Engineering project cost evaluation system based on data dynamic prediction

Abstract

The application discloses an engineering project cost evaluation system based on data dynamic prediction, relates to the technical field of intelligent engineering cost evaluation, and comprises data collection, benchmark generation, factor collection and cost synthesis modules. The data collection module extracts precise and adaptive cost composition elements from a historical database by constructing an engineering cost feature correlation graph and adopting a dynamic similarity weighting method. The benchmark generation module generates a dynamic cost benchmark of a subproject which can reflect market fluctuation trends in real time based on time series analysis and a market factor correction algorithm. The system comprehensively combines the dynamic benchmark and real-time collected cost influence factors, and finally outputs a real-time dynamic total cost evaluation value of the project through cost superposition and risk integration. The system effectively solves the problems of rough historical data utilization and static lag of cost benchmarks, and realizes the dynamic and accurate evaluation of engineering costs.

Description

Technical Field

[0001] This invention belongs to the field of intelligent engineering cost assessment technology, specifically an engineering project cost assessment system based on dynamic data prediction. Background Technology

[0002] Traditional engineering project cost assessments often rely on static historical databases and fixed pricing quotas. At the data application level, these methods typically directly call upon the overall cost of historical projects or the average cost of similar projects, lacking in-depth mining and correlation analysis of the fine structure within massive amounts of historical cost data. The matching between historical data and the project to be assessed is crude, failing to dynamically adjust the weight and selection range of reference data according to the specific characteristics of the project, resulting in insufficient relevance of the assessment basis and biased prediction starting points.

[0003] Existing technical solutions generally use periodically published market information prices or adjustment coefficients to update cost benchmarks, which is essentially a periodic static correction. When projects are large-scale and complex, such as in the cost assessment and management of ultra-large span double-layer steel truss cross-sea channels, the use of ultra-large span double-layer steel truss girders to improve traffic capacity greatly increases the self-weight of the bridge structure and the difficulty of load control. This makes it difficult for traditional cost control methods to achieve accurate review, and conventional cost control measures fail, resulting in significant challenges in cost control.

[0004] Furthermore, traditional methods cannot capture the intrinsic changing trends of cost elements over time in real time, nor can they respond sensitively to the immediate impact of external factors such as policy adjustments and market supply and demand fluctuations. This results in a lag and disconnect between cost benchmarks and the dynamic changes in the real market environment, making it difficult for assessment results to accurately reflect the real-time cost situation during project execution.

[0005] This invention aims to solve two core problems: how to intelligently extract highly relevant cost components from unstructured historical data, and how to generate a dynamic cost benchmark that integrates the accuracy of indicators, quota supplementation, and process settlement in the cost control of ultra-large-scale complex projects, and can automatically evolve with time and market environment. Summary of the Invention

[0006] This invention aims to solve at least one of the technical problems existing in the prior art; To this end, the present invention proposes an engineering project cost assessment system based on dynamic data prediction, comprising: The data acquisition module is used to collect the historical cost database of the target project, perform structured parsing on the historical cost database to form a cost feature association map, and use a dynamic similarity weighted cost data extraction method to extract the corresponding historical cost component data from the historical cost database to form an initial cost reference data pool. The benchmark generation module is used to input the initial cost reference data pool into the dynamic cost benchmark generation model for processing. The dynamic cost benchmark generation model is based on time series analysis and market factor correction algorithm to generate dynamic cost benchmarks for sub-projects applicable to the project to be evaluated. The factor acquisition module is used to collect dynamic cost impact factor data in the current market environment in real time. The cost synthesis module is used to input the dynamic cost benchmark of the sub-project and the dynamic cost impact factor data into the real-time cost synthesis engine. The real-time cost synthesis engine performs cost superposition and risk coefficient integration item by item to calculate and generate the real-time dynamic total cost assessment value of the project to be evaluated.

[0007] Furthermore, the formation of the initial cost reference data pool includes: Collect the historical cost database of the target project. The historical cost database contains a complete cost list of completed projects and their corresponding project feature information. The complete cost list in the historical cost database is structured and parsed to extract the cost components of each sub-item project. A mapping relationship is established between the cost components of each sub-item project and its corresponding project feature information to form a project cost feature association map. Obtain the initial project feature set of the project to be evaluated, perform pattern matching between the initial project feature set of the project to be evaluated and the project cost feature association map, and identify the set of historical sub-projects similar to the project to be evaluated. Based on the aforementioned set of similar historical sub-projects, a dynamic similarity-weighted cost data extraction method is used to extract the corresponding historical cost component data from the historical cost database to form an initial cost reference data pool.

[0008] Furthermore, the step of performing structured parsing of the complete cost list in the historical cost database, extracting the cost components of each sub-item, and establishing a mapping relationship between the cost components of each sub-item and its corresponding project feature information to form a project cost feature association map includes: According to the standard bill of quantities specifications, the complete cost list is broken down to the level of individual items. For each sub-item of the project, its direct cost composition is further analyzed. The direct cost composition includes labor consumption, material consumption, machine shift consumption and their corresponding historical unit price data. Geographic environment attributes, structural type attributes, building scale attributes, and construction technology complexity attributes are extracted from the feature information of the engineering project; Establish a multi-dimensional statistical correlation model between the labor consumption, material consumption, and machine shift consumption in the direct cost composition of each sub-item of the project and the geographical environment attribute, structural type attribute, building scale attribute, and construction technology complexity attribute, respectively. Based on the multi-dimensional statistical correlation model of all sub-items of the project, the project cost feature correlation graph is constructed. The project cost feature correlation graph includes project feature attribute nodes and sub-item project nodes. Sub-item project nodes are connected to related project feature attribute nodes through edges, and the weight of the edges is set based on the corresponding correlation strength.

[0009] Further, an initial project feature set of the project to be evaluated is obtained, and pattern matching is performed between the initial project feature set of the project to be evaluated and the project cost feature association map to identify a set of historical sub-projects similar to the project to be evaluated, including: The design drawings and technical specifications of the project to be evaluated are analyzed, and the geographical environment characteristics, structural type characteristics, building scale characteristics and technical requirements characteristics of the project to be evaluated are extracted as the initial project feature set. The features in the initial project feature set are transformed into feature vectors that can be recognized by the engineering cost feature association map; In the engineering cost feature association map, the multidimensional similarity between the feature vector of the project to be evaluated and the feature vector of each historical sub-project node in the map is calculated. Set a multidimensional similarity threshold, filter out all historical sub-project nodes whose multidimensional similarity exceeds the threshold, and classify them into the set of similar historical sub-projects.

