an embodiment of the present invention; Figure 3 It is a schematic block diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0014] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0015] It should also be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0016] It should be further understood that the term "and / or" used in the specification and appended claims of this application refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations. Figure 1 , an embodiment of the project risk management method in the embodiment of the present application includes: Step 100: extract features and divide dimensions of the project monitoring data to obtain a risk feature data set; It is understandable that the execution subject of the present application may be a project risk management system, or a terminal or a server, which is not specifically limited here. The present application embodiment is described by taking a server as the execution subject as an example.

[0017] Specifically, the project monitoring data is cleaned and standardized to obtain a standardized monitoring data matrix. Through data cleaning, outliers, missing values ​​or other data noise that affect subsequent analysis in the monitoring data are removed to obtain a relatively complete and reliable monitoring data set. Standardization calculation normalizes or standardizes data of different dimensions so that each data dimension has the same scale, eliminates the deviation caused by dimensional differences, and obtains a standardized monitoring data matrix. The standardized monitoring data matrix is ​​numerically decomposed and attribute mapped to obtain a basic characteristic attribute matrix. It is divided according to the four basic attributes of risk probability, impact, duration and correlation. The selection of attributes is derived from the core concerns in risk management, that is, the probability of occurrence of each risk, the impact it brings, the duration of the impact, and the degree of correlation with other risks. These attributes together constitute the basic description of the risk. Through numerical decomposition and attribute mapping, the data in the standardized monitoring data matrix are mapped to a specific attribute space to obtain a basic characteristic attribute matrix, which contains the basic attribute description of each risk feature. The basic feature attribute matrix is ​​mapped and transformed nonlinearly to fully explore the nonlinear relationship in the data, making the risk characteristics more diverse and revealing the complex risk patterns hidden in the original monitoring data. The initial risk feature vector is obtained through nonlinear transformation, which effectively reflects the distribution state of the risk characteristics in the multidimensional space. The linear correlation between the features is evaluated by calculating the covariance matrix of the initial risk feature vector, and the principal component weight coefficient matrix is ​​obtained through eigenvalue decomposition. Based on the principal component weight coefficient matrix, the risk feature data is linearly combined and reconstructed to generate a risk dimension sequence. The linear combination reconstructs the different dimensions of the risk characteristics through the weighted summation of the principal components. The contribution threshold is set based on the risk dimension sequence, and different risk features are classified according to their contribution. Contribution is an indicator to measure the importance of each feature in describing the overall risk data. By setting a contribution threshold, the features that contribute more to the overall risk description are classified into the core feature set, while the features that contribute less are classified into the auxiliary feature set. This classification method helps to simplify the risk management model, retaining only the features that are important for risk assessment and discarding the minor ones, reducing the complexity of the model and improving the accuracy and response speed of the risk management method. Through the above steps, a risk feature dataset is finally obtained.

[0018] Step 200: Use a bidirectional correlation calculation mechanism to perform a cross-impact analysis of quality risk-cost risk and schedule risk-resource risk on the risk feature data set to obtain a risk correlation intensity matrix; Specifically, the risk feature data set is bidirectionally grouped. The data of quality risk and cost risk are constructed as the first association group, and the data of schedule risk and resource risk are constructed as the second association group, so as to obtain bidirectionally associated grouped data. Each group of risk data is analyzed using Pearson correlation calculation to obtain the risk dimension correlation coefficient matrix. Pearson correlation calculation is a method to measure the strength of the linear relationship between two variables. By calculating the correlation coefficient, the degree of correlation between different risks can be understood. Based on the risk dimension correlation coefficient matrix, the target risk factor pairs are determined. These factor pairs are risk factor pairs with strong correlation in the risk feature data set. By determining the target risk factor pairs, it is identified which risks have important mutual influences. The target risk factor pairs are input into the threshold recurrent network model, and the data are filtered by the gated threshold to remove the time series noise and obtain the filtered risk association data. The filtered risk association data is processed by wavelet transform. Wavelet transform is an effective signal denoising method, which decomposes the signal into different frequency components, thereby realizing fine denoising operations on the data. Through this process, the denoised risk association sequence is obtained. Based on the risk association sequence after noise reduction, a directed risk propagation network is constructed, and the in-degree, out-degree and betweenness centrality of the risk nodes are calculated to obtain the influence index of the risk nodes. In the directed risk propagation network, each node represents a specific risk factor, and the directed edges between nodes represent the risk propagation path. By calculating the in-degree and out-degree of each node, the position and importance of each risk factor in the propagation network can be understood; and the betweenness centrality is used to measure the intermediary role of the node in the entire network, that is, whether a node is on the critical path between other nodes. Through calculation, the influence index of each risk node is quantified, so as to identify the key risks with significant propagation effects in the project. Based on the risk node influence index, a standardized risk association intensity value is generated, and the standardized risk association intensity value is matrix-reconstructed according to the risk dimension to generate the final risk association intensity matrix. The standardization process is to quantify the influence of the risk node, so that the intensity between different risks can be compared and evaluated horizontally, so as to better understand the cross-influence between various risks. Through matrix reconstruction, the relationship between different risk dimensions is reintegrated into a structured risk association intensity matrix, reflecting the degree of association and interactive influence relationship between various risks.

[0019] Step 300: Based on the risk correlation strength matrix and historical risk evolution data, a comprehensive assessment is performed on the risk occurrence probability, impact degree and duration to obtain a risk level assessment vector; It should be noted that the risk correlation intensity matrix is ​​characterized by decomposing the intrinsic connection and important characteristics between risks, thereby effectively reducing the complexity of the data and retaining the key information in the risk data, and obtaining the risk correlation feature matrix. The risk correlation feature matrix is ​​matched with the historical risk evolution data in time series. By comparing the evolution mode of the historical data, the correlation characteristics between the current risk characteristics and the historical data are obtained to form a risk evolution feature sequence. The risk evolution feature sequence is input into the Bayesian network model for probability evaluation. Bayesian network is an effective probabilistic graph model used to analyze the uncertainty relationship in complex random systems. By calculating the conditional probability, the probability distribution of risk occurrence is obtained, and the risk occurrence probability vector is generated. The Bayesian network model uses the dependency relationship of conditional probability to comprehensively consider multiple possible risk paths and the mutual influence between risks, so as to more accurately describe the probability of occurrence of each risk. Based on the risk occurrence probability vector, the impact degree of each risk is evaluated to obtain the risk impact degree vector. The evaluation of the risk impact degree aims to quantify the specific impact of each risk on the project after it occurs, including the direct and indirect impact on the cost, schedule, and resources of the project, so as to clarify the potential destructiveness and severity of each risk. On this basis, the risk impact vector is analyzed in the time dimension to obtain the risk duration vector, so as to determine the time duration of each risk, that is, how long the risk will affect the project after it occurs. The risk probability vector, the risk impact vector and the risk duration vector are integrated in multiple dimensions to conduct a comprehensive risk assessment and obtain a comprehensive risk assessment value. The comprehensive risk assessment value represents the threat level of each risk to the overall project under the current conditions. The comprehensive risk assessment value is graded to obtain the risk grade classification result. Risks are classified according to their comprehensive threat level, so that project managers can focus on responding to high-level and major threatening risks. In the process of grade classification, the comprehensive risk assessment value is divided into multiple levels, such as low, medium and high, according to the pre-set risk assessment standards. Based on the risk grade classification results, each risk element is vectorized to generate a risk grade assessment vector. The risk grade assessment vector is a way to quantitatively represent risk characteristics, which can describe the level of each risk in a standardized form, which is convenient for calculation and analysis in subsequent risk management.

