An engineering cost whole life cycle intelligent management and optimization system

Abstract

The application discloses an engineering cost full life cycle intelligent management and optimization system, which comprises a data acquisition and preprocessing module, a feature extraction and fingerprint construction module, and the like, wherein the data acquisition and preprocessing module is used for acquiring cost data and multi-source heterogeneous associated data of each stage of the engineering cost full life cycle, and performing data cleaning and standardized preprocessing; the feature extraction and fingerprint construction module is used for performing time sequence behavior feature extraction on the preprocessed data of each stage based on engineering stage division results, and constructing a cost behavior fingerprint vector of the corresponding stage, and the application relates to the technical field of engineering cost management. The engineering cost full life cycle intelligent management and optimization system can greatly improve the cost management capability of engineering projects, reduce the fluctuation and risk of engineering cost, and ensure that the project is efficiently completed according to the budget through intelligent data processing, accurate cost analysis, dynamic weight adjustment and cross-stage optimization decision, and has a wide application prospect and market value.

Description

Technical Field

[0001] This invention relates to the field of engineering cost management, and more specifically, to an intelligent control and optimization system for the entire life cycle of engineering costs. Background Technology

[0002] In engineering projects, cost management typically covers all stages, from initial planning and design to construction and post-construction operation. Traditional cost management suffers from issues such as manual management, delayed data analysis, and inaccurate cost control, leading to risks like budget overruns and the inability to promptly identify and adjust cost deviations. Therefore, achieving precise control, timely optimization, and dynamic adjustments across different stages of cost management has become a pressing challenge for the industry.

[0003] In existing technologies, some intelligent cost management systems optimize cost management through data processing and analysis methods, but most lack comprehensive control over the entire life cycle of a project. They also have shortcomings in weight assessment, data credibility processing, and cross-stage optimization, making it difficult to effectively connect and dynamically adjust between different stages and achieve comprehensive cost control and optimization.

[0004] Therefore, there is an urgent need for an intelligent full lifecycle management and optimization system that can integrate multiple data sources, automatically analyze and generate optimization decisions to improve the accuracy, efficiency and sustainability of engineering cost management. Summary of the Invention

[0005] The purpose of this invention is to provide an intelligent management and optimization system for the entire life cycle of engineering cost. This system solves the problem that existing intelligent cost management systems optimize cost management through data processing and analysis methods, but most of them lack comprehensive control over the entire life cycle of the project. Furthermore, they still have shortcomings in weight assessment, data credibility processing, and cross-stage optimization. They cannot effectively connect and dynamically adjust between different stages, making it difficult to achieve comprehensive cost control and optimization, and thus failing to meet user needs.

[0006] This invention achieves the above objectives through the following technical solution: an intelligent management and optimization system for the entire lifecycle of engineering cost, the system comprising:

[0007] The module includes: data acquisition and preprocessing module, feature extraction and fingerprint construction module, cost deviation analysis and positioning module, impact factor evaluation and weight assignment module, weight dynamic decay processing module, and cross-stage optimization decision generation module.

[0008] The data acquisition and preprocessing module is used to acquire cost data and multi-source heterogeneous correlation data at each stage of the entire life cycle of engineering cost, and to perform data cleaning and standardization preprocessing.

[0009] The feature extraction and fingerprint construction module is used to extract time-series behavioral features from the preprocessed data of each stage based on the engineering stage division results, and construct the cost behavior fingerprint vector of the corresponding stage.

[0010] The cost deviation analysis and location module is used to automatically generate an interpretable cost deviation cause model using the cost behavior fingerprint vector, thereby locating the source of cost anomalies.

[0011] The impact factor assessment and weight assignment module is used to assess the credibility of cost impact factors formed at each stage and assign dynamic credibility weights.

[0012] The dynamic weight decay processing module is used to decay the credibility weight according to the data source, timeliness and environmental changes when switching engineering stages.

[0013] The cross-stage optimization decision generation module is used to generate a cross-stage cost optimization decision scheme based on the decayed credibility weight and the causes of cost deviations in each stage.

[0014] Furthermore, the data acquisition and preprocessing module includes:

[0015] Data partitioning unit, data cleaning unit, format unification unit, and standardization processing unit;

[0016] The data segmentation unit is used to divide the entire life cycle of project cost into multiple core stages, and to divide the data of each stage according to time nodes and business processes, so as to ensure strong correlation between data and stages.

[0017] The data cleaning unit is used to process missing values ​​using interpolation filling and to remove outliers using statistical identification.

[0018] The format unification unit is used to standardize date formats, numerical units, and encoding rules according to a unified standard.

[0019] The standardization processing unit is used to convert data of different magnitudes and dimensions into a unified scale using a normalization algorithm to meet the needs of subsequent data processing.

[0020] Furthermore, the cost data collected by the data acquisition and preprocessing module includes:

[0021] Data related to labor costs, material and equipment prices, construction process costs, and bill of quantities;

[0022] The multi-source heterogeneous correlation data includes policy and regulatory data, market data, geological and environmental data, and project management data.

[0023] The data cleaning unit uses an interpolation filling method based on the nearest neighbor algorithm, and the outlier identification method is a statistical identification method based on quantiles. The normalization algorithm used by the standardization processing unit is Z-score normalization.

[0024] Furthermore, the feature extraction and fingerprint construction module includes:

[0025] Feature extraction unit, feature weight setting unit, redundant feature removal unit, and fingerprint vector construction unit;

[0026] The feature extraction unit is used to extract core temporal behavioral features based on the temporal characteristics of data at each stage, including trend features, fluctuation features, mutation features, and correlation features.

[0027] The feature weight setting unit is used to calculate the weight of each feature based on its contribution to the cost deviation.

[0028] The redundant feature removal unit is used to identify and remove redundant features through statistical testing.

[0029] The fingerprint vector construction unit is used to fuse the selected features in terms of dimensions to construct a cost behavior fingerprint vector that condenses the core features of cost behavior at each stage.

[0030] Furthermore, the feature extraction unit calculates trend features using a linear regression correlation algorithm, calculates fluctuation features using a moving average statistical algorithm, identifies abrupt change features using a residual test correlation algorithm, and calculates association features using a correlation analysis algorithm.

[0031] The feature weight setting unit uses a feature importance evaluation algorithm to calculate feature weights and set weight thresholds. Features below the threshold are considered minor features.