[0010] Furthermore, based on the aforementioned set of similar historical sub-projects, a dynamic similarity-weighted cost data extraction method is used to extract corresponding historical cost component data from the historical cost database, forming an initial cost reference data pool, including: For each historical sub-project in the set of similar historical sub-projects, its weight coefficient is calculated based on its multidimensional similarity with the project to be evaluated. According to the aforementioned weighting coefficients, the historical unit price data corresponding to each historical sub-project is weighted; The weighted historical unit price data is then categorized and summarized according to their corresponding labor, material, and machinery categories. Outlier removal and statistical smoothing are performed on the aggregated unit price data for each category to generate a reference unit price sequence divided by category, which constitutes the initial cost reference data pool.

[0011] Further, the initial cost reference data pool is input into the dynamic cost benchmark generation model for processing. The dynamic cost benchmark generation model, based on time series analysis and a market factor correction algorithm, generates dynamic cost benchmarks for the sub-items of the project to be evaluated, including: For each category of reference unit price sequence in the initial cost reference data pool, a time series prediction model is established to predict the future unit price change trend line within the planned construction period of the project to be evaluated; Acquire macro market index data, which includes industry inflation rate forecasts and regional economic development indicators. The future unit price change trend line is fitted and calibrated with the macro market index data to generate a calibrated future benchmark unit price curve. Based on the estimated quantities of the project to be evaluated, the calibrated future benchmark unit price curve is converted into a comprehensive unit price for each sub-item, which is the dynamic cost benchmark for the sub-item.

[0012] Furthermore, real-time data on dynamic cost influencing factors under the current market environment is collected, including: The dynamic cost influencing factor data includes material price fluctuation data, labor cost index, and equipment rental rate change data; An application programming interface that connects to multiple external data sources, including commodity trading platforms, human resources service agencies, and equipment supplier information systems; The application programming interface of the commodity trading platform periodically obtains spot and futures price data of major building materials; The human resources service agency periodically obtains wage guidance prices and social security contribution rate data for different job types in different regions from its application programming interface. Real-time rental quotes and maintenance cost data for various types of construction machinery are periodically obtained from the application programming interface of the equipment supplier information system. The obtained spot and futures price data, wage guidance price and social security contribution rate data, real-time rental quotation and maintenance cost data are cleaned and standardized to form the structured dynamic cost influencing factor data.

[0013] Furthermore, the dynamic cost benchmark of the sub-projects and the dynamic cost impact factor data are input into the real-time cost synthesis engine. The real-time cost synthesis engine then performs cost aggregation and risk coefficient integration item by item to calculate and generate the real-time dynamic total cost assessment value of the project to be evaluated, including: The dynamic cost benchmark of each sub-item project is compared with the latest dynamic cost impact factor data corresponding to the main materials, labor and machinery required for the sub-item project. When the deviation between the dynamic cost impact factor data of a certain category and the predicted value used when generating the dynamic cost benchmark of the sub-item exceeds the preset fluctuation threshold, the original predicted value is replaced with the latest dynamic cost impact factor data, and the comprehensive unit price of the sub-item is recalculated in real time. Real-time recalculation is performed on all sub-projects of the project to be evaluated, and the updated direct engineering costs are obtained by summarizing them. Based on the updated direct engineering costs, indirect costs, profits, and taxes are calculated according to preset rates; The updated direct engineering costs, indirect costs, profits, and taxes are added together to obtain the real-time dynamic total cost assessment value.

[0014] Furthermore, the real-time cost synthesis engine performs cost aggregation and risk coefficient integration item by item, and also includes risk assessment and integration steps: Analyze the historical volatility and current volatility of each element in the dynamic cost influencing factor data; Identify cost elements whose volatility exceeds the risk threshold and mark them as high-risk cost elements; For each high-risk cost element, calculate the maximum additional cost increment it causes to the corresponding sub-item cost based on its historical maximum fluctuation range; The maximum additional cost increments calculated from all sub-projects are aggregated according to a probability distribution model to calculate the overall cost risk reserve value for the project to be evaluated.

[0015] Furthermore, the system also includes: The report output module outputs the real-time dynamic total cost assessment value of the project to be evaluated and the dynamic cost benchmarks of all sub-projects as a structured cost assessment report, including: Fill the preset report template with the real-time dynamic total cost assessment value, the dynamic cost benchmark of each sub-project, the list of identified high-risk cost elements, and the overall cost risk reserve value. The report template displays historical trends, current values, and forecasts of major material prices in chart form. The report template generates a project cash flow forecast table that scrolls over time. The cash flow forecast table is generated based on the dynamic cost benchmark of the sub-project and the project schedule. The generated complete structured cost assessment report will be converted into a standard document format for storage and output.

[0016] Compared with the prior art, the beneficial effects of the present invention are: By constructing a correlation map of engineering cost characteristics and performing structured analysis on historical data, a more explicit expression of the multidimensional and complex relationships within cost data is achieved. Based on this, a dynamic similarity-weighted data extraction method is employed to calculate the matching weight between the target project and historical data items in real time, according to the specific characteristics of the target project. This ensures that the reference data extracted from the historical database is no longer a simple classification or averaging result, but rather a data set highly adapted to the current evaluation object, accurate down to the element level. The precision and relevance of historical data utilization are significantly improved, laying a more accurate and reliable data foundation for subsequent evaluations.

[0017] By applying time series analysis algorithms to process historical cost data, it is possible to effectively identify and fit the long-term trends, seasonal cycles, and other temporal patterns inherent in the cost data. Combined with market factor correction algorithms, dynamic variables such as macroeconomic indicators and material price indices are introduced into the model in real time as correction parameters. This technical solution ensures that the generated dynamic cost benchmarks for each sub-item of the project not only inherit historical statistical patterns but also adaptively adjust to real-time changes in the external market environment. The cost benchmarks thus possess foresight and environmental sensitivity, enabling a more realistic simulation of cost changes during future project execution, overcoming the lag problem caused by static benchmarks or periodic revisions. Attached Figure Description

[0018] Figure 1 This is a timeline diagram of the engineering project cost assessment system based on dynamic data prediction as described in this invention. Figure 2 This is a flowchart illustrating the process of creating a correlation map of engineering cost characteristics in the data-driven dynamic prediction-based engineering project cost assessment system described in this invention. Figure 3 This is a graph showing the weight coefficients of historical sub-projects under different weighting strategies in the data-driven dynamic prediction-based engineering project cost assessment system of the present invention. Figure 4 This is a diagram illustrating the cost risk assessment and reserve allocation for engineering projects in the data-driven dynamic prediction-based engineering project cost assessment system described in this invention. Figure 5 This is a chart showing the price trend analysis of the main materials in the data-driven dynamic prediction-based engineering project cost assessment system described in this invention. Detailed Implementation

[0019] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] See Figure 1 The overall implementation scheme of the engineering project cost assessment system based on data dynamic prediction is as follows.