[0020] Step 400: hierarchically decompose the risk level assessment vector to generate a risk decomposition hierarchy tree; Specifically, the risk level assessment vector is split in time series according to the five project stages of project initiation, planning, execution, monitoring, and closing to obtain the risk sequence of different stages. By matching the risk data with the project life cycle, the specific risk characteristics of each stage are effectively identified to form a staged risk sequence, so that the risks of each stage can be independently evaluated and handled in the project management process. Based on the stage risk sequence, the relative importance between each two risk factors is calculated to obtain the risk priority judgment matrix. The impact of different risk factors in each stage is compared, and the priority of each risk is determined based on these comparisons. Through the risk priority judgment matrix, a global understanding of different risks is formed. Based on the risk priority judgment matrix, the power iteration method is used to calculate the maximum eigenvalue of the matrix and its corresponding eigenvector to obtain the risk weight vector of each stage. The power iteration method is an efficient numerical calculation method used to extract the main eigenvalue of the matrix and its corresponding eigenvector, determine the relative importance of the risk priority, and thus assign a weight value to each risk. These weights reflect the importance of each risk in a specific stage. The consistency check of the stage risk weight vector is carried out to verify the consistency of the judgment matrix, that is, to evaluate whether the judgment matrix is ​​logically consistent, to ensure that no artificial bias is introduced in the weight calculation process, and to obtain the revised stage risk weight. The revised stage risk weight is multiplied by the risk sequence of the corresponding stage one by one, the impact of the risk in each stage is calculated, and the quantitative value of the risk cumulative effect is calculated by the least squares method. The least squares method is a statistical optimization method used to minimize the sum of squares of errors so that the final cumulative effect can truly reflect the impact of different risks in each stage on the overall project. The quantitative value of the risk cumulative effect can reflect the overall influence of the risk in each stage. The quantitative value of the risk cumulative effect is hierarchically decomposed, and the transmission probability between different risk nodes is calculated to obtain the risk transmission coefficient matrix. The hierarchical decomposition operation helps to better understand the transmission and interaction mechanism between risks by gradually disassembling the complex cumulative effect into simpler impact units. The calculation of the transmission probability can quantify the possibility of risk transmission from one node to another and reveal the transmission path of risk in the project stage. By quantitatively analyzing the relationship between these risk nodes, a risk transmission coefficient matrix is ​​constructed to depict the transmission characteristics of risk in the entire project system. Based on the risk transfer coefficient matrix, forward calculation is performed to obtain the risk diffusion path diagram, which shows the transmission path and interaction mode of each risk in the project. Through graphical representation, it reflects how a certain risk affects other risks through different nodes. The risk diffusion path diagram is node traversed and hierarchical processed to finally generate a risk decomposition hierarchy tree.By traversing the risk nodes, the subordinate relationships and mutual influence paths between the nodes are identified, and through hierarchical processing, the complex risk propagation relationships are organized in a hierarchical form, so that the final risk decomposition hierarchy tree can show the propagation and evolution of risks in a hierarchical manner.

[0021] Step 500: Perform multi-level risk classification on the risk decomposition hierarchy tree to construct a multi-level risk classification matrix; Specifically, each risk node in the risk decomposition hierarchy tree is hierarchically extracted, and the risk nodes are split according to the five stages of project initiation, planning, execution, monitoring and closing to obtain a set of risk nodes in each stage. All risks in the project are organized in chronological order, so that specific risks in each stage can be analyzed independently, and it is helpful to more targetedly evaluate and deal with risks in different stages during the project management process. For each risk node in the risk node set of each stage, the values ​​of the three characteristics of risk occurrence frequency, impact range and duration are calculated to construct a multidimensional risk feature vector. The risk occurrence frequency reflects the probability of risk occurring in a certain stage, the impact range describes the potential impact of the risk on the project, and the duration is used to evaluate the duration of the risk impact on the project. Through these three characteristics, the characteristics of each risk node are described, and a risk feature vector containing multidimensional information is generated. The multidimensional risk feature vector is input into the k-means clustering algorithm, and the risk nodes in the same stage are grouped by similarity by calculating the Euclidean distance and iteratively updating the cluster center to obtain a preliminary classification result. In the clustering process, the k-means algorithm minimizes the Euclidean distance between risk nodes, assigns similar risk nodes to the same category, and realizes the grouping of different risks in each stage. Through this clustering method, it is effective to identify which risk nodes have similar characteristics, so that they are classified into the same group. The preliminary classification results are analyzed and sorted, and these classification results are sorted from high to low according to the risk level. The boundary value of the risk level is determined by calculating the average distance of nodes in the same risk level and the minimum distance between nodes of different risk levels. The risk level boundary value is an important indicator for distinguishing different risk categories. It can clearly define the scope and boundary of each risk category and ensure that each risk level is independent and non-overlapping when dividing the risk level. By calculating these boundary values, the risk level can be divided more accurately, so that the relative importance of different risks can be clearly defined. According to the determined risk level boundary value, the risk nodes at each stage are scored to obtain the quantitative weight of the risk node. The quantitative weight of each risk node reflects its relative importance in a specific stage. In order to reflect the comprehensive influence of the risk node in the entire project, the quantitative weight is multiplied by the weight coefficient of the level where the risk node is located to obtain a comprehensive risk weight value, which reflects the position and importance of the risk node in the entire project structure. The comprehensive risk weight value is mapped by a piecewise function, and the risk nodes are divided into three levels: high risk, medium risk and low risk, to obtain the risk classification results for each stage. The piecewise function mapping divides the comprehensive weight value into intervals to ensure that each risk node can be assigned to an appropriate risk level according to its influence, making the risk classification results clearer and more operational.The risk grading results of each stage are organized into a matrix form, where m represents the number of project stages and n represents the number of risk levels, ultimately generating a multi-level risk classification matrix.

[0022] Step 600: Establish a risk response decision table based on the multi-level risk classification matrix and generate a risk management control strategy.

[0023] Specifically, the multi-level risk classification matrix is ​​processed in blocks, and the risk nodes in the matrix are disassembled according to different stages and risk levels to obtain a hierarchical risk data set. For each risk node in the hierarchical risk data set, the exponential smoothing model is applied to calculate the changing trend of risk frequency, impact scope and duration in the time series to generate a dynamic risk indicator matrix. The exponential smoothing model effectively captures the changes in risk characteristics over time and emphasizes the impact of the latest data, so as to reasonably predict the dynamic evolution of risk characteristics and obtain a matrix covering the dynamic change trend of risk. The dynamic risk indicator matrix is ​​mined for association rules to generate association rules between risk indicators. Through association rule mining, the potential relationship between different risk indicators is revealed, including the interaction between risk occurrence, impact scope and duration. The risk indicator association rules are input into the CART decision tree model to calculate the feature splitting node through the Gini coefficient and determine the key factors affecting the development of risk. The CART decision tree is a classification and regression tree model that measures the purity of the data set through the Gini coefficient, determines the best feature splitting node, and finds the most critical decision point affecting the risk. In the process of generating the decision tree, the decision path is pruned to avoid model overfitting and ensure the simplicity and effectiveness of the risk response rule set. The purpose of the pruning operation is to remove branches that contribute less to the prediction accuracy of the model, retain the most representative risk response rules, and obtain a streamlined and effective risk response rule set. The risk response rule set is further processed, and the weight coefficient of each rule is calculated by information entropy to generate a priority table of risk prevention and control measures. Information entropy is an important indicator for measuring system uncertainty. By calculating the information entropy for each risk response rule, the contribution of each rule in reducing system uncertainty is quantified, so that each rule is assigned a weight coefficient to determine the priority of each risk response measure and generate a priority table. Based on the priority table, the risk status of each stage is recursively estimated to obtain a real-time risk status sequence. The recursive estimation process can dynamically reflect the state changes of risks at different stages of the project, ensuring the adaptability and timeliness of risk response measures. Fuzzy reasoning operations are performed on the real-time risk status sequence to obtain the optimization scheme of the prevention and control strategy. Fuzzy reasoning operations are a method of processing uncertainty information. Under the fuzzy description of the risk status, appropriate prevention and control measures can be derived, thereby formulating the optimal prevention and control strategy for different risk scenarios. Through fuzzy reasoning, we can more flexibly deal with various complex and uncertain risk situations in the project and develop the most practical optimization prevention and control strategies. When organizing the optimization plan of the prevention and control strategy in time sequence, we consider the various stages of the project and the different levels of risk to ensure the rationality of the prevention and control strategy in time and its applicability in different stages of the project. According to the various stages and risk levels of the project, a dual index structure is established to generate risk management control strategies.The dual index structure helps project managers quickly find and apply appropriate control strategies when facing specific stages and specific risk levels. By organizing risk management control strategies according to stages and levels, it can ensure that each stage has targeted control measures, and the prevention and control strategies are reasonably allocated and applied according to the severity of the risk, thereby minimizing the potential risk impact in the project.