[0032] The redundant feature removal unit uses the variance test method to identify redundant features, and the fingerprint vector construction unit normalizes the features to a uniform numerical range to form a fingerprint vector with fixed dimensions.

[0033] Furthermore, the cost deviation analysis and positioning module includes:

[0034] Model building unit, model training unit, core feature selection unit, causal rule generation unit, and anomaly localization unit;

[0035] The model building unit is used to construct a self-explanatory model that balances prediction accuracy and interpretability by taking cost behavior fingerprint vectors as input and cost deviation related indicators as output labels.

[0036] The model training unit is used to complete model construction through dataset partitioning, hyperparameter optimization, iterative training, and model evaluation.

[0037] The core feature screening unit is used to screen core features that have a significant impact on cost deviation through an information gain correlation algorithm.

[0038] The causal rule generation unit is used to generate a structured set of cost deviation causal rules based on model-based decision logic;

[0039] The anomaly localization unit is used to compare the real-time cost behavior fingerprint vector with the rule set through a rule matching mechanism, locate the specific source of the cost anomaly, and output an explanation report.

[0040] Furthermore, the impact factor assessment and weight assignment module includes:

[0041] The system comprises an impact factor classification unit, an evaluation system construction unit, a subjective weight calculation unit, an objective weight calculation unit, and a dynamic weight fusion unit.

[0042] The influencing factor classification unit is used to divide the cost influencing factors into material price fluctuation factors, policy adjustment factors, construction efficiency factors, management level factors, and market supply and demand factors.

[0043] The evaluation system construction unit is used to construct a three-level credibility evaluation system that includes a target layer, a criterion layer, and an indicator layer.

[0044] The subjective weight calculation unit is used to calculate the subjective weight of each influencing factor using the analytic hierarchy process (AHP) algorithm.

[0045] The objective weight calculation unit is used to calculate the objective weight of each influencing factor using the entropy weight method;

[0046] The dynamic weight fusion unit is used to introduce weight allocation coefficients and fuse subjective weights and objective weights through a weighted summation method to obtain the initial value of the dynamic credibility weight of each influencing factor.

[0047] Furthermore, the criteria layer of the evaluation system includes:

[0048] Data reliability, timeliness, and relevance; the indicator layer includes data collection accuracy, update frequency, and correlation with cost.

[0049] The weighting coefficients are dynamically adjusted according to the project stage. In the early stage of the project, subjective weights are emphasized, while in the later stage, objective weights are emphasized. The dynamic weight fusion unit also sets a weight threshold. Influence factors below the threshold are considered as low-credibility factors and are only used as a decision-making reference.

[0050] Furthermore, the weight dynamic decay processing module includes:

[0051] Attenuation triggering unit, attenuation coefficient calculation unit, and weight update unit;

[0052] The attenuation triggering unit is used to set the engineering stage switching as the triggering condition for weight attenuation;

[0053] The attenuation coefficient calculation unit is used to calculate the attenuation coefficient by comprehensively considering factors such as the reliability of the data source, the attenuation of data timeliness, and the influence of environmental changes, and by using a multiplicative model.

[0054] The weight update unit is used to update and adjust the credibility weight of each influencing factor according to the calculated attenuation coefficient. The updated weight is still normalized to a preset value range, and a weight threshold after attenuation is set. Influencing factors below the threshold can be temporarily excluded from the current stage decision-making system.

[0055] Furthermore, the cross-stage optimization decision generation module includes:

[0056] The system includes a deviation cause collection unit, a core deviation screening unit, an optimization measure generation unit, a decision priority ranking unit, and a scheme output unit.

[0057] The deviation cause collection unit is used to classify the cost deviation causes identified at each stage according to the associated influencing factors, forming a multidimensional correlation matrix.

[0058] The core deviation screening unit is used to calculate the weighted influence of various deviation causes using the attenuated confidence weight as a weighting coefficient, and to screen the core deviation causes.

[0059] The optimization measure generation unit is used to generate targeted optimization measures based on the core causes of deviations and the characteristics and business processes of each stage of the project.

[0060] The decision priority ranking unit is used to construct a multi-dimensional scoring model to comprehensively score and prioritize optimization measures.

[0061] The output unit is used to output the sorted cross-stage cost optimization decision-making scheme, clarify the implementation elements and risk control points of each measure, and form a closed-loop management and control system.

[0062] The beneficial effects of this invention are as follows:

[0063] 1. Through the data acquisition and preprocessing module, it is possible to comprehensively acquire cost data and multi-source heterogeneous correlation data at all stages of the entire life cycle of engineering cost, and use scientific and reasonable methods to clean, unify and standardize the data to ensure the quality and consistency of the data, providing a reliable data foundation for subsequent analysis.

[0064] 2. The feature extraction and fingerprint construction module accurately extracts core temporal behavioral features based on the temporal characteristics of data at each stage and constructs cost behavior fingerprint vectors. The cost deviation analysis and location module uses these fingerprint vectors to construct an interpretable cost deviation cause model, which can accurately locate the source of cost anomalies and provide strong support for cost control.

[0065] 3. The impact factor assessment and weight assignment module classifies cost-influencing factors and constructs a three-level credibility assessment system. It uses a combination of subjective and objective methods to calculate the dynamic credibility weight of each impact factor, which can accurately reflect the actual impact of each factor on the cost and improve the scientificity and rationality of the assessment.

[0066] 4. When switching between engineering phases, the dynamic weight decay processing module comprehensively considers factors such as data source, timeliness, and environmental changes. It uses a multiplication model to calculate the decay coefficient and dynamically decays the credibility weight, making the weight more adaptable to changes in actual conditions and ensuring the scientific nature of decision-making.