[0021] When the system is in operation, the data acquisition module first starts to collect historical cost data from the target project database. This database contains detailed cost information for multiple completed projects. The data acquisition module performs structured parsing of the historical cost database, identifies and extracts the basic elements that constitute the project cost, and analyzes the inherent relationships between these elements and the characteristics of various projects, forming a cost characteristic correlation map that reflects the relationship between cost composition and project characteristics.

[0022] Subsequently, the data acquisition module adopts a dynamic similarity-weighted cost data extraction method. Using the generated graph, it intelligently filters out historical cost components that are highly related to the current project to be evaluated from massive historical data. These extracted data are aggregated to form an initial cost reference data pool specifically for the project to be evaluated.

[0023] The benchmark generation module receives the initial cost reference data pool and inputs it into an embedded dynamic cost benchmark generation model. The core of this model is based on time series analysis and market factor correction algorithms. It analyzes the patterns of price changes over time in the data pool and, combined with judgments on future macroeconomic factors, performs calculations and corrections, ultimately outputting a series of forward-looking dynamic cost benchmarks applicable to each component of the project to be evaluated.

[0024] The factor acquisition module works independently and continuously. It connects to external data sources to collect dynamic cost impact factor data that reflects the fluctuations in the current market environment in real time. This data includes, but is not limited to, real-time changes in material, labor, and machinery costs.

[0025] As the system's decision-making terminal, the cost synthesis module synchronously acquires the dynamic cost benchmarks for each sub-project output by the benchmark generation module and the dynamic cost impact factor data provided by the factor acquisition module. The built-in real-time cost synthesis engine of the cost synthesis module merges, compares, and calculates these two types of data, verifying and updating the cost of each sub-project item by item, and integrating risk factors. Through a systematic cost overlay and integration process, it ultimately calculates and generates a real-time dynamic total cost assessment value for the project to be evaluated, reflecting the latest market conditions.

[0026] In one embodiment of the present invention, the process by which the data acquisition module forms an initial cost reference data pool is as follows.

[0027] Collect the historical cost database of the target project. This historical cost database contains complete cost lists for multiple completed projects and their corresponding project characteristic information, such as project location, structural form, building area, and construction period.

[0028] This module can directly recognize various formats of pricing result files (non-standard Excel, XML, PDF, JPG, JPEG, PNG, etc.) and parse the data within them; Non-standard Excel files, such as those from Glodon, Swell, Yida, Tongwang, and Zongheng, are the most common and widely used non-standard pricing documents in the construction industry in China. Due to the high flexibility and complex formulas involved in manual compilation, they are used by each company. However, in project management, they have always been a challenge for information management of construction data.

[0029] This invention relies on deep neural network (CNN) natural language processing technology, developed over many years, for intelligent recognition and extraction. It extracts data from tables to be parsed, obtaining a dataset to be parsed. A machine learning model is then used to classify the table headers of this dataset, resulting in several header-classified datasets. Furthermore, the machine learning model is used to classify the columns of each header-classified dataset, encapsulating them into column-classified datasets. These column-classified datasets are then assembled to obtain a standard JSON data structure, which is then returned. The machine learning model can identify, classify, and clean irregular tables and data based on headers, column data, and cell data, extracting a standardized JSON data structure. Compared to traditional manual classification methods or manually written fixed classification models, the machine learning model possesses extremely high intelligence and automation. Moreover, as the amount of data processed increases, the classification accuracy and processing efficiency of the machine learning model will continue to improve.

[0030] S11: Perform structured parsing of the complete cost list in the historical cost database. This parsing process aims to extract the cost components of each sub-item of the list; for example, the quantity of labor, materials, and machinery consumed in a certain sub-item and their historical unit prices. Establish a mapping relationship between the cost components of each sub-item and its corresponding project characteristic information. This mapping relationship reveals the typical performance of cost elements under different project characteristics. All mapping relationships together constitute a networked engineering cost characteristic association map.

[0031] When an engineering project needs to be evaluated, its initial project feature set is obtained. This feature set originates from the project's early design documents and plans. Pattern matching is performed between this initial feature set and a pre-constructed engineering cost feature association map. By calculating feature similarity, the system can identify a set of historical sub-projects that are similar in features to the project under evaluation. Based on the identified set of similar historical sub-projects, a dynamic similarity-weighted cost data extraction method is used. This method assigns different weights to the data of each historical sub-project according to the degree of similarity, extracting corresponding historical cost component data from the historical cost database. After weighted aggregation, an initial cost reference data pool is formed.

[0032] For example, a complex and large-scale project may require research covering multiple engineering fields such as roads, bridges, interchanges, and service areas, comprehensively collecting and analyzing the construction technology, procedures, consumption quantities, and unit prices of each part of the project. The cost assessment object not only includes the research on existing relevant cost standards, but also involves multiple aspects such as project construction technology procedures, quota item division, collection and research on labor, material, and machinery consumption quantities, research on construction material cost quotas, research on machinery shift cost quotas, calculation of quota standard results, and compilation of research results, thus forming a comprehensive and systematic research system.

[0033] This invention, through detailed research and analysis of construction technology and procedures, clarifies the technological content, sequence of procedures, construction conditions and requirements, and engineering characteristics of each task, laying the foundation for accurately classifying quota items and determining consumption.

[0034] In the study of quota item division, based on the research results of process procedures, the names of quota items to be formulated, quota units, and corresponding work contents of each item are clarified. The results are then compared and analyzed with existing cost standards to ensure the rationality and accuracy of the item division, so as to meet the needs of detailed accounting of engineering projects in cost management.

[0035] In the collection and research of labor, material, and machinery consumption, detailed data collection forms were developed. Through on-site observation and drawing analysis, relevant data such as the model specifications, performance, and consumption of labor, materials, and machinery were recorded and organized. For the research on material cost quotas, invoices, receipts, contracts, and other documents were collected to study the names, specifications, performance parameters, ex-factory prices, taxes, freight, and transportation loss rates of various construction materials. For some materials, a traceability approach was used to analyze the price composition to accurately determine current prices. The research on machinery shift cost quotas investigated the use of automated and specialized machinery in the project to comprehensively determine the relevant cost parameters for newly added machinery.