[0024] In the embodiment of the present application, through multi-dimensional risk feature extraction and bidirectional correlation analysis, the cross-impact calculation between quality risk-cost risk and schedule risk-resource risk is realized, and the correlation strength between different risk dimensions is accurately quantified. The hierarchical risk decomposition and linear regression model are adopted to accurately calculate the risk accumulation effect, establish a complete risk decomposition hierarchy tree, and reveal the transmission law of risk in different project stages. Based on the multi-level risk classification model, the risk nodes are subjected to feature similarity calculation and weight distribution, and a multi-level risk classification matrix is ​​constructed to make the risk classification more accurate and reasonable. Through the calculation of the dynamic risk indicator matrix and the real-time risk state sequence, a complete risk response decision-making mechanism is established, so that risk prevention and control measures can be dynamically adjusted according to actual conditions. Through the application of association rule mining and decision tree model, a systematic risk management control strategy is formed, which enhances the scientificity and operability of risk prevention and control. The prevention and control strategy is optimized by fuzzy reasoning operation, and a risk management system with a dual index structure is established to make the risk management process more systematic and standardized.

[0025] In a specific embodiment, the process of executing step 100 may specifically include the following steps: Perform data cleaning and standardization calculation on the project monitoring data to obtain a standardized monitoring data matrix; According to the four basic attributes of risk probability, impact, duration and correlation, the standardized monitoring data matrix is ​​numerically decomposed and attribute mapped to obtain the basic characteristic attribute matrix; Perform feature mapping and nonlinear transformation on the basic feature attribute matrix to obtain the initial risk feature vector, and perform covariance matrix calculation and eigenvalue decomposition operation on the initial risk feature vector to obtain the principal component weight coefficient matrix; Based on the principal component weight coefficient matrix, the risk characteristic data is linearly combined and reconstructed to obtain the risk dimension sequence; A contribution threshold is set based on the risk dimension sequence, and risk features with a contribution greater than the threshold are classified into the core feature set, and risk features with a contribution less than the threshold are classified into the auxiliary feature set to obtain a risk feature data set.

[0026] Specifically, the monitoring data of the engineering project is cleaned and standardized to ensure the quality and consistency of the data. Data cleaning is the process of eliminating noise, missing values ​​and outliers in the data to ensure the integrity and reliability of the data. For example, there are missing data caused by sensor failure and outliers caused by erroneous measurements in the monitoring data. Through data cleaning, data that interferes with subsequent analysis is removed. Standardized calculations are performed on the data to convert the original monitoring data into a standardized form to form a standardized monitoring data matrix. Assume that the original data set is a matrix ,in represents the number of data samples, Indicates the number of monitored variables. Eliminate the impact between different data dimensions and scale the data of each variable to a unified range for calculation. The standardization process uses the Z-score standardization method, and its formula is as follows: ; in, is the normalized data value, is the original data, Indicates The mean of the variables, For the After standardization, each variable in the data set will become a standard normal distribution with a mean of zero and a standard deviation of one, and a standardized monitoring data matrix will be obtained. According to the four basic attributes of risk probability, impact, duration and correlation, the standardized monitoring data matrix is ​​numerically decomposed and attribute mapped to obtain the basic characteristic attribute matrix. Assume that the standardized monitoring data matrix is , decompose each monitoring data and map it to four risk attributes. For each attribute (risk occurrence probability , Impact , Duration and relevance ), respectively Perform weighted summation to form the basic feature attribute matrix , and its calculation process is expressed as: ; in, For the The variable in The weights on the risk attributes are For sample In the Through this process, the basic feature attribute matrix It contains the characteristic description of each monitoring sample in terms of the four attributes of risk probability, impact, duration and correlation. The basic characteristic attribute matrix is ​​subjected to feature mapping and nonlinear transformation to obtain the initial risk feature vector. Assume that the basic characteristic attribute matrix is , applying a nonlinear transformation function to it, such as , get the initial risk feature vector : ; in, is a matrix The Row vector, representing samples Characteristic description of It is a nonlinear mapping function, such as polynomial transformation or RBF (radial basis function), which maps the basic features to a high-dimensional space through nonlinear transformation. The covariance matrix and eigenvalue decomposition operation are performed on the initial risk feature vector to obtain the principal component weight coefficient matrix. Calculate the covariance matrix of the initial risk feature vector : ; in, represents the mean of all initial risk feature vectors, is the vector after removing the mean. Covariance matrix Reflects the linear correlation between different features. Perform eigenvalue decomposition on the covariance matrix to obtain the eigenvalues ​​and corresponding eigenvectors: ; in, represents the eigenvalue, Represents the eigenvector, all eigenvectors form the principal component weight coefficient matrix Principal component analysis finds the direction with the largest variance in the data through eigenvalue decomposition, so as to project the data into a low-dimensional space while retaining the main information of the data. Based on the principal component weight coefficient matrix, linear combination and reconstruction operations are performed on the risk characteristic data to obtain the risk dimension sequence. Assume that the initial risk characteristic data is a matrix , the principal component weight coefficient matrix is , the risk dimension sequence matrix is ​​obtained through linear combination : ; in, is the number of features after dimensionality reduction. Through linear combination, the high-dimensional initial risk feature vector is reduced to a low-dimensional space to obtain the risk dimension sequence , which retains the main variation information of the data while reducing noise and redundancy. Based on the risk dimension sequence, a contribution threshold is set, and risk features with a contribution greater than the threshold are classified into the core feature set, and risk features with a contribution less than the threshold are classified into the auxiliary feature set to obtain a risk feature data set. The contribution of each feature is calculated, that is, the proportion of the feature in describing the total variation of the data. Assume that the eigenvalue vector is , then The contribution of each feature is: ; Set a contribution threshold ,if , the corresponding features are classified into the core feature set; otherwise, they are classified into the auxiliary feature set. In this way, the features that play a major role in the overall risk description are retained, while the features with less contribution are classified into the auxiliary set, forming a data set containing core and auxiliary features for subsequent risk management and decision support.