[0067] 5. The cross-stage optimization decision generation module classifies the causes of cost deviations identified in each stage according to related influencing factors, screens the core causes of deviations, and generates targeted optimization measures for the core causes of deviations. By constructing a multi-dimensional scoring model, the optimization measures are comprehensively scored and prioritized, and the ranked cross-stage cost optimization decision scheme is output. The implementation elements and risk control points of each measure are clarified, forming a closed-loop management and control system, effectively realizing the full life cycle optimization and control of project costs, and improving the economic and social benefits of engineering construction projects. Attached Figure Description

[0068] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0069] Figure 1 This is a system block diagram of the present invention;

[0070] Figure 2 This is a flowchart of the data acquisition and preprocessing module of the present invention;

[0071] Figure 3 This is a flowchart of the feature extraction and fingerprint construction module of the present invention;

[0072] Figure 4 This is a flowchart of the cost deviation analysis and positioning module of the present invention;

[0073] Figure 5 This is a flowchart of the cross-stage optimization decision generation module of the present invention. Detailed Implementation

[0074] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0075] Example 1:

[0076] Please see Figure 1-5 This invention provides a technical solution: an intelligent management and optimization system for the entire lifecycle of engineering cost, the system comprising:

[0077] The module includes: data acquisition and preprocessing module, feature extraction and fingerprint construction module, cost deviation analysis and positioning module, impact factor evaluation and weight assignment module, weight dynamic decay processing module, and cross-stage optimization decision generation module.

[0078] The data acquisition and preprocessing module is used to acquire cost data and multi-source heterogeneous correlation data at each stage of the entire life cycle of engineering cost, and to perform data cleaning and standardization preprocessing.

[0079] Cost data across the entire lifecycle of a project encompasses all cost-related data generated at each stage from project initiation to completion, such as estimated costs during the planning phase, preliminary costs during the design phase, budgeted and settled costs during the construction phase, and cost data during the operation and maintenance phase. Multi-source heterogeneous correlated data refers to data from multiple different sources, such as design drawings, construction records, market price information, and policy and regulatory documents. Heterogeneity indicates that these data have different structures and formats, potentially including text, images, tables, database records, and other forms. Correlated data emphasizes the logical connections between these data from different sources and structures, collectively influencing project costs. Data cleaning involves checking, filtering, correcting, and supplementing the collected raw data to remove errors, duplicates, missing values, outliers, and other inaccurate or incomplete information, improving data quality and reliability. Standardized preprocessing transforms and unifies the cleaned data according to certain rules and standards, ensuring a consistent representation of data from different sources and formats, facilitating subsequent analysis and processing.

[0080] The feature extraction and fingerprint construction module is used to extract time-series behavioral features from the preprocessed data of each stage based on the engineering stage division results, and construct the cost behavior fingerprint vector of the corresponding stage.

[0081] The project phase division results are based on the characteristics and patterns of project construction, dividing the entire life cycle of project cost into several clearly defined and characteristic phases, such as the planning phase, design phase, construction phase, completion and acceptance phase, and operation and maintenance phase. These division results provide a foundation for subsequent data processing and analysis for different phases. Time-series behavioral feature extraction extracts time-series-related feature information from the pre-processed data of each phase. These features reflect the changing trends and fluctuations of project cost at different points in time, as well as various time-related behavioral patterns. For example, the monthly cost expenditure changes during the construction phase and the impact of seasonal material price fluctuations on cost. The cost behavior fingerprint vector is a vector representation constructed based on the extracted time-series behavioral features using specific algorithms and models. Similar to a human fingerprint, it is unique and identifiable, accurately depicting the behavioral characteristics and patterns of project cost at corresponding project phases, providing a basis for subsequent cost deviation analysis and positioning.

[0082] The cost deviation analysis and location module is used to automatically generate an interpretable cost deviation cause model by utilizing cost behavior fingerprint vectors, thereby locating the source of cost anomalies.

[0083] Among them, the cost deviation cause model is a model that automatically generates through data analysis, machine learning and other methods using cost behavior fingerprint vectors to explain the causes of cost deviations. This model can reveal the intrinsic connection and influence mechanism between various factors and cost deviations, helping managers understand the root cause of cost anomalies. The source location of cost anomalies is based on the cost deviation cause model to determine the specific link, factor or event that caused the cost anomaly. For example, it can be located that a design change of a certain construction sub-project led to cost overruns, or that a sharp increase in material prices within a specific period caused cost deviations.

[0084] The impact factor assessment and weight assignment module is used to assess the credibility of cost impact factors formed at each stage and assign dynamic credibility weights.

[0085] Among them, cost influencing factors are various factors that can affect the cost throughout the entire life cycle of project cost, including but not limited to design factors, construction factors, market factors, and policy and regulatory factors; credibility assessment is a comprehensive and objective evaluation of cost influencing factors formed at each stage, judging the accuracy and reliability of their impact on cost, and considering factors such as the authority of the data source, the timeliness of the data, and the strength of the correlation between the influencing factors and the cost; dynamic credibility weight is a numerical weight assigned to each cost influencing factor based on the credibility assessment results, reflecting its importance and credibility. This weight is not fixed, but will be dynamically adjusted with changes in project stages, updates to data, and changes in environmental factors, in order to more accurately reflect the actual impact of each influencing factor on the cost under different circumstances;

[0086] The dynamic weight decay processing module is used to decay the credibility weight based on data source, timeliness, and environmental changes when switching engineering stages.

[0087] Among these, project phase transition refers to the process of a project moving from one phase to another, such as transitioning from the design phase to the construction phase, or from the construction phase to the completion and acceptance phase. Different phases have different characteristics and tasks, and the factors and degrees of influence on project cost will also change. Data sources are the origins of data related to cost influencing factors, such as design documents provided by design units, construction logs recorded by construction units, and price information released by market research institutions. Different data sources may differ in terms of accuracy, timeliness, and completeness. During project phase transition, these differences will affect the adjustment of credibility weights. Timeliness refers to the impact of the time attribute of data on its value and credibility. Over time, some data may lose its original accuracy and representativeness. For example, market price data from a long period of time may not accurately reflect the current market. When switching project phases, the timeliness of data needs to be considered to reduce the credibility weights, making the weights more reflective of the current stage. Environmental changes include changes in the macroeconomic environment, policy and regulatory environment, and natural environment. For example, the introduction of new national tax policies or natural disasters causing material supply disruptions will have varying degrees of impact on project costs. Therefore, the credibility weights need to be adjusted accordingly when switching project phases to adapt to the new environmental conditions. Credibility weight reduction is a process of reducing and adjusting the dynamic credibility weights previously assigned to each cost influencing factor according to certain rules and methods, based on factors such as data source, timeliness, and environmental changes, when switching project phases. The purpose is to make the weights more accurately reflect the actual impact of each influencing factor on the cost at the current project stage, thereby improving the scientific nature and accuracy of cost management.