[0036] The above structured standard analysis enables the quota standards to accurately reflect the resource consumption and cost composition during the project construction process, thus enabling refined management of project cost control, cost accounting, bidding and quotation.

[0037] In practical implementation, a planned prestressed concrete continuous beam bridge is used as an example to illustrate the project to be evaluated. When the data acquisition module operates, the historical cost database collected contains complete cost lists for multiple completed bridge projects and some high-rise buildings, along with their corresponding project characteristic information. The complete cost lists specify the quantities and total prices of each sub-item, while the project characteristic information records the project location, bridge span, structural form, construction year, and construction technical standards. A structured analysis of the complete historical cost list of a completed cable-stayed bridge is performed to extract the cost components of the sub-item "C50 concrete pouring for the main tower." These cost components include the consumption of 525 cement, medium sand, crushed stone, and the mechanical shift consumption of the concrete pump truck, as well as their historical unit price data during construction in 2019. For each sub-item project (e.g., "C50 concrete pouring for the main tower") in the historical cost database, a multi-dimensional statistical correlation model is established between its comprehensive unit price or unit quantity cost and a set of suitable project characteristic attributes (e.g., geographical environment, structural type, building scale, and construction technology complexity). This model quantifies the contribution or influence coefficient of each characteristic attribute on the cost through methods such as multiple regression analysis. By integrating the statistical correlation models of all sub-items, a cost characteristic correlation graph is constructed. In this graph, nodes are divided into two categories: project characteristic attribute nodes and sub-item project nodes. Edges connect sub-item project nodes to their related project characteristic attribute nodes, with the edge weights assigned by the corresponding influence coefficients or correlation strengths calculated in the statistical correlation model. This graph depicts the inherent patterns of how different engineering characteristics jointly influence the cost of sub-items.

[0038] S12: Obtain the initial project feature set of the prestressed concrete continuous beam bridge to be evaluated. The initial project feature set is extracted by parsing the design drawings and technical specifications of the bridge. The extracted content includes the geographical environment feature of mountainous valley, the structural type feature of prestressed concrete continuous beam, the construction scale feature, and the technical requirement feature of cantilever casting construction.

[0039] The text features in the initial project feature set are transformed into numerical feature vectors. The feature vectors are represented by multi-dimensional arrays, and each dimension of the array corresponds to a feature attribute and its quantified value defined in the engineering cost feature association map.

[0040] In selecting features, it is necessary to adhere to the industry standard system. The study follows the relevant regulations of the ministry-issued quotas in terms of on-site measurement methods, basic data collection forms, calculation methods, units, and determination of amplitude differences, and ensures that the various specifications are coordinated, complementary, and systematically compatible.

[0041] For example, properly handling the relationship with the "Highway Engineering Budget Quota", "Highway Engineering Estimate Quota", "Highway Engineering Construction Project Estimate and Budget Compilation Method", "Highway Engineering Construction Quota Measurement and Compilation Procedure" (JTG / Tb3811-2020) and other relevant cost standards and quotas in the highway industry requires compliance with the top-level design of the highway industry standard and specification system.

[0042] In the engineering cost feature association map, the multidimensional similarity between the feature vector of the project to be evaluated and the feature vector of the historical sub-project node "concrete pouring of the arch ring of a certain arch bridge" stored in the map is calculated. The calculation simultaneously considers the matching degree of material properties, structural complexity, and environmental conditions. The multidimensional similarity threshold is set to 0.7. The system filters out all historical sub-project nodes with a multidimensional similarity exceeding 0.7. For example, nodes such as "concrete pouring of a certain bridge pier" and "high-strength concrete pouring of a certain building core tube" are classified into a set of similar historical sub-projects.

[0043] In some embodiments, the multidimensional similarity is calculated using the cosine of the vector angle, and its expression is: Where: symbol Represents the feature vector of the project to be evaluated, with the symbol... The first element in the correlation graph of engineering cost characteristics The feature vector of each historical sub-project node, symbol The multidimensional similarity value is calculated, and the symbol is... Identifiers and symbols representing the engineering projects to be evaluated. This represents the identifier of a historical sub-project node. It's understandable that identifying similar sets of historical sub-projects relies entirely on the quantization calculation of feature vectors and threshold comparison.

[0044] S13: Based on a set of similar historical sub-projects, a dynamic similarity-weighted cost data extraction method is used. For the historical sub-project "concrete pouring of a bridge pier" in the set, its multidimensional similarity with the project to be evaluated is calculated to be 0.85, and its weight coefficient is calculated to be 0.85. Using a weight coefficient of 0.85, the historical unit price data corresponding to the historical sub-project "concrete pouring of a bridge pier" is weighted, and its historical unit price of C30 concrete (300 yuan / cubic meter) becomes 255 yuan / cubic meter after weighting. The historical unit price data of concrete, steel reinforcement, and formwork installation labor for all historical sub-projects after weighting are classified and summarized to form multiple price data sets such as concrete, steel reinforcement, and labor. The aggregated concrete unit price data set undergoes outlier removal and statistical smoothing. Data points with unit prices significantly deviating from the group mean are removed, and the remaining data are smoothed using a moving average method to generate a reference unit price sequence for concrete. This reference unit price sequence, along with the processed reference unit price sequences for steel reinforcement and labor, constitutes the initial cost reference data pool. In practice, the initial cost reference data pool exists in the form of a structured data table, which includes fields for material category, reference unit price, and timestamp of price data source.

[0045] It is understandable that the dynamic similarity-weighted cost data extraction method assigns higher weight to historical project data that is more similar to the current project's characteristics when forming the initial cost reference data pool. Optionally, the weight coefficients can be calculated by directly using normalized multidimensional similarity values, or by using a step function assignment method based on similarity ranking. In some embodiments, outlier removal uses a box plot statistical method, considering data points in the price data that exceed 1.5 times the interquartile range of the upper quartile or fall below 1.5 times the interquartile range of the lower quartile as outliers. Optionally, in addition to the moving average method, the statistical smoothing process can also use the kernel density estimation method to smooth the price data distribution. In specific implementation, the initial cost reference data pool finally output by the data acquisition module serves as the sole data input source for the subsequent dynamic cost benchmark generation model.

[0046] See Figure 2 In one embodiment of the present invention, a structured parsing of the complete cost list in the historical cost database is performed, specifically including: S111: In accordance with the standard bill of quantities specifications, the complete cost list is broken down level by level to the level of individual items.