[0027] In a specific embodiment, the process of executing step 200 may specifically include the following steps: Performing bidirectional grouping processing on the risk feature data set, constructing the quality risk and cost risk data into a first association group, and constructing the schedule risk and resource risk data into a second association group, to obtain bidirectional association group data; Perform Pearson correlation calculation on each group of risk data in the bidirectional correlation grouped data to obtain the risk dimension correlation coefficient matrix, and generate the target risk factor pair based on the risk dimension correlation coefficient matrix; The target risk factor is input into the threshold recurrent network model, and the time series noise is filtered through the gated threshold to obtain the filtered risk association data, and the filtered risk association data is subjected to wavelet transformation to obtain the risk association sequence after noise reduction; Based on the denoised risk association sequence, a directed risk propagation network is constructed, and the in-degree, out-degree and betweenness centrality of the risk nodes are calculated to obtain the risk node influence index; A standardized risk correlation strength value is generated based on the risk node influence index, and the standardized risk correlation strength value is reconstructed according to the risk dimension to generate a risk correlation strength matrix.

[0028] Specifically, the risk feature data set is bidirectionally grouped. The data of quality risk and cost risk are constructed as the first association group, and the data of schedule risk and resource risk are constructed as the second association group to obtain bidirectionally associated grouped data. For example, quality risk will have a direct impact on cost, resulting in project budget overruns, while there is a close correlation between schedule risk and resource risk, such as insufficient resource allocation leading to project schedule delays. Targeted correlation analysis can be easily performed through grouping. After completing the bidirectional grouping, Pearson correlation calculation is performed for each set of risk data to evaluate the linear correlation between the risk factors. Suppose there are two risk data sets, namely the quality risk and cost risk data sets. and , Pearson correlation coefficient The calculation formula is: ; in, and They are and The mean of Representation variables and The correlation coefficient between them ranges from [-1,1]. When the value is close to 1, it means that the two variables are positively correlated; When the value is close to -1, it means that the two variables are negatively correlated; When the value is close to 0, it means that there is no significant linear relationship between the two variables. By performing Pearson correlation calculation on each group of risk data in the bidirectional correlation grouped data, the risk dimension correlation coefficient matrix is ​​obtained. ,in is the number of risk dimensions, each element in the matrix Indicates risk factors and risk factors The degree of correlation between them. Generate target risk factor pairs based on the risk dimension correlation coefficient matrix, screen out risk combinations with strong correlation, and obtain target risk factor pairs, which represent the mutual relationship between risks that need to be focused on in subsequent risk management. Input the target risk factor pairs into the threshold recurrent network model to filter the time series noise. The threshold recurrent network is an improved model based on the recurrent neural network, which can effectively deal with the long-term dependency and noise problems in time series data. In the threshold recurrent network model, the gating threshold is used to filter out irrelevant time series noise, and the time series data closely related to risk propagation is retained. Through the operation of the gating unit, the filtered risk association data is obtained, so that the noise of the data is significantly reduced, and more critical risk information is retained. The filtered risk association data is subjected to wavelet transform to remove residual nonlinear noise. Wavelet transform is an effective signal analysis tool that decomposes the signal into different frequency components, thereby realizing refined noise reduction operations on the time series. Assume that the filtered risk association data is signal , then the mathematical expression of wavelet transform is: ; in, is the wavelet transform coefficient, and denote the scale and location parameters respectively, is the complex conjugate of the wavelet mother function. Through wavelet transform, multi-scale analysis of risk association data is performed to extract useful information and remove interference to obtain the risk association sequence after noise reduction. Based on the risk association sequence after noise reduction, a directed risk propagation network is constructed. In this network, each node represents a specific risk factor, and the directed edges between nodes represent the causal relationship or impact path between risk factors. To construct a directed risk propagation network, the in-degree, out-degree and betweenness centrality of each risk node are calculated to evaluate the role and influence of each node in the network. Assuming a directed graph represents the risk propagation network, where Represents a collection of nodes. represents a set of directed edges, then the node In-degree Indicates the number of incoming edges connected to the node, outgoing degree Indicates the number of outgoing edges connected to the node. The calculation formula is: ; in, Representation Node To Node The number of shortest paths, Indicates the nodes passed through in these shortest paths The number of nodes in the network. Betweenness centrality reflects the intermediary role of a node in the network and its importance to the information dissemination between other nodes. By calculating the in-degree, out-degree and betweenness centrality, the influence index of each risk node is obtained. Based on the risk node influence index, a standardized risk association strength value is generated. This makes the various risk indicators comparable, which is convenient for subsequent analysis and processing. Assume that the influence index of a node is , then the standardized risk association intensity value It is expressed as: ; in, and Represent the minimum and maximum values ​​of all influence indicators respectively, obtained by standardization The value range is between [0,1]. According to the risk dimension, the standardized risk correlation strength value is reconstructed into a matrix to generate a risk correlation strength matrix .matrix Each element in Indicates risk factors Risk factors The intensity of impact.

[0029] In a specific embodiment, the process of executing step 300 may specifically include the following steps: Perform eigendecomposition on the risk correlation intensity matrix to obtain the risk correlation feature matrix, and perform time series matching between the risk correlation feature matrix and the historical risk evolution data to obtain the risk evolution feature sequence; Input the risk evolution feature sequence into the Bayesian network model, calculate the probability distribution of risk occurrence through conditional probability, and obtain the risk occurrence probability vector; Conduct risk impact assessment based on the risk probability vector to obtain the risk impact degree vector, and conduct time dimension analysis on the risk impact degree vector to obtain the risk duration vector; Perform multi-dimensional fusion and comprehensive risk assessment on the risk occurrence probability vector, risk impact degree vector and risk duration vector to obtain a comprehensive risk assessment value; The comprehensive risk assessment value is graded to obtain a risk grade classification result, and the risk factors are vectorized based on the risk grade classification result to generate a risk grade assessment vector.

[0030] Specifically, the risk correlation intensity matrix is ​​decomposed to reveal the inherent correlation between risks. Assume that the risk correlation intensity matrix is ,in Represents the number of risk factors, each element Indicates risk factors Risk factors In order to perform eigenvalue decomposition on this matrix, the eigenvalue decomposition method is used to transform the matrix Decomposed into the following form: ; in, is the matrix The matrix composed of the eigenvectors of is a diagonal matrix, and the elements on the diagonal are matrices The complex risk relationship is simplified and expressed as a combination of eigenvalues ​​and eigenvectors to obtain the risk association characteristic matrix . The risk association feature matrix Perform time series matching with historical risk evolution data to obtain the risk evolution feature sequence. Assume that the historical risk evolution data is , indicating that at time The state of each risk factor at the moment. The risk correlation feature matrix Match with historical data to obtain the risk evolution feature sequence , and its calculation formula is: ; The historical risk evolution data is transformed through the feature matrix to capture the key change patterns in the historical data and obtain the risk evolution feature sequence that can reflect the dynamic changes of risks. The role of the risk evolution feature sequence is to integrate the current risk correlation and the changes in historical data. The risk evolution feature sequence is input into the Bayesian network model to evaluate the probability distribution of each risk by calculating the conditional probability. The Bayesian network is a risk assessment tool based on a probabilistic graphical model that can calculate the conditional probability using the causal relationship between risk factors. Assume that the risk evolution feature sequence is , each node in the Bayesian network represents a risk factor, and the edges between nodes represent the causal relationship between risk factors. Using Bayesian theorem, the conditional probability of each risk occurrence is calculated, and the formula is: ; in, Indicates that in a given risk evolution feature sequence In the case of risk The probability of occurrence, For the characteristic sequence at risk The likelihood of occurrence, For risk The prior probability of is the total probability of the feature sequence. The probability distribution of each risk factor at the current moment is obtained through the Bayesian network to form a risk occurrence probability vector ,in Indicates The probability of occurrence of risk factors. Based on the risk probability vector , evaluate the impact of each risk and obtain the risk impact vector. Assuming that the impact of each risk factor is related to its probability of occurrence, the impact of the risk is evaluated by combining the probability of risk occurrence with its historical impact data. For example, the risk impact It is expressed as: ; in, Indicates risk The historical maximum impact value, is the impact degree of the current risk. Risk impact degree vector Indicates the potential impact of each risk under current conditions. Analyze the risk impact vector in the time dimension to determine the duration of each risk. Assuming that the duration of a risk is proportional to its probability of occurrence, the duration of each risk is calculated by the cumulative impact. : ; in, Indicates at time Risk at any time By integrating the risk impact, we can get the risk duration vector , which represents the time range in which each risk factor affects the project. , risk impact vector and the risk duration vector Conduct multi-dimensional integration to conduct comprehensive risk assessment and obtain a comprehensive risk assessment value Assuming that the comprehensive assessment value is determined by the weighted sum of the risk probability, impact level and duration, the calculation formula for the comprehensive risk assessment value is: ; in, , , are weight coefficients, which respectively represent the contribution of risk probability, impact and duration to the comprehensive risk assessment. By weighted summation, the comprehensive risk assessment vector is obtained: , each Indicates The comprehensive assessment results of the risk factors are classified into different levels to determine the risk level of each risk. The comprehensive assessment values ​​are divided into different risk levels, such as low, medium and high. Assume that the boundary value of the risk level is set to and , then the risk level It is expressed as: ; Through this process, each risk factor is graded and the risk grade classification result vector is obtained. Based on the risk level classification results, each risk factor is vectorized to generate a risk level assessment vector , which is expressed as: ; Among them, the function It is a mapping function that combines the risk level and the comprehensive assessment value into the final assessment vector.