[0088] The cross-stage optimization decision generation module is used to generate cross-stage cost optimization decision schemes based on the decayed credibility weights and the causes of cost deviations in each stage.

[0089] Among them, the causes of cost deviations at each stage are determined by the cost deviation analysis and positioning module, which identifies the specific reasons and influencing factors that lead to cost deviations at each stage of the entire project cost lifecycle. These causes are an important basis for formulating optimization decisions. The cross-stage cost optimization decision scheme is based on the attenuated credibility weight and the causes of cost deviations at each stage. It comprehensively considers the situation of the entire project cost lifecycle and uses optimization algorithms and decision models to generate a series of measures and schemes aimed at reducing costs and improving efficiency. These schemes may involve multiple aspects such as design optimization, construction management improvement, and procurement strategy adjustment. They also need to consider the mutual influence and coordination between different stages in order to achieve optimal control of the entire lifecycle cost.

[0090] It should be noted that during use, the data acquisition and preprocessing module acquires multi-source data and cleans and standardizes it, providing a high-quality foundation for subsequent analysis. The feature extraction and fingerprint construction module can accurately characterize the cost behavior characteristics at each stage, making it easy to grasp cost dynamics. The cost deviation analysis and location module can automatically generate causal models, quickly locate the source of anomalies, and improve problem-solving efficiency. The impact factor assessment and weight assignment module assigns dynamic weights, making the assessment more realistic. The dynamic weight decay processing module considers various factors during stage switching, making the weights more reasonable. The cross-stage optimization decision generation module, based on decay weights and deviation cause generation schemes, can coordinate the entire life cycle to achieve cost optimization. The overall design realizes intelligent and refined management from data acquisition to decision generation, effectively reducing costs, improving efficiency, and enhancing project competitiveness.

[0091] In one embodiment, cost data and multi-source heterogeneous correlation data at each stage of the entire project cost lifecycle are acquired, and data cleaning and standardization preprocessing are performed, including:

[0092] The entire lifecycle of project cost is clearly divided into the design stage, procurement stage, construction stage, and settlement stage. Data for each stage is precisely divided according to time nodes and business processes to ensure a strong correlation between data and stage.

[0093] Cost data includes labor cost data, material and equipment price data, construction process cost data, and bill of quantities data. Among them, labor cost data includes wage standards for different types of work, labor hours consumed, and labor efficiency coefficients; material and equipment price data includes market prices of main / auxiliary materials, equipment purchase prices, and transportation and installation costs; construction process cost data includes unit prices for each process step, process complexity coefficients, and quality control costs; and bill of quantities data includes the quantities of sub-items, comprehensive unit prices, and costs of provisional items.

[0094] Multi-source heterogeneous correlation data includes policy and regulatory data, market data, geological and environmental data, and project management log data. Among them, policy and regulatory data includes pricing specifications, tax rate standards, and industry regulatory requirements; market data includes material price indices, labor cost fluctuation trends, and equipment supply and demand changes; geological and environmental data includes topographic parameters, geological survey report data, and climate condition influencing factors; and project management log data includes construction progress records, quality acceptance results, and change order records.

[0095] During data cleaning, missing values ​​are handled using a method based on... Interpolation filling using the nearest neighbor algorithm, The value was determined using cross-validation, and the value with the smallest 10-fold cross-validation error was selected. value, The value is 5;

[0096] Outliers were identified using the interquartile range method, with a threshold set at [value missing]. ,in The lower quartile of the data is... Quantiles The upper quartile is... Quantiles Data exceeding this range is identified as outliers and removed.

[0097] At the same time, data format standardization was completed, and the date format was standardized to [format missing]. Numerical units are normalized according to industry standards, and coding rules adopt national standard coding.

[0098] The standardization process uses Z-score normalization to convert data of different magnitudes and dimensions into a uniform scale. The expression is:

[0099]

[0100] in, The original data values, The cost behavior fingerprint vector is the mean of this type of data, calculated from the full sample. The cost-related behavior fingerprint vector is the standard deviation of this type of data. The cost behavior fingerprint vector is a standardized data value to ensure that the data meets the requirements for subsequent feature extraction and model training.

[0101] This design clearly divides the entire lifecycle of project cost into different stages, defines the cost data and multi-source heterogeneous related data content of each stage in detail, and standardizes the data cleaning, standardization and normalization process. It accurately divides the stages and defines the data, ensuring that the data is closely related to the stages, providing a solid foundation for subsequent analysis. Scientific data cleaning and standardization can eliminate noise and differences in the data, improve data quality, and unify the data format and scale, making data from different sources and of different types comparable and compatible, which facilitates subsequent feature extraction and model training, and provides reliable data support for the accurate operation and effective decision-making of the entire system.

[0102] In one embodiment, based on the engineering phase division results, temporal behavioral features are extracted from the preprocessed data of each phase to construct a cost behavior fingerprint vector for the corresponding phase, including:

[0103] Based on the temporal continuity and stage characteristics of data at each stage, we focus on extracting four core temporal behavioral characteristics: trend characteristics, fluctuation characteristics, mutation characteristics, and correlation characteristics, to comprehensively depict the dynamic change patterns of cost data.

[0104] Trend characteristics are calculated using linear regression slopes, constructing a cost behavior fingerprint vector using a linear regression model with time as the independent variable and cost data as the dependent variable:

[0105]

[0106] Among them, slope This refers to the trend characteristic value. This indicates that costs are trending upward. This indicates that costs are trending downwards. This indicates that costs are stabilizing;

[0107] The rules for setting trend feature weights are as follows:

[0108] Based on the contribution of features to cost bias, feature importance scores are calculated using the random forest algorithm, and after normalization, they are used as trend feature weights. The weight threshold is set to 0.1, and trend features below this threshold are considered minor features.