[0047] S112: For each sub-item in the list, further analyze its direct cost composition. The direct cost composition includes the labor consumption, material consumption, and machine shift consumption of the sub-item, as well as the corresponding historical unit price data in historical projects.

[0048] S113: At the same time, key attributes are extracted from the project feature information that is associated with the historical cost list, including geographical environment attributes, structural type attributes, building scale attributes, and construction technology complexity attributes.

[0049] S114: Establish a correlation model. For each sub-item of the project, analyze the statistical relationship between the labor consumption, material consumption, and machinery shift consumption in its direct cost composition and the geographical environment attribute, structural type attribute, building scale attribute, and construction technology complexity attribute, respectively, to form a multi-dimensional statistical correlation model.

[0050] S115: Based on the multi-dimensional statistical correlation model of all sub-items of the project, construct the project cost feature correlation graph. The project cost feature correlation graph is a heterogeneous information network containing two types of nodes: project feature attribute nodes and sub-item project nodes. Sub-item project nodes are connected to their related project feature attribute nodes through edges, and the weight of the edges is determined by the influence coefficient of the corresponding attribute in the statistical correlation model of that sub-item project.

[0051] The engineering project cost assessment system based on dynamic data prediction of this invention obtains an initial project feature set of the engineering project to be assessed, performs pattern matching between the initial project feature set of the engineering project to be assessed and the engineering cost feature association map, and identifies a set of historical sub-projects similar to the engineering project to be assessed, specifically including: S121: By analyzing the design drawings and technical specifications of the project to be evaluated, the geographical environment features, structural type features, building scale features, and technical requirement features contained in the project are extracted. The set of these features constitutes the initial project feature set. The textual or category features in the initial project feature set are transformed into feature vectors that can be mathematically calculated in the engineering cost feature association map. After the transformation, the multidimensional similarity between the feature vector of the project to be evaluated and the feature vectors of each historical sub-project node pre-stored in the map is calculated. S124: A multidimensional similarity threshold is set. The system automatically filters out all historical sub-project nodes whose multidimensional similarity exceeds the threshold and groups these nodes into a set, namely the set of historical sub-projects similar to the project to be evaluated.

[0052] In the specific implementation, a planned comprehensive stadium project is used as an example to illustrate the project to be evaluated. The stadium includes a large-span reticulated shell roof and a multi-story concrete grandstand structure. A complete cost list of a completed stadium project in the historical cost database is structured and analyzed. The analysis process strictly follows the "Construction Engineering Quantity List Pricing Specification". The complete cost list is broken down to the level of sub-items, such as "high-altitude bulk assembly of reticulated shell steel structure", "prestressed concrete pouring of C40 grandstand", and "installation of precast grandstand panels". For the sub-item "C40 Grandstand Prestressed Concrete Pouring", the direct cost composition was further analyzed. The analyzed direct cost composition includes: the consumption of senior skilled workers' labor is 12 man-days per 10 cubic meters of concrete; the consumption of 525 silicate cement is 420 kg per cubic meter of concrete; the consumption of medium sand is 650 kg per cubic meter of concrete; the consumption of crushed stone is 1100 kg per cubic meter of concrete; the consumption of concrete pump machine shifts is 0.8 shifts per 100 cubic meters of concrete; and the historical unit price data corresponding to these consumptions during the construction period in 2021, such as the daily wage of senior skilled workers being 350 yuan, the price per ton of 525 silicate cement being 520 yuan, the price per ton of medium sand being 90 yuan, the price per ton of crushed stone being 70 yuan, and the machine shift cost of concrete pump being 1200 yuan. Four attributes were extracted from the project's feature information: geographical environment attribute ("coastal area, high wind speed"), structural type attribute ("prestressed frame structure, partial steel structure"), building scale attribute ("building area 80,000 square meters, seating capacity 40,000 seats"), and construction technology complexity attribute ("large-volume concrete construction, prestressed tensioning"). Taking the "C40 grandstand prestressed concrete pouring" sub-item as an example, with its comprehensive unit price as the dependent variable and multiple feature attributes such as "geographical environment attribute" (e.g., whether it is coastal), "structural type attribute," "building scale attribute," and "construction technology complexity attribute" (e.g., whether it is large-volume concrete construction) as independent variables, a multiple linear regression model was established. The model quantifies that, controlling for other attributes, the "coastal area" feature may increase the comprehensive unit price of this sub-item by approximately X%; the "large-volume concrete construction" feature may cause a Y% change in its comprehensive unit price. Here, X% and Y% are the influence coefficients calculated in the correlation model. Similarly, a multi-dimensional statistical correlation model is established between labor consumption and structural type attributes, and between material consumption and building scale attributes. Based on the multi-dimensional statistical correlation model of all sub-items in the database, a correlation graph of engineering cost characteristics is constructed. The nodes in the correlation graph of engineering cost characteristics represent specific sub-items such as "C40 grandstand prestressed concrete pouring" and "high-altitude bulk assembly of reticulated steel structure", and the edges between nodes represent the attribute correlation strength. The weight value of the edge is determined by the correlation coefficient in the statistical correlation model.

[0053] The design drawings and technical specifications of the comprehensive gymnasium project to be evaluated were analyzed. The geographical environmental characteristics of the project were extracted as "inland plain, seismic fortification intensity 7 degrees," the structural type characteristics as "reinforced concrete frame-shear wall structure, locally large-span steel grid shell," the building scale characteristics as "building area 65,000 square meters, 35,000 fixed seats," and the technical requirements characteristics as "seamless construction of ultra-long structures, overall hoisting of large components." These features constitute the initial project feature set. The textual features in the initial project feature set were transformed into feature vectors recognizable by the engineering cost feature association graph. The transformation method was one-hot encoding and numerical normalization. For example, the "structural type feature" was encoded as a multi-dimensional vector, where the dimension value of "reinforced concrete frame-shear wall structure" was 1, the dimension value of "locally large-span steel grid shell" was 1, and the dimension value of other structural types was 0. In the engineering cost feature association map, the multidimensional similarity between the feature vector of the project to be evaluated and the feature vector of the historical sub-project node "steel structure installation of a convention center" in the map is calculated. The calculation considers four dimensions: geographical environment matching degree, structural type matching degree, scale matching degree, and technical complexity matching degree. The multidimensional similarity threshold is set to 0.65. The system filters out all historical sub-project nodes with a multidimensional similarity exceeding 0.65. For example, nodes such as "large-span roof installation of a theater" and "prestressed beam construction of a high-speed railway station" are classified into a similar set of historical sub-projects.