[0031] In a specific embodiment, the process of executing step 400 may specifically include the following steps: According to the five stages of project initiation, planning, execution, monitoring, and closing, the risk level assessment vector is split in time series to obtain the stage risk sequence; Based on the stage risk sequence, the relative importance between every two risk factors is calculated to obtain the risk priority judgment matrix; Based on the risk priority judgment matrix, the power iteration method is used to calculate the maximum eigenvalue and its corresponding eigenvector to obtain the stage risk weight vector; Conduct consistency check on the stage risk weight vector to obtain the revised stage risk weight; Multiply the modified stage risk weights by the risk sequence of the corresponding stage, and calculate the quantitative value of the risk cumulative effect by the least squares method; Perform hierarchical decomposition operations on the quantified value of the risk cumulative effect, and calculate the transmission probability between risk nodes to obtain the risk transmission coefficient matrix; The risk transfer coefficient matrix is ​​forward calculated to obtain the risk diffusion path diagram, and the risk diffusion path diagram is node traversed and hierarchically processed to generate a risk decomposition hierarchy tree.

[0032] Specifically, according to the five stages of project initiation, planning, execution, monitoring and closing, the risk level assessment vector is split into time series to obtain the risk sequence of different stages. Assume that the risk level assessment vector is ,in Represents the number of different risk factors, and the vector Split the project into five stages to obtain the risk sequence for each stage , The risk sequence of each stage represents the level of all risk factors related to the project in that stage. Based on the stage risk sequence, the relative importance between every two risk factors is calculated to generate a risk priority judgment matrix. There are risk factors , evaluate the importance relationship between each two risk factors. Risk priority judgment matrix Elements Indicates risk factors and risk factors The relative importance of is defined as: ; in, and Risk factors and By comparing the importance of each risk factor, a risk priority judgment matrix is ​​constructed. , and use this matrix to analyze the priority relationship between different risks. Based on the risk priority judgment matrix , use the power iteration method to calculate its maximum eigenvalue and its corresponding eigenvector, and obtain the weight vector of stage risk. The power iteration method is a numerical method used to estimate the maximum eigenvalue of a matrix and its corresponding eigenvector. Assume that the initial vector is , the recursive formula of the power iteration method is: ; in, For the The vector at the iteration, for The norm of is used to normalize the vector to avoid the value being too large or too small. After multiple iterations, when When converged, we get the matrix The largest characteristic root and its corresponding eigenvector , which is the weight vector of stage risk. In order to ensure the consistency of the risk priority judgment matrix, the consistency test of the stage risk weight vector is performed. The consistency of the judgment matrix is ​​measured by the consistency ratio (CR). Calculate the consistency index (Cl): ; in, is the maximum eigenvalue of the judgment matrix, is the order of the matrix. Then calculate the consistency ratio (CR): ; in, is a random consistency indicator, according to the order of the matrix Take different values. When , the judgment matrix is ​​considered to have satisfactory consistency, otherwise the judgment matrix needs to be revised to ensure consistency. After the consistency test, the revised stage risk weight vector is obtained. . The modified stage risk weight vector The risk sequence of the corresponding stage Multiply them together to calculate the comprehensive impact of the risk. For each stage, the comprehensive impact of the risk is expressed as: ; in, For stage In order to quantify the cumulative effect of risk, the cumulative effect quantification value is calculated by the least squares method. The goal of the least squares method is to find a set of parameters that minimizes the sum of squares of the fitting errors. Assume that the cumulative effect model is: ; in, represents the cumulative effect of risk, is the risk characteristic matrix, is the parameter vector to be estimated, is the error term. By using the least squares method, we get the parameter vector : ; The cumulative effect quantification value calculated by the least squares method is used to describe the overall impact of different risk factors on the project in each stage. Hierarchical decomposition operation is performed on the risk cumulative effect quantification value, and the transmission probability between risk nodes is calculated to obtain the risk transmission coefficient matrix. Assume that the cumulative effect quantification value is , the transmission probability Indicates that from the risk node To risk node The risk transfer probability and the transfer coefficient matrix The elements of are defined as: ; By calculating the risk transfer coefficient matrix , quantifying the intensity of risk transmission between different nodes, reflecting the interaction relationship between various risk nodes. The risk transfer coefficient matrix is ​​forward calculated to obtain the risk diffusion path diagram. In the risk diffusion path diagram, nodes represent risk factors, and directed edges represent the transmission paths of risks between different nodes. Through forward calculation, the overall picture of risk diffusion is obtained, describing the risk transmission process in the project. In order to analyze the hierarchical relationship of risks, the risk diffusion path diagram is node traversed and hierarchically processed to generate a risk decomposition hierarchy tree. Decomposing the complex risk propagation process into multiple levels allows the system to intuitively understand the transmission relationship between different risk factors, thereby formulating more effective risk response strategies.

[0033] In a specific embodiment, the process of executing step 500 may specifically include the following steps: The risk nodes in the risk decomposition hierarchy tree are extracted hierarchically, and the risk nodes are split into five levels: project initiation stage, planning stage, execution stage, monitoring stage, and closing stage, to obtain the risk node set of each stage; For each risk node in the risk node set of each stage, the values ​​of the three characteristics of risk occurrence frequency, impact scope and duration are calculated to obtain a multi-dimensional risk feature vector; The multi-dimensional risk feature vector is input into the k-means clustering algorithm, and the risk nodes at the same stage are grouped by similarity by calculating the Euclidean distance and iteratively updating the cluster center to obtain the preliminary classification results. The preliminary classification results are sorted from high to low according to the risk level, and the risk level boundary value is obtained by calculating the average distance of nodes within the same risk level and the minimum distance between nodes of different risk levels; Score the risk nodes at each stage according to the risk level boundary value to obtain the quantitative weight of the risk node, and multiply the quantitative weight of the risk node by the weight coefficient of the level where the risk node is located to obtain the comprehensive risk weight value; The comprehensive risk weight value is mapped by piecewise function, and the risk nodes are divided into three levels: high risk, medium risk, and low risk. The risk grading results of each stage are obtained, and the risk grading results of each stage are organized into an m×n matrix form, where m represents the number of project stages and n represents the number of risk levels, to generate a multi-level risk classification matrix.