[0109] Fluctuation characteristics are calculated using moving standard deviation. The moving window size is dynamically adjusted based on the project cycle; for small projects, the window size is 15 days. For small projects, the window is 30 days; for medium-sized projects, it is 1 year. Project cycle For medium-sized projects, the window size is 60 days; for large projects, the project cycle is [not specified]. The year is for large-scale projects; the standard deviation of the data within the sliding calculation window indicates that the larger the standard deviation, the more drastic the cost fluctuation during that period; the fluctuation characteristic threshold is set at 1.2 times the average fluctuation of similar projects in the industry, and exceeding this threshold is judged as abnormal fluctuation;

[0110] Mutation characteristics are identified using the residual test method, and a time series prediction model is established based on historical data. The model training steps are as follows:

[0111] Determine the difference order :pass Stationarity test, if but =1, otherwise =0;

[0112] Sure and Value: Based on the truncation properties of the autocorrelation coefficient and partial autocorrelation coefficient. Pick truncation order, Pick truncation order;

[0113] Model training: Parameters were estimated using maximum likelihood estimation, with 1000 iterations and a convergence threshold of 1e-6. The residuals between the actual and predicted values ​​were calculated, with a mutation threshold set to three standard deviations. The principle is established that the absolute value of the residual exceeds... ,in The mean of the residuals, The standard deviation of the residual is used. When the absolute value of the residual exceeds this threshold, it is determined to be a mutation point. The time of occurrence of the mutation, the magnitude of the mutation, and the related influencing factors are recorded.

[0114] The correlation characteristics were calculated using the Pearson correlation coefficient to analyze the linear correlation between different cost indicators, such as material prices and total cost, and labor costs and construction progress. The correlation coefficient ranged from [value missing]. The rules for setting the association strength weight are as follows:

[0115] Classified by the absolute value of the correlation coefficient The weight of the cost behavior fingerprint vector is 0.4. The weight of the cost-related behavior fingerprint vector is 0.3. The weight of the cost behavior fingerprint vector is 0.2. The weight of the cost behavior fingerprint vector is 0.1;

[0116] The four types of features are dimensionally fused, redundant features are removed, and variance inflation factor is used. test, The threshold is set to 10. The features are considered redundant; the final constructed dimension is... Cost behavior fingerprint vector:

[0117]

[0118] Among them, cost behavior fingerprint vector Cost behavior fingerprint vector representation of the first The normalized eigenvalues ​​of each behavioral feature are normalized to... interval, The total number of features is determined based on the project type and data size, and is typically... This vector comprehensively encapsulates the core characteristics of cost-related behaviors at each stage.

[0119] This design extracts four core time-series behavioral features from the data at each stage and constructs a cost behavior fingerprint vector. These four features comprehensively depict the dynamic changes in cost data, accurately reflecting the characteristics of cost behavior at different stages. By scientifically calculating various feature values ​​and reasonably setting weights and thresholds, important and secondary features can be effectively distinguished. The constructed cost behavior fingerprint vector condenses the core features of cost behavior at each stage, providing a key basis for subsequent cost deviation analysis. This helps to accurately grasp cost change trends, promptly identify potential problems, and improve the accuracy and foresight of cost control.

[0120] In one embodiment, a cost behavior fingerprint vector is used to automatically generate an interpretable cost deviation causal model, thereby locating the source of cost anomalies, including:

[0121] Fingerprint vector of cost behavior The core input is cost deviation rate, and the output label is cost deviation rate. The formula for calculating cost deviation rate is:

[0122]

[0123] Construct a self-explanatory model based on gradient boosting decision trees, balancing model prediction accuracy and interpretability;

[0124] The specific steps for model training are as follows:

[0125] Dataset partitioning: The preprocessed feature data is divided into training set, validation set and test set in a ratio of 7:2:1. Stratified sampling is used for partitioning to ensure that the cost deviation rate distribution of each dataset is consistent.

[0126] Hyperparameter optimization: A grid search combined with 5-fold cross-validation is used, with the search space set as the learning rate. Decision tree depth Number of leaf node samples Number of iterations The optimal hyperparameters are determined with the goal of minimizing the mean square error of the validation set.

[0127] Model training: The optimal hyperparameters were set as follows: learning rate 0.1, decision tree depth 8, leaf node sample size 20, and number of iterations 300. The loss function was mean squared error, the optimizer was gradient descent, and the learning rate decayed every 50 iterations. The convergence threshold is set to 1e-5;

[0128] Model evaluation: Calculate the evaluation metric, root mean square error, on the test set. Mean absolute percentage error Coefficient of determination If the model training is deemed successful, otherwise the hyperparameters are readjusted and training is iterated.

[0129] Key behavioral features were screened using information gain ratio (IFR). An IFR threshold of 0.05 was set based on industry data statistical analysis; features below this threshold were considered to have minimal impact on bias. Features were then sorted in descending order of IFR and selected from the top... The characteristics are taken as the core influencing features, and the correlation strength and positive or negative correlation between each feature and cost deviation are clarified. A positive correlation indicates that an increase in the feature value leads to an increase in deviation, while a negative correlation indicates that an increase in the feature value inhibits deviation.

[0130] Based on the splitting path of the decision tree, a structured set of rules for the causes of cost deviations is generated. The rule form is as follows:

[0131] If features and characteristics And...and features The cause of the deviation is The contribution of the deviation is ,

[0132] The deviation contribution weight is calculated by weighting the feature importance weight and the rule matching degree, with the feature importance weight accounting for a certain percentage. Rule matching degree weight ratio The threshold for deviation contribution is set as follows: Causes below this threshold are considered secondary causes;

[0133] Through a rule-based matching mechanism, the real-time constructed cost behavior fingerprint vector is compared with the rule set, and the matching degree threshold is set to... That is, the degree of overlap between the vector and the regular feature interval. If the match is successful, the specific source of the cost anomaly can be quickly located, such as deviations caused by sudden changes in material prices or cost overruns caused by adjustments in construction techniques. A detailed explanation report of the cause of the deviation is output, including key influencing characteristics, logical deduction process and quantitative impact degree, which solves the problem that traditional methods can only warn of deviation values ​​but cannot trace the cause.

[0134] This design utilizes cost behavior fingerprint vectors to construct a self-explanatory model, generating a cost deviation causal model and locating the source of anomalies. The self-explanatory model balances prediction accuracy and interpretability, clearly presenting the correlation between cost deviations and various features. Through rigorous dataset partitioning, hyperparameter optimization, and model evaluation, the model's reliability and effectiveness are ensured. By selecting key features and generating a structured logical rule set, the source of anomalies can be quickly located, and a detailed explanation report can be output. This solves the problem that traditional methods can only warn of deviation values ​​but cannot trace their causes, providing targeted decision-making basis for cost control and helping to take timely measures to correct deviations and reduce cost risks.