[0054] In some embodiments, the multidimensional similarity is calculated using a weighted Manhattan distance method, the formula of which is expressed as: Where: symbol This represents the weighted distance between the feature vector of the project to be evaluated and the feature vectors of historical sub-project nodes. The symbol p represents the project to be evaluated, and the symbol q represents a historical sub-project node in the project cost feature correlation graph. The symbol represents the total number of dimensions of the feature vector. Indicates the first Preset weights for each feature dimension, with symbols Indicates the project to be evaluated in the [number]th [year]. Values ​​in each feature dimension, symbol Indicates the historical sub-project node at the [number]th The values ​​on each feature dimension. This can be understood as weighted distance. The smaller the value, the higher the multidimensional similarity. Similar nodes can be filtered out by setting an upper limit threshold for distance.

[0055] It is understandable that the extraction of geographical environment attributes, structural type attributes, building scale attributes, and construction technology complexity attributes relies on natural language processing and key information extraction techniques for unstructured engineering project documents. Optionally, the construction technology complexity attribute can be obtained by parsing the process descriptions in the construction organization design documents and combining them with an expert knowledge base for quantitative scoring. In some embodiments, the establishment of a multidimensional statistical correlation model adopts a multiple linear regression method, using the unit consumption of sub-projects as the dependent variable and multiple project characteristic attributes as independent variables to fit the correlation equation. Optionally, the construction of the engineering cost feature correlation map uses a graph database for storage and traversal, with each node storing the identifier and consumption data of sub-projects, and each edge storing the name and correlation coefficient of the correlation attribute. In specific implementation, the transformation process of the initial project feature set needs to be completely consistent with the encoding rules used when constructing the engineering cost feature correlation map to ensure that the feature vectors are in the same metric space.

[0056] In one embodiment of the present invention, S13: The process of forming an initial cost reference data pool based on a set of similar historical sub-projects using a dynamic similarity weighted cost data extraction method is as follows.

[0057] S131: For each historical sub-project in the set of similar historical sub-projects, calculate its corresponding weight coefficient based on the multi-dimensional similarity calculation results with the project to be evaluated. The higher the similarity, the greater the weight. According to the calculated weight coefficient, the historical unit price data corresponding to each historical sub-project is weighted, so that the price data of highly similar projects have a greater impact on the results. The weighted historical unit price data is then categorized and summarized according to their corresponding resource categories such as labor, materials, and machinery, forming weighted price sets for each resource category. S134: Outlier removal and statistical smoothing are performed on the summarized unit price data sets for each category to eliminate data noise. After processing, a relatively pure reference unit price sequence is generated by category. These sequences together constitute the initial cost reference data pool.

[0058] In this specific implementation, a planned prestressed concrete continuous beam bridge is used as the project to be evaluated. This embodiment details the process of forming an initial cost reference data pool based on a set of identified similar historical sub-projects. For each historical sub-project in the set of similar historical sub-projects, its weight coefficient is calculated based on its multidimensional similarity with the project to be evaluated. The multidimensional similarity value is calculated by feature vector matching in the previous embodiment.

[0059] For example, a historical sub-project node named "Pouring C40 Concrete for a Highway Bridge Pier" in the set has a multidimensional similarity of 0.82 with the prestressed concrete continuous beam bridge project to be evaluated. Therefore, its weight coefficient is directly set to 0.82. Based on the calculated weight coefficient, the historical unit price data corresponding to each historical sub-project is weighted. For the node "Pouring C40 Concrete for a Highway Bridge Pier," the unit price of "C40 commercial concrete" in its historical unit price data is 450 yuan / cubic meter. After weighting, its contribution value is 450 yuan / cubic meter multiplied by the weight coefficient 0.82, which is 369 yuan / cubic meter. The historical unit price data of all historical sub-projects after weighting are classified and summarized according to their corresponding labor, material and machinery categories. For example, the weighted unit price data of "C40 commercial concrete" in all historical sub-projects is summarized into one set, the weighted unit price data of "secondary steel reinforcement" is summarized into another set, and the weighted unit price data of "bridge formwork installation technicians" is summarized into a third set.

[0060] Outlier removal and statistical smoothing were performed on the aggregated unit price data sets for each category. For the "C40 ready-mixed concrete" reference unit price sequence, which contains 20 weighted historical unit price data points, outlier removal employed the three-sigma principle based on the normal distribution assumption. The mean μ and standard deviation σ of the sequence were calculated, and all data points outside the interval [μ-3σ, μ+3σ] were removed. Statistical smoothing then used a locally weighted regression method to fit the remaining data points after outlier removal, generating a smooth unit price-time relationship curve. This curve constitutes the reference unit price sequence for "C40 ready-mixed concrete." Outlier removal and statistical smoothing were repeated for other categories such as "Grade II steel reinforcement" and "formwork installation technicians," generating their respective reference unit price sequences. All these sequences together constitute the initial cost reference data pool specifically for this prestressed concrete continuous beam bridge project.

[0061] In some embodiments, the weighting coefficients can be calculated directly using normalized multidimensional similarity values, or they can be assigned values ​​using an exponential decay function based on similarity ranking. For example, similar historical sub-project sets can be sorted from high to low multidimensional similarity, with the weighting coefficient of the first-ranked node set to 1.0, the weighting coefficient of the second-ranked node set to λ, the weighting coefficient of the third-ranked node set to λ^2, and so on, where λ is a decay factor between 0 and 1. The formula is expressed as: ; Where: symbol Indicates the sorting order The weight coefficient of the historical sub-item project node, symbol Indicates the preset attenuation factor, symbol This indicates the order of nodes arranged from highest to lowest similarity. It can be understood that using an exponential decay function can further amplify the influence of historical data with high similarity while reducing the contribution of historical data with low similarity.

[0062] It is understandable that classification and summarization is the process of aggregating data from different sources and with different weights but belonging to the same resource category. This is a necessary step in constructing a structured initial cost reference data pool. In some embodiments, outlier removal can be achieved using box plots, identifying outliers as those below the lower quartile minus 1.5 times the interquartile range or above the upper quartile plus 1.5 times the interquartile range. Optionally, statistical smoothing can be performed using moving averages or Gaussian process regression to eliminate random fluctuations and reveal potential trends. In specific implementation, the initial cost reference data pool is ultimately organized into a database table, where each record represents a resource category. Record fields include resource category code, resource name, processed reference unit price sequence data, and the time range corresponding to the sequence data. The dynamic similarity-weighted cost data extraction method ensures that the data in the initial cost reference data pool not only originates from history but also has a high degree of correlation and representativeness with the current project to be evaluated.