[0034] Specifically, the risk nodes in the risk decomposition hierarchy tree are extracted hierarchically. The risk decomposition hierarchy tree decomposes all risks in the project according to different impact relationships and propagation paths, and splits these risk nodes into five levels: the project start-up phase, planning phase, execution phase, monitoring phase, and closing phase, to obtain the risk node set of each stage. Suppose there is a risk decomposition hierarchy tree, and the risk node set in it is recorded as , each node Represents a specific risk factor. Through hierarchical extraction, these nodes are divided into five project stages to obtain a set of risk nodes for the five stages. , where each set Contains the risk nodes belonging to this stage. For each stage of the risk node set, each risk node is quantitatively analyzed, including calculating the values ​​of the three characteristics of risk frequency, impact range and duration, to obtain a multi-dimensional risk feature vector. Assuming that a certain stage The risk node in , then the frequency of risk occurrence Scope of influence and duration They represent the probability of risk occurrence, the degree of impact of risk on the project, and the time span of risk impact. Multidimensional risk feature vector It is expressed as: ; Each risk node is described as a vector containing three features, so that the system can perform subsequent analysis and processing on each risk node in the form of data. The multidimensional risk feature vector is input into the k-means clustering algorithm to group the risk nodes at the same stage by similarity. k-means is an unsupervised learning algorithm used to group data points into clusters based on their feature similarity. clusters. The algorithm randomly selects The initial cluster center is then calculated by calculating the Euclidean distance from each risk node to the cluster center to determine the category to which the node belongs. Assume that the cluster center is ,in Indicates cluster centers, then the risk nodes To cluster center The Euclidean distance is expressed as: ; in, Represents cluster centers Coordinates in the three dimensions of frequency of occurrence, scope of impact, and duration. By calculating the Euclidean distance from all risk nodes to each cluster center, each risk node is assigned to the cluster to which the cluster center closest to it belongs. Update the cluster center of each cluster to the mean position of all risk nodes in the cluster, and repeat this process until the cluster center no longer changes significantly, and obtain the preliminary classification results of each risk node. Analyze the preliminary classification results, and sort the risk nodes from high to low according to the risk level to determine the risk level boundary value. By calculating the average distance of nodes within the same risk level and the minimum distance between nodes of different risk levels, the boundary values ​​of different risk levels are obtained. Assume that the preliminary classification results contain Risk level, for level The set of all nodes , and its average distance is expressed as: ; in, Representing a collection The number of nodes in Representation Node With Node By calculating the average distance of nodes within each level and the minimum distance between nodes of different levels, the boundaries between risk levels are clarified and the boundary values ​​of risk levels are obtained. According to the obtained risk level boundary value, the risk nodes at each stage are scored to determine the quantitative weight of each risk node. The risk level is , then its quantized weight It is expressed as: ; in, is the boundary value of the corresponding level, is the multidimensional feature vector of the risk node. Through the scoring process, the quantitative weight of each risk node reflects its potential threat to the project. The quantitative weight is combined with the weight coefficient of the level where the risk node is located. Multiply them together to get the comprehensive risk weight value : ; in, It is the weight coefficient of the stage where the risk node is located. Different stages have different importance. For example, the execution stage and the monitoring stage have higher weight coefficients to reflect their key role in the overall project. Perform piecewise function mapping to divide risk nodes into three levels: high risk, medium risk, and low risk, and obtain risk grading results for each stage. Set the threshold of the piecewise function and , the comprehensive risk weight value Mapping to risk levels : ; Through this process, the risk nodes in each stage are divided into risk levels, and the classification results are organized into The matrix form of represents the number of project phases (i.e. 5 phases), Indicates the number of risk levels (low, medium, high), generating a multi-level risk classification matrix , where each element Indicated in of the project phases The number of risk nodes at each risk level.

[0035] In a specific embodiment, the process of executing step 600 may specifically include the following steps: The multi-level risk classification matrix is ​​divided into blocks to obtain a hierarchical risk data set, and an exponential smoothing model is applied to each risk node in the hierarchical risk data set to calculate the changing trends of risk occurrence frequency, impact scope, and duration in time series, and obtain a dynamic risk indicator matrix; Perform association rule mining on the dynamic risk indicator matrix to generate risk indicator association rules, and input the risk indicator association rules into the CART decision tree model. Calculate the feature splitting nodes through the Gini coefficient, and prune the decision path to obtain the risk response rule set. The risk response rule set is used to calculate the weight coefficient of each rule according to the information entropy, generate a priority table of risk prevention and control measures, and recursively estimate the risk status based on the priority table to obtain a real-time risk status sequence; Fuzzy reasoning operations are performed on the real-time risk status sequence to obtain the prevention and control strategy optimization plan, and the prevention and control strategy optimization plan is organized in time series. A dual index structure is established according to the project stage and risk level to generate a risk management control strategy.

[0036] Specifically, the multi-level risk classification matrix is ​​processed into blocks to obtain a graded risk data set. Indicated in Each risk level in the project phase The number of risk nodes. The data is processed in blocks to obtain a hierarchical risk data set, where each block represents a subset of a certain stage or a specific risk level. The exponential smoothing model is applied to each risk node in the hierarchical risk data set to calculate the changing trend of the risk frequency, impact scope and duration in the time series, and obtain a dynamic risk indicator matrix. The exponential smoothing model is a method for time series prediction. It smoothes the data by assigning exponentially decreasing weights to historical data, so that recent data has a greater impact on future trends. Suppose the historical occurrence frequency of a risk node is the time series , then its exponential smoothing model is expressed as: ; in, is the smoothed predicted value at the next moment, is the actual value at the current moment, is the smoothing coefficient, and its value range is (0,1). Similarly, the same exponential smoothing method is applied to the time series of the impact range and duration to obtain the corresponding smoothed prediction value. By performing exponential smoothing on the three characteristic values ​​of each risk node, the changing trend of the risk characteristics over time can be better identified, thereby constructing a dynamic risk indicator matrix ,in is the number of risk nodes, and each row vector represents the dynamic indicator of a risk node. Association rule mining is performed on the dynamic risk indicator matrix to reveal the potential association between different risk features and generate risk indicator association rules. The purpose of association rule mining is to discover the symbiotic relationship or causal relationship between risk features. Suppose the dynamic risk indicator is ,in Indicates The association rule mining process is to Perform frequent item set analysis to generate rules that meet the support and confidence thresholds. Assume that the support is , represents the feature set The frequency of occurrence in the risk node set, and the confidence Indicated in When it appears The probability of occurrence is expressed as: ; Through these association rules, we can identify which features have strong mutual relationships. The generated risk indicator association rules are input into the CART decision tree model to analyze and classify risk factors. The CART (Classification and Regression Tree) model constructs a decision tree by recursively partitioning the data set. The key is to find the best feature splitting node. The selection of the splitting node is based on the Gini coefficient, which is used to measure the purity of the data set and is defined as: ; in, is the Gini coefficient, is the number of categories, Indicates The smaller the Gini coefficient, the purer the data set and the better the splitting effect. The CART model selects the feature with the smallest Gini coefficient for splitting and finally constructs a decision tree. During the construction process, in order to prevent the decision tree from overfitting, the decision path is pruned and the branches that have little impact on the prediction accuracy are removed to maintain the simplicity and generalization ability of the model and obtain the risk response rule set. The risk response rule set is weighted to determine the relative importance of each rule. The weight coefficient of each rule is calculated using information entropy. It is used to measure the uncertainty of the system and is defined as: ; in, Indicates The probability of a rule in the rule set. By calculating the information entropy of each rule, its importance in reducing system uncertainty is quantified, so that each rule is assigned a corresponding weight coefficient. Based on the weight coefficient, a priority table of risk prevention and control measures is generated, and rules with high priority need to be given priority in risk prevention and control. Based on the priority table, the risk state is recursively estimated to obtain a real-time risk state sequence. Recursive estimation is a dynamic update method that predicts the future risk state based on the current risk state and its changing trend. Suppose the risk state at the current moment is , then the risk state at the next moment Calculated by the following recursive formula: ; in, It represents the risk change at the current moment, which is estimated by the implementation of risk prevention and control measures and the changing trend of risk characteristics. Through this recursive process, a real-time risk state sequence is generated to reflect the evolution of risk in different stages of the project. Fuzzy reasoning operation is performed on the generated real-time risk state sequence to develop an optimization plan for the prevention and control strategy. Fuzzy reasoning is a reasoning method based on fuzzy logic, which is suitable for dealing with risk problems with uncertainty and ambiguity. Suppose the risk state Represented as a fuzzy set, such as "low risk", "medium risk" and "high risk", the optimization scheme of the prevention and control strategy is obtained through reasoning with fuzzy rules. The optimization scheme of the prevention and control strategy is organized in time series, and a double index structure is established according to the project stage and risk level to generate the final risk management control strategy. The establishment of the double index structure is to ensure the rationality of the prevention and control strategy in time and space, so as to take appropriate control measures for different risk levels at different stages of the project. Assume that the prevention and control strategy at a certain stage is ,in represents the project stage, then the double index structure is expressed as ,in Indicates the risk level (e.g. "high risk", "medium risk", "low risk"). Through structured representation, you can quickly find the prevention and control measures for a specific stage and specific risk level, thereby improving the efficiency and response speed of risk management.