[0135] In one embodiment, the credibility of cost-influencing factors formed at each stage is assessed, and dynamic credibility weights are assigned, including:

[0136] Cost influencing factors include material price fluctuation factors, policy adjustment factors, construction efficiency factors, management level factors, and market supply and demand factors. Among them, the material price fluctuation factor reflects the impact of market fluctuations in the prices of major building materials; the policy adjustment factor reflects the impact of changes in pricing policies, tax policies, etc.; the construction efficiency factor is related to factors such as construction progress and equipment utilization rate; the management level factor involves the standardization of project management processes and the professional capabilities of personnel; and the market supply and demand factor reflects the impact of industry supply and demand on cost. Other specific influencing factors can be expanded according to project characteristics.

[0137] Credibility assessment employs a combination of the analytic hierarchy process (AHP) and entropy weighting to construct a three-level assessment system:

[0138] The target layer is the credibility of the influencing factors; the criteria layer includes three dimensions: data reliability, timeliness, and relevance; and the indicator layer includes specific indicators such as data collection accuracy, update frequency, and correlation with cost. Among them, data collection accuracy includes the accuracy of the collection equipment and the error of manual recording; update frequency includes the data update cycle and real-time guarantee; and correlation with cost is the feature importance score.

[0139] The rule for setting the weights of the criteria layer is to pass. Pairwise comparison matrix calculations were performed, with data reliability weighted at 0.4, timeliness weighted at 0.3, and relevance weighted at 0.3. These weights were determined based on industry expert scoring statistics, and consistency verification indicators were used. The weight of the indicator layer is set as the ratio of data collection accuracy to data reliability. Human recording errors account for Update frequency accounts for timeliness Real-time assurance accounts for Feature importance score accounts for a significant portion of relevance. ;

[0140] The training steps for the entropy weight method model are as follows:

[0141] Data standardization: Normalizing the evaluation index data of each influencing factor to... interval;

[0142] Calculate the first The first indicator The proportion of each sample:

[0143]

[0144] Calculate the first The entropy value of each indicator:

[0145]

[0146] in, , The number of samples;

[0147] Calculate the first Coefficient of variation for each indicator:

[0148]

[0149] Calculate objective weights:

[0150]

[0151] in, For the number of indicators;

[0152] The subjective weights of the criterion layer and indicator layer, and the cost behavior fingerprint vector, are determined using the analytic hierarchy process. Then, the objective weights of each factor are calculated using the entropy weight method to obtain the cost behavior fingerprint vector. Finally, the weighting coefficient is introduced. , The default value is 0.6, and the setting rule is:

[0153] Early stage of the project The design / procurement phase is the early stage of a project and focuses on subjective experience.

[0154] Later stages of the project The construction / settlement phase is the later stage of the project and focuses on objective data. ;

[0155] The initial values ​​of the dynamic credibility weights are obtained by weighted summation, expressed as follows:

[0156]

[0157] in, Indicates a stage time node, such as the initial moment of the design stage. =0, initial moment of the procurement phase =1, etc., the initial value range of the weights is normalized to... The weight threshold is set to 0.2. Influence factors below this threshold are considered low-confidence factors and are only used as a reference in decision-making.

[0158] This design assesses the credibility of cost-influencing factors at each stage and assigns them dynamic weights. It comprehensively considers multiple influencing factors, adopts a combined weighting method, and combines subjective experience with objective data to make the weight assessment more scientific and comprehensive. By constructing a three-level assessment system and clarifying the weight setting rules for each level, it ensures that the assessment process is standardized and orderly. The weight allocation coefficients are dynamically adjusted according to the project stage, so that the weights can change in real time with the project progress, making them more in line with the actual situation. Assigning dynamic credibility weights can accurately measure the degree of influence of each influencing factor on the cost, providing a reliable reference for subsequent decision-making and improving the scientific and rational nature of the decision.

[0159] In one embodiment, during the engineering phase transition, the credibility weight is attenuated based on data source, timeliness, and environmental changes, including:

[0160] Set credibility decay trigger conditions for engineering phase transitions, such as transitioning from the design phase to the procurement phase, or from the construction phase to the settlement phase, to ensure dynamic adaptation of the weights of influencing factors when making decisions across phases.

[0161] Attenuation coefficient Taking into account three major factors—data source reliability, timeliness decay, and the impact of environmental changes—a multiplicative model is used for calculation, and the expression is:

[0162]

[0163] in, The cost-related behavior fingerprint vector serves as a data source credibility coefficient, with the following rule: official channels. Government cost estimation platforms and industry associations are official channels; authoritative third-party platforms... Unofficial channels The specific figures will be determined by the project management team in conjunction with channel qualification assessments.

[0164] The cost behavior fingerprint vector is a time-decrease coefficient, calculated using an exponential decay model. The calculation formula is as follows:

[0165]

[0166] The rules for the decay rate parameter are set as follows:

[0167] Core Impact Factor =0.03, weight As the core impact factor;

[0168] Important Influence Factors =0.05, It is an important influencing factor;

[0169] General Influence Factors =0.08, weight This is a general impact factor;

[0170] The data duration is the time interval between the current stage and the data generation stage, measured in months.

[0171] The cost behavior fingerprint vector is the environmental change impact coefficient, and the rules are set as follows:

[0172] The steps for constructing an environmental change assessment model, i.e., a logistic regression model, are as follows:

[0173] Feature selection: frequency of policy changes, market price fluctuation range, technology update rate, and degree of change in the geopolitical environment;

[0174] Label settings: 0 for no obvious change, 1 for slight change, and 2 for significant change;

[0175] Model training: Stochastic gradient descent was used, with 500 iterations, a learning rate of 0.01, and a regularization coefficient of 0.001.

[0176] Model Evaluation: Accuracy It is considered qualified;

[0177] The model output is scored, with a score range of 0-10. No significant change results in a positive score. point, slight changes , Significant changes are the score. , =0.5-0.7;

[0178] The decayed confidence weights are updated using the following formula:

[0179]

[0180] in, +1 indicates the time node after the switch, and the updated weights are still normalized to... The weight threshold after attenuation is set to 0.1. Factors with influence below this threshold can be temporarily excluded from the current stage decision-making system. This effectively prevents early decision factors, such as the preliminary estimated factors in the design stage, from having an unreasonable amplified impact on subsequent stages due to decreased timeliness or environmental changes, thus ensuring the scientific nature of cross-stage decision-making.