[0063] See Figure 3 This is a chart showing the weight coefficients of historical sub-projects under different weighting strategies. It illustrates the changing trends of the weight coefficients of historical sub-projects with similarity ranking under three different weighting calculation strategies in engineering project cost assessment. The exponential decay strategy (λ=0.8 or 0.9) effectively highlights the reference value of historical data with high similarity, which is more in line with the business logic of "the more similar to the project to be assessed, the higher the reference value." The choice of the decay factor λ directly affects the weight distribution. The closer λ is to 1, the smoother the decay, and the higher the weight of low similarity data; the smaller λ is, the more drastic the decay, and the stronger the dominance of high similarity data. While directly using the similarity strategy is simple, it may cause unnecessary interference from low-similarity historical data in the cost assessment results.

[0064] In one embodiment of the engineering project cost assessment system based on dynamic data prediction described in this invention, the benchmark generation module is used to input the initial cost reference data pool into the dynamic cost benchmark generation model for processing as follows.

[0065] S20: For each category of reference unit price sequence in the initial cost reference data pool, such as steel reinforcement, concrete, and skilled worker wages, establish a time series forecasting model. Analyze historical price fluctuation patterns through the model to predict the future unit price trend line within the planned construction period of the project to be evaluated. S21: Simultaneously, acquire macroeconomic market index data from economic research institutions or statistical departments. This macroeconomic market index data includes industry inflation rate forecasts and regional economic development indicators. S22: Fit and calibrate the predicted future unit price trend line with the acquired macroeconomic market index data. Use the macroeconomic index to correct the microeconomic trend, generating a calibrated future benchmark unit price curve that better reflects macroeconomic market expectations. S23: Combine the estimated quantities of the project to be evaluated with the calibrated future benchmark unit price curve to convert it into a comprehensive unit price required to complete each sub-item of the project within the planned construction period. This comprehensive unit price is the dynamic cost benchmark for each sub-item.

[0066] In one embodiment of the data-driven dynamic prediction-based engineering project cost assessment system of the present invention, the factor acquisition module is used to collect dynamic cost impact factor data under the current market environment in real time through the following method.

[0067] S30: Dynamic cost impact factor data mainly includes material price fluctuation data, labor cost index and equipment rental rate change data.

[0068] The system connects to application programming interfaces (APIs) from multiple external data sources, including commodity trading platforms, human resource service agencies, and equipment supplier information systems. It periodically retrieves spot and futures price data for major building materials from the commodity trading platform's API.

[0069] We periodically retrieve wage guidelines and social security contribution rate data for different job types in different regions from the application programming interface of human resources service agencies.

[0070] Real-time rental quotes and maintenance cost data for various types of construction machinery are periodically obtained from the application programming interface of the equipment supplier's information system.

[0071] S35: Perform data cleaning and format standardization on the obtained spot and futures price data, wage guidance price and social security contribution rate data, and real-time rental quotation and maintenance cost data, remove invalid records, unify data format, and finally form a structured dynamic cost impact factor data stream.

[0072] For example, during project implementation, the consumption and price parameters of quota items can be dynamically adjusted to further promote quality engineering and green highway construction technologies.

[0073] In practical implementation, a planned commercial complex project is used as an example to illustrate the project to be evaluated, with a planned construction period from January 2025 to December 2026. The benchmark generation module inputs the initial cost reference data pool into the dynamic cost benchmark generation model for processing. The processing procedure is as follows: A time series forecasting model is established for each category of reference unit price series in the initial cost reference data pool. For example, for the category of reference unit price series for "HRB400 threaded steel bars for high-rise buildings," this series provides weighted monthly historical reference unit prices from January 2020 to December 2024. A seasonal autoregressive integral moving average model is used to analyze this series, predicting the future unit price change trend line for 24 months from January 2025 to December 2026. The future unit price change trend line shows cyclical fluctuations caused by seasonal procurement and production capacity factors. Macroeconomic market index data is obtained, including the construction industry's predicted inflation rate data for the next two years released by the National Bureau of Statistics, and the regional GDP annual growth forecast indicator data released by the municipal government where the project is located. The predicted future unit price trend line for "HRB400 threaded steel bars for high-rise buildings" was fitted and calibrated with the obtained macroeconomic market index data. The baseline level of the trend line was shifted and corrected using the construction industry's predicted inflation rate, and the long-term slope of the trend line was adjusted using the regional GDP growth forecast indicator, generating a calibrated future benchmark unit price curve. Combined with the estimated quantities of the commercial complex project to be evaluated (derived from preliminary design drawings), the estimated quantity of the "main structure steel bar fabrication and installation" sub-item was 8500 tons. The calibrated future benchmark unit price curve was integrated according to the planned construction progress months and allocated to the unit quantity, calculating a comprehensive unit price of RMB 5850 per ton for the "main structure steel bar fabrication and installation" sub-item within the planned construction period. This comprehensive unit price serves as the dynamic cost benchmark for this sub-item.

[0074] Real-time collection of dynamic cost impact factor data is achieved through the factor collection module. This data includes material price fluctuations, labor cost indices, and equipment rental rate changes. The factor collection module connects to multiple external data source application programming interfaces (APIs), including the "MySteel.com" commodity trading platform API, the "Local Construction Engineering Cost Management Station" human resources service agency API, and the "XCMG Group" equipment supplier information system API. The "MySteel.com" commodity trading platform API periodically retrieves spot and futures price data for major building materials; for example, it retrieves the average spot transaction price and three-month futures price of HRB400 rebar in major cities nationwide every hour. The "Local Construction Engineering Cost Management Station" human resources service agency API periodically retrieves wage guidance prices and social security contribution rates for different trades in different regions; for example, it retrieves the daily wage guidance prices and the latest social insurance contribution rates for rebar workers, formwork workers, and concrete workers in the project location at the beginning of each quarter. Real-time rental quotes and maintenance cost data for various construction machinery are periodically obtained from the application programming interface (API) of the XCMG Group's equipment supplier information system. For example, monthly rental market quotes and routine maintenance cost estimates for tower cranes and construction hoists are obtained daily. The obtained spot and futures price data, wage guidance prices and social security contribution rate data, and real-time rental quotes and maintenance cost data undergo data cleaning and format standardization. Data cleaning includes removing obviously erroneous negative prices or extreme outliers. Format standardization unifies the timestamps of data from different sources to Beijing time, the currency unit to RMB, and the data precision to two decimal places. After cleaning and standardization, a structured dynamic cost influencing factor data stream is formed, which is stored in the form of a time-series database table.