[0037] The above describes the project risk management method in the embodiment of the present application. The following describes the project risk management system 10 in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the engineering project risk management system 10 includes: The feature extraction module 11 is used to extract features and divide dimensions of the project monitoring data to obtain a risk feature data set; The cross-impact analysis module 12 is used to perform a cross-impact analysis of quality risk-cost risk and schedule risk-resource risk on the risk feature data set using a bidirectional correlation calculation mechanism to obtain a risk correlation intensity matrix; Comprehensive evaluation module 13, used to comprehensively evaluate the risk occurrence probability, impact degree and duration based on the risk correlation strength matrix and historical risk evolution data, and obtain a risk level evaluation vector; A hierarchical decomposition module 14 is used to perform hierarchical decomposition on the risk level assessment vector to generate a risk decomposition hierarchical tree; The multi-level risk classification module 15 is used to perform multi-level risk classification on the risk decomposition hierarchy tree and construct a multi-level risk classification matrix; The generation module 16 is used to establish a risk response decision table according to the multi-level risk classification matrix and generate a risk management control strategy.

[0038] Through the collaboration of the above components, through multi-dimensional risk feature extraction and bidirectional correlation analysis, the cross-impact calculation between quality risk-cost risk and schedule risk-resource risk is realized, and the correlation strength between different risk dimensions is accurately quantified. The hierarchical risk decomposition and linear regression model are used to accurately calculate the cumulative effect of risk, establish a complete risk decomposition hierarchy tree, and reveal the transmission law of risk in different project stages. Based on the multi-level risk classification model, the feature similarity calculation and weight allocation of risk nodes are carried out, and a multi-level risk classification matrix is ​​constructed to make the risk classification more accurate and reasonable. Through the calculation of the dynamic risk indicator matrix and the real-time risk state sequence, a complete risk response decision-making mechanism is established, so that risk prevention and control measures can be dynamically adjusted according to actual conditions. Through the application of association rule mining and decision tree model, a systematic risk management control strategy is formed, which enhances the scientificity and operability of risk prevention and control. Fuzzy reasoning operation is used to optimize the prevention and control strategy, and a risk management system with a dual index structure is established to make the risk management process more systematic and standardized.

[0039] See also Figure 3 , Figure 3 This is a schematic block diagram of the structure of an electronic device 300 provided in an embodiment of the present application. The electronic device 300 includes a processor 301 and a memory 302. The processor 301 and the memory 302 are connected via a system bus 303, wherein the memory 302 may include a non-volatile storage medium and an internal memory.

[0040] The non-volatile storage medium can store a computer program. The computer program includes program instructions, and when the program instructions are executed by the processor 301, the processor 301 can execute any of the above-mentioned engineering project risk management methods.

[0041] The processor 301 is used to provide computing and control capabilities to support the operation of the entire electronic device 300 .

[0042] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor 301, the processor 301 can execute any of the above-mentioned engineering project risk management methods.

[0043] Those skilled in the art will understand that Figure 3 The structure shown in the figure is only a block diagram of a partial structure related to the present application scheme, and does not constitute a limitation on the electronic device 300 involved in the present application scheme. The specific electronic device 300 may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0044] It should be understood that the processor 301 may be a central processing unit (CPU), and the processor 301 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0045] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the electronic device 300 described above can refer to the corresponding process of the aforementioned engineering project risk management method, and will not be repeated here.

[0046] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by one or more processors, the one or more processors implement the engineering project risk management method provided in the embodiment of the present application.

[0047] The computer-readable storage medium may be an internal storage unit of the electronic device 300 in the aforementioned embodiment, such as a hard disk or memory of the electronic device 300. The computer-readable storage medium may also be an external storage device of the electronic device 300, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc. equipped with the electronic device 300.

[0048] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0049] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable an electronic device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program codes.

[0050] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for risk management of engineering projects, characterized in that: The engineering project risk management method includes: Extract features and divide dimensions of engineering project monitoring data to obtain risk feature data sets; A bidirectional correlation calculation mechanism is used to perform a cross-impact analysis of quality risk-cost risk and schedule risk-resource risk on the risk feature data set to obtain a risk correlation intensity matrix; Based on the risk correlation intensity matrix and historical risk evolution data, a comprehensive assessment is performed on the probability of risk occurrence, degree of impact and duration to obtain a risk level assessment vector; Decomposing the risk level assessment vector hierarchically to generate a risk decomposition hierarchy tree; Performing multi-level risk classification on the risk decomposition hierarchy tree to construct a multi-level risk classification matrix; A risk response decision table is established based on the multi-level risk classification matrix, and a risk management control strategy is generated.

2. The engineering project risk management method according to claim 1, characterized in that: The feature extraction and dimension division of the engineering project monitoring data to obtain a risk feature data set includes: Perform data cleaning and standardization calculation on the project monitoring data to obtain a standardized monitoring data matrix; According to the four basic attributes of risk occurrence probability, impact degree, duration and correlation, the standardized monitoring data matrix is ​​numerically decomposed and attribute mapped to obtain a basic characteristic attribute matrix; Performing feature mapping and nonlinear transformation on the basic feature attribute matrix to obtain an initial risk feature vector, and performing covariance matrix calculation and eigenvalue decomposition operation on the initial risk feature vector to obtain a principal component weight coefficient matrix; Performing linear combination and reconstruction operations on the risk characteristic data based on the principal component weight coefficient matrix to obtain a risk dimension sequence; A contribution threshold is set based on the risk dimension sequence, risk features with a contribution greater than the threshold are classified into a core feature set, and risk features with a contribution less than the threshold are classified into an auxiliary feature set, to obtain a risk feature data set.