[0181] This design, during project phase transitions, attenuates the credibility weights based on multiple factors. Phase transitions are set as trigger conditions to ensure dynamic weight adaptation to cross-phase decisions. It comprehensively considers three major factors: data source, timeliness, and environmental changes. A multiplicative model is used to calculate the attenuation coefficient, fully taking into account various factors affecting weight changes. By rationally setting the calculation rules and parameters for each factor, the attenuation process is made more precise and reasonable. Attenuating the weights prevents early decision factors from having unreasonable impacts on subsequent phases due to decreased timeliness and environmental changes, ensuring that cross-phase decisions are based on the latest and most accurate information, thus improving the scientific rigor and effectiveness of decision-making.

[0182] In one embodiment, based on the attenuated confidence weights and combined with the causes of cost deviations at each stage, a cross-stage cost optimization decision scheme is generated, including:

[0183] The causes of cost deviations identified in each stage through the self-explanatory model are classified and grouped according to the associated cost influencing factors, forming a three-dimensional correlation matrix of influencing factors-cause of deviations-stage distribution.

[0184] Using the attenuated credibility weight as the weighting coefficient, the weighted influence of various deviation causes is calculated. Weighted influence = deviation amplitude × corresponding influence factor weight. The deviation causes are sorted in descending order of weighted influence, and the screening threshold is set to Top 3, that is, the three core deviation causes with the highest weighted influence are selected as the optimization focus.

[0185] Based on the root causes of the core deviations, and considering the engineering characteristics and business processes at the corresponding stages, targeted optimization measures are generated:

[0186] For deviations related to material price fluctuations, formulate dynamic procurement strategies and establish a price early warning mechanism;

[0187] For deviations related to construction efficiency, optimize the construction organization plan and improve equipment utilization.

[0188] For deviations related to policy adjustments, strengthen policy tracking and interpretation, and adjust cost planning in advance;

[0189] For deviations related to management level, improve management processes and strengthen staff training;

[0190] For market supply and demand discrepancies, optimize resource allocation and expand supply channels;

[0191] The steps for constructing a cross-stage optimization decision prioritization model, also known as a multi-dimensional scoring model, are as follows:

[0192] Sample construction: Collect historical data on implementation costs, expected benefits, and feasibility of optimization measures, and mark them with comprehensive priorities, which are divided into 1-5 levels;

[0193] Feature engineering: quantifies implementation costs, expected benefits, and feasibility into numerical features ranging from 0 to 10.

[0194] Model training: A weighted linear regression model was used. ,in The implementation cost is calculated by subtracting the implementation cost score from the actual cost score. Score for expected benefits, Feasibility score;

[0195] Parameter optimization: The weights are fitted using the least squares method. The dimensional weights are set as follows: implementation cost weight 0.3, expected benefit weight 0.4, and feasibility weight 0.3, based on the cost-benefit analysis principle. The fitting error threshold is set to 1e-4.

[0196] Calculate the overall score: Overall score = Implementation cost weight × (10 - Implementation cost score) + Expected benefit weight × Expected benefit score + Feasibility weight × Feasibility score;

[0197] The overall score threshold is set at 6 points; optimization measures that fall below this threshold need to be readjusted.

[0198] Based on the comprehensive score, the optimization measures are sorted in descending order, and the sorted cross-stage cost optimization decision-making scheme is output. The implementing entities, implementation steps, time nodes, expected goals and risk control points of each measure are clearly defined, forming a closed-loop management and control system to achieve intelligent optimization of the entire life cycle of project cost.

[0199] This design generates cross-stage cost optimization decision-making schemes based on attenuated weights and the causes of cost deviations. By constructing a three-dimensional correlation matrix, the relationship between influencing factors, deviation causes, and stages is clearly presented. The weighted influence degree is calculated using attenuated weights as weighting coefficients to screen core deviation causes, making the optimization focus clearer. Targeted optimization measures are generated for different core deviation causes to improve the effectiveness of the measures. A cross-stage optimization decision priority ranking model is constructed, which comprehensively considers implementation costs, expected benefits, and feasibility to provide a scientific basis for ranking optimization measures. Detailed decision-making schemes are output, forming a closed-loop management and control system to achieve intelligent optimization of the entire life cycle of project costs and improve project economic benefits.

[0200] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0201] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A smart management and optimization system for the entire lifecycle of engineering cost, characterized in that, The system includes: The module includes: data acquisition and preprocessing module, feature extraction and fingerprint construction module, cost deviation analysis and positioning module, impact factor evaluation and weight assignment module, weight dynamic decay processing module, and cross-stage optimization decision generation module. The data acquisition and preprocessing module is used to acquire cost data and multi-source heterogeneous correlation data at each stage of the entire life cycle of engineering cost, and to perform data cleaning and standardization preprocessing. The feature extraction and fingerprint construction module is used to extract time-series behavioral features from the preprocessed data of each stage based on the engineering stage division results, and construct the cost behavior fingerprint vector of the corresponding stage. The cost deviation analysis and location module is used to automatically generate an interpretable cost deviation cause model using the cost behavior fingerprint vector, thereby locating the source of cost anomalies. The impact factor assessment and weight assignment module is used to assess the credibility of cost impact factors formed at each stage and assign dynamic credibility weights. The dynamic weight decay processing module is used to decay the credibility weight according to the data source, timeliness and environmental changes when switching engineering stages. The cross-stage optimization decision generation module is used to generate a cross-stage cost optimization decision scheme based on the decayed credibility weight and the causes of cost deviations in each stage.

2. The intelligent management and optimization system for the entire lifecycle of engineering cost as described in claim 1, characterized in that, The data acquisition and preprocessing module includes: Data partitioning unit, data cleaning unit, format unification unit, and standardization processing unit; The data segmentation unit is used to divide the entire life cycle of project cost into multiple core stages, and to divide the data of each stage according to time nodes and business processes, so as to ensure strong correlation between data and stages. The data cleaning unit is used to process missing values ​​using interpolation filling and to remove outliers using statistical identification. The format unification unit is used to standardize date formats, numerical units, and encoding rules according to a unified standard. The standardization processing unit is used to convert data of different magnitudes and dimensions into a unified scale using a normalization algorithm to meet the needs of subsequent data processing.