[0075] In some embodiments, the time series forecasting model may employ a long short-term memory neural network model, which is capable of capturing long-term dependencies and nonlinear patterns in the reference unit price series. Future unit price trend line. With macro market index Fit calibration can be achieved through a linear adjustment function, expressed as follows: Where: symbol Indicates a point in time The calibrated future benchmark unit price, symbol Represents the original future unit price predicted by the time series model, with the symbol... Indicates a point in time The macro market index value, symbol This represents the average value of a macro market index over its historical forecast period, with the symbol [symbol missing]. This represents the preset sensitivity coefficient. It can be understood that the calibration process quantifies overall expected changes in the macroeconomy into specific adjustments to the forecast of micro-material unit prices.

[0076] It is understandable that the acquisition of macroeconomic market index data needs to be matched with the start and end dates of the project's planned construction period to ensure the effectiveness of the correction. Optionally, regional economic development indicator data can be refined to the city-level fixed asset investment growth rate or construction industry output growth rate. In some embodiments, the application programming interface's timed acquisition task is implemented through a configured independent scheduling service, which automatically triggers data capture and updates at a set frequency. Optionally, the data cleaning process can include verification based on business rules, such as checking whether the price of steel bars is within the reasonable fluctuation range of the information price for the current year. The structured dynamic cost impact factor data collected in real time is cached in tabular form, see Table 1.

[0077] Table 1: Data Table of Dynamic Cost Impact Factors See Figure 4 This is a diagram illustrating the cost risk assessment and reserve allocation for engineering projects. It shows the risk levels and corresponding risk reserves for five core cost risk factors in engineering projects, clearly reflecting the matching relationship between risk levels and financial reserves. Steel price fluctuations have the highest risk level and the largest corresponding risk reserve, indicating that the project owner considers it the primary source of cost uncertainty. Policy and regulatory changes have the lowest risk level and the smallest reserve, reflecting that their direct impact on project costs is relatively controllable. The allocation logic of risk reserves is clear, following the principle of "the higher the risk, the more adequate the reserve," which aligns with the risk integration requirements in engineering cost assessment. Project management should prioritize monitoring the market prices of bulk materials such as steel and concrete, and hedge risks through hedging, long-term contracts, etc. Simultaneously, a dynamic adjustment mechanism should be established to update risk levels and reserve sizes promptly based on market changes.

[0078] The project cost assessment system based on dynamic data prediction, as described in this invention, includes a cost synthesis module in one embodiment. This module is used to input dynamic cost benchmarks and dynamic cost influencing factor data for each sub-item of the project into a real-time cost synthesis engine. The engine then performs cost aggregation and risk coefficient integration item by item to calculate and generate a real-time dynamic total cost assessment value for the project to be evaluated. Specifically, this includes: S40: For each sub-item of the project, the real-time cost synthesis engine compares its dynamic cost benchmark for that sub-item with the latest dynamic cost impact factor data corresponding to the main materials, labor, and machinery required for that sub-item. When the deviation between the dynamic cost impact factor data of a certain category and the predicted value used when generating the dynamic cost benchmark for the sub-item exceeds a preset fluctuation threshold, the original predicted value is replaced with the latest dynamic cost impact factor data, and the comprehensive unit price of the sub-item is recalculated in real time based on this. The above real-time recalculation process is performed on all sub-items of the project to be evaluated, and the costs of each recalculated sub-item are summarized to obtain the updated direct project costs. Based on the updated direct project costs, indirect costs, profits, and taxes are calculated according to the industry or enterprise preset rates. S44: The updated direct project costs, the calculated indirect costs, profits, and taxes are added together to obtain the real-time dynamic total cost assessment value.

[0079] The real-time cost synthesis engine performs a step-by-step cost aggregation and risk coefficient integration process, which also includes risk assessment and integration: S45: Analyze the various elements in the dynamic cost impact factor data, such as the historical volatility and current volatility of specific material prices. Identify cost elements whose volatility exceeds a set risk threshold and mark them as high-risk cost elements. For each marked high-risk cost element, simulate and calculate the maximum additional cost increment it may cause to the corresponding sub-item cost based on its historical maximum volatility. S48: Aggregate the maximum additional cost increments calculated for all sub-items according to a probability distribution model to estimate the overall cost risk reserve for the project under evaluation.

[0080] The project cost assessment system based on dynamic data prediction of the present invention further includes: The report output module outputs a structured cost assessment report, which combines the real-time dynamic total cost assessment value of the project to be evaluated with the dynamic cost benchmarks of all sub-projects. The report output module is used for: The system automatically populates the pre-defined report template framework with real-time dynamic total cost assessment, dynamic cost benchmarks for each sub-project, a list of identified high-risk cost elements, and the overall cost risk reserve. The report template displays historical trends in major material prices and compares current market values ​​with system forecasts using line charts or bar charts. It also generates a time-based rolling project cash flow forecast table based on the dynamic cost benchmarks for each sub-project and the project's established schedule. The report output module converts the generated structured cost assessment report into standard PDF or DOCX document formats for storage and can then be exported to the user.

[0081] In practical implementation, a commercial complex project planned for construction in 2025 is used as an example to illustrate the project to be evaluated. The real-time cost synthesis engine receives the dynamic cost benchmark of the sub-items from the benchmark generation module and the dynamic cost impact factor data provided by the factor acquisition module. For the sub-item "Pouring C40P8 impermeable concrete for the exterior wall of the third basement level", the comprehensive unit price prediction value of its sub-item dynamic cost benchmark in March 2025 is 650 yuan / cubic meter. The generation of this sub-item dynamic cost benchmark depends on the prediction of historical prices of cement, sand, and gravel. The real-time cost synthesis engine compares this sub-item dynamic cost benchmark with the latest spot price data of "PO42.5 cement" in the dynamic cost impact factor data. The latest data shows that the spot price of cement in early March 2025 had risen to 380 yuan / ton due to tight supply, while the prediction value used when generating the sub-item dynamic cost benchmark was 350 yuan / ton. With a preset fluctuation threshold of 5%, the deviation between the latest spot price of 380 yuan / ton and the predicted value of 350 yuan / ton is 8.57%, exceeding the preset fluctuation threshold. The real-time cost synthesis engine replaces the original predicted value with the latest cement price data of 380 yuan / ton and updates the price differences of other materials in the concrete mix design. It recalculates the comprehensive unit price of the "C40P8 anti-seepage concrete pouring for the third basement floor exterior wall" sub-project in real time, updating the comprehensive unit price to 678 yuan / cubic meter. For all sub-projects of