3. The engineering project risk management method according to claim 2, characterized in that: The bidirectional correlation calculation mechanism is used to perform a cross-impact analysis of quality risk-cost risk and schedule risk-resource risk on the risk feature data set to obtain a risk correlation intensity matrix, including: Performing bidirectional grouping processing on the risk feature data set, constructing the quality risk and cost risk data into a first associated group, and constructing the schedule risk and resource risk data into a second associated group, to obtain bidirectional associated group data; Performing Pearson correlation calculation on each group of risk data in the bidirectionally associated grouped data to obtain a risk dimension correlation coefficient matrix, and generating a target risk factor pair based on the risk dimension correlation coefficient matrix; Input the target risk factor into the threshold recurrent network model, filter the time series noise through the gated threshold to obtain filtered risk association data, and perform wavelet transform on the filtered risk association data to obtain a risk association sequence after noise reduction; Based on the risk association sequence after noise reduction, a directed risk propagation network is constructed, and the in-degree, out-degree and betweenness centrality of the risk node are calculated to obtain the risk node influence index; A standardized risk association strength value is generated based on the risk node influence index, and the standardized risk association strength value is matrix-reconstructed according to the risk dimension to generate a risk association strength matrix.

4. The engineering project risk management method according to claim 3 is characterized in that: Based on the risk correlation intensity matrix and historical risk evolution data, a comprehensive assessment is performed on the risk occurrence probability, impact degree and duration to obtain a risk level assessment vector, including: Performing feature decomposition on the risk association intensity matrix to obtain a risk association characteristic matrix, and performing time series matching on the risk association characteristic matrix with historical risk evolution data to obtain a risk evolution characteristic sequence; Inputting the risk evolution feature sequence into the Bayesian network model, calculating the probability distribution of risk occurrence through conditional probability, and obtaining a risk occurrence probability vector; Performing a risk impact assessment based on the risk occurrence probability vector to obtain a risk impact degree vector, and performing a time dimension analysis on the risk impact degree vector to obtain a risk duration vector; Performing multi-dimensional fusion and comprehensive risk assessment on the risk occurrence probability vector, the risk impact degree vector and the risk duration vector to obtain a comprehensive risk assessment value; The comprehensive risk assessment value is graded to obtain a risk grade classification result, and risk factors are vectorized based on the risk grade classification result to generate a risk grade assessment vector.

5. The engineering project risk management method according to claim 4, characterized in that: The step of performing hierarchical decomposition of the risk level assessment vector to generate a risk decomposition hierarchy tree includes: According to the five stages of project initiation, planning, execution, monitoring and closing, the risk level assessment vector is split in time series to obtain a stage risk sequence; Based on the risk sequence of the stage, the relative importance between every two risk factors is calculated to obtain a risk priority judgment matrix; Based on the risk priority judgment matrix, the power iteration method is used to calculate the maximum eigenvalue and its corresponding eigenvector to obtain the stage risk weight vector; Performing a consistency check on the stage risk weight vector to obtain a revised stage risk weight; The modified stage risk weights are multiplied by the risk sequence of the corresponding stage respectively, and the quantitative value of the risk cumulative effect is calculated by the least square method; Performing a hierarchical decomposition operation on the quantified value of the risk cumulative effect, and calculating the transmission probability between risk nodes to obtain a risk transmission coefficient matrix; The risk transfer coefficient matrix is ​​forward calculated to obtain a risk diffusion path diagram, and the risk diffusion path diagram is node traversed and hierarchically processed to generate a risk decomposition hierarchical tree.

6. The engineering project risk management method according to claim 5, characterized in that: The risk decomposition hierarchy tree is subjected to multi-level risk classification to construct a multi-level risk classification matrix, including: Performing hierarchical extraction on the risk nodes in the risk decomposition hierarchy tree, splitting the risk nodes into five levels: project initiation stage, planning stage, execution stage, monitoring stage, and closing stage, and obtaining a set of risk nodes at each stage; Calculating the values ​​of three characteristics of risk occurrence frequency, impact scope and duration for each risk node in the risk node set of each stage to obtain a multi-dimensional risk characteristic vector; The multi-dimensional risk feature vector is input into the k-means clustering algorithm, and the risk nodes at the same stage are grouped by similarity by calculating the Euclidean distance and iteratively updating the cluster center to obtain a preliminary classification result; The preliminary classification results are sorted from high to low according to the risk level, and the risk level boundary value is obtained by calculating the average distance of nodes within the same risk level and the minimum distance between nodes of different risk levels; Scoring the risk nodes at each stage according to the risk level boundary value to obtain the quantitative weight of the risk node, and multiplying the quantitative weight of the risk node by the weight coefficient of the level where the risk node is located to obtain a comprehensive risk weight value; The comprehensive risk weight value is mapped by piecewise function, and the risk nodes are divided into three levels: high risk, medium risk, and low risk to obtain the risk grading results of each stage, and the risk grading results of each stage are organized into an m×n matrix form, where m represents the number of project stages and n represents the number of risk levels, to generate a multi-level risk classification matrix.

7. The engineering project risk management method according to claim 6, characterized in that: The step of establishing a risk response decision table according to the multi-level risk classification matrix and generating a risk management control strategy includes: The multi-level risk classification matrix is ​​processed in blocks to obtain a hierarchical risk data set, and an exponential smoothing model is applied to each risk node in the hierarchical risk data set to calculate the changing trends of the risk occurrence frequency, impact scope, and duration in the time series to obtain a dynamic risk indicator matrix; Performing association rule mining on the dynamic risk indicator matrix to generate risk indicator association rules, and inputting the risk indicator association rules into a CART decision tree model, calculating feature splitting nodes through the Gini coefficient, and pruning the decision path to obtain a risk response rule set; Calculating the weight coefficient of each rule of the risk response rule set according to information entropy, generating a priority table of risk prevention and control measures, and recursively estimating the risk state based on the priority table to obtain a real-time risk state sequence; A fuzzy reasoning operation is performed on the real-time risk state sequence to obtain a prevention and control strategy optimization scheme, and the prevention and control strategy optimization scheme is organized in time series, a dual index structure is established according to the project stage and risk level, and a risk management control strategy is generated.

8. A project risk management system, characterized in that: Used to execute the engineering project risk management method according to any one of claims 1 to 7, the engineering project risk management system comprises: The feature extraction module is used to extract features and divide dimensions of the project monitoring data to obtain a risk feature data set; A cross-impact analysis module, used to perform a cross-impact analysis of quality risk-cost risk and schedule risk-resource risk on the risk feature data set using a bidirectional correlation calculation mechanism to obtain a risk correlation intensity matrix; A comprehensive assessment module, for comprehensively assessing the probability of risk occurrence, degree of impact and duration based on the risk correlation strength matrix and historical risk evolution data, to obtain a risk level assessment vector; A hierarchical decomposition module, used for performing hierarchical decomposition on the risk level assessment vector to generate a risk decomposition hierarchical tree; A multi-level risk classification module is used to perform multi-level risk classification on the risk decomposition hierarchy tree and construct a multi-level risk classification matrix; The generation module is used to establish a risk response decision table according to the multi-level risk classification matrix and generate a risk management control strategy.

9. An electronic device, characterized in that: The electronic device comprises: a memory and at least one processor, wherein instructions are stored in the memory; The at least one processor calls the instructions in the memory so that the electronic device executes the engineering project risk management method according to any one of claims 1 to 7.

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

Images