3. The intelligent management and optimization system for the entire lifecycle of engineering cost as described in claim 2, characterized in that, The cost data collected by the data acquisition and preprocessing module includes: Data related to labor costs, material and equipment prices, construction process costs, and bill of quantities; The multi-source heterogeneous correlation data includes policy and regulatory data, market data, geological and environmental data, and project management data. The data cleaning unit uses an interpolation filling method based on the nearest neighbor algorithm, and the outlier identification method is a statistical identification method based on quantiles. The normalization algorithm used by the standardization processing unit is Z-score normalization.

4. The intelligent management and optimization system for the entire lifecycle of engineering cost as described in claim 1, characterized in that, The feature extraction and fingerprint construction module includes: Feature extraction unit, feature weight setting unit, redundant feature removal unit, and fingerprint vector construction unit; The feature extraction unit is used to extract core temporal behavioral features based on the temporal characteristics of data at each stage, including trend features, fluctuation features, mutation features, and correlation features. The feature weight setting unit is used to calculate the weight of each feature based on its contribution to the cost deviation. The redundant feature removal unit is used to identify and remove redundant features through statistical testing. The fingerprint vector construction unit is used to fuse the selected features in terms of dimensions to construct a cost behavior fingerprint vector that condenses the core features of cost behavior at each stage.

5. The intelligent management and optimization system for the entire life cycle of engineering cost as described in claim 4, characterized in that: The feature extraction unit calculates trend features using a linear regression correlation algorithm, calculates fluctuation features using a moving average statistical algorithm, identifies abrupt change features using a residual test correlation algorithm, and calculates association features using a correlation analysis algorithm. The feature weight setting unit uses a feature importance evaluation algorithm to calculate feature weights and set weight thresholds. Features below the threshold are considered minor features. The redundant feature removal unit uses the variance test method to identify redundant features, and the fingerprint vector construction unit normalizes the features to a uniform numerical range to form a fingerprint vector with fixed dimensions.

6. The intelligent management and optimization system for the entire lifecycle of engineering cost as described in claim 1, characterized in that, The cost deviation analysis and location module includes: Model building unit, model training unit, core feature selection unit, causal rule generation unit, and anomaly localization unit; The model building unit is used to construct a self-explanatory model that balances prediction accuracy and interpretability by taking cost behavior fingerprint vectors as input and cost deviation related indicators as output labels. The model training unit is used to complete model construction through dataset partitioning, hyperparameter optimization, iterative training, and model evaluation. The core feature screening unit is used to screen core features that have a significant impact on cost deviation through an information gain correlation algorithm. The causal rule generation unit is used to generate a structured set of cost deviation causal rules based on model-based decision logic; The anomaly localization unit is used to compare the real-time cost behavior fingerprint vector with the rule set through a rule matching mechanism, locate the specific source of the cost anomaly, and output an explanation report.

7. The intelligent management and optimization system for the entire lifecycle of engineering cost as described in claim 1, characterized in that, The impact factor evaluation and weight assignment module includes: The system comprises an impact factor classification unit, an evaluation system construction unit, a subjective weight calculation unit, an objective weight calculation unit, and a dynamic weight fusion unit. The influencing factor classification unit is used to divide the cost influencing factors into material price fluctuation factors, policy adjustment factors, construction efficiency factors, management level factors, and market supply and demand factors. The evaluation system construction unit is used to construct a three-level credibility evaluation system that includes a target layer, a criterion layer, and an indicator layer. The subjective weight calculation unit is used to calculate the subjective weight of each influencing factor using the analytic hierarchy process (AHP) algorithm. The objective weight calculation unit is used to calculate the objective weight of each influencing factor using the entropy weight method; The dynamic weight fusion unit is used to introduce weight allocation coefficients and fuse subjective weights and objective weights through a weighted summation method to obtain the initial value of the dynamic credibility weight of each influencing factor.

8. The intelligent management and optimization system for the entire lifecycle of engineering cost as described in claim 7, characterized in that, The criteria layer of the evaluation system includes: Data reliability, timeliness, and relevance; the indicator layer includes data collection accuracy, update frequency, and correlation with cost. The weighting coefficients are dynamically adjusted according to the project stage. In the early stage of the project, subjective weights are emphasized, while in the later stage, objective weights are emphasized. The dynamic weight fusion unit also sets a weight threshold. Influence factors below the threshold are considered as low-credibility factors and are only used as a decision-making reference.

9. The intelligent management and optimization system for the entire life cycle of engineering cost as described in claim 1, characterized in that, The weight dynamic decay processing module includes: Attenuation triggering unit, attenuation coefficient calculation unit, and weight update unit; The attenuation triggering unit is used to set the engineering stage switching as the triggering condition for weight attenuation; The attenuation coefficient calculation unit is used to calculate the attenuation coefficient by comprehensively considering factors such as the reliability of the data source, the attenuation of data timeliness, and the influence of environmental changes, and by using a multiplicative model. The weight update unit is used to update and adjust the credibility weight of each influencing factor according to the calculated attenuation coefficient. The updated weight is still normalized to a preset value range, and a weight threshold after attenuation is set. Influencing factors below the threshold can be temporarily excluded from the current stage decision-making system.

10. The intelligent management and optimization system for the entire lifecycle of engineering cost as described in claim 1, characterized in that, The cross-stage optimization decision generation module includes: The system includes a deviation cause collection unit, a core deviation screening unit, an optimization measure generation unit, a decision priority ranking unit, and a scheme output unit. The deviation cause collection unit is used to classify the cost deviation causes identified at each stage according to the associated influencing factors, forming a multidimensional correlation matrix. The core deviation screening unit is used to calculate the weighted influence of various deviation causes using the attenuated confidence weight as a weighting coefficient, and to screen the core deviation causes. The optimization measure generation unit is used to generate targeted optimization measures based on the core causes of deviations and the characteristics and business processes of each stage of the project. The decision priority ranking unit is used to construct a multi-dimensional scoring model to comprehensively score and prioritize optimization measures. The output unit is used to output the sorted cross-stage cost optimization decision-making scheme, clarify the implementation elements and risk control points of each measure, and form a closed-loop management and control system.

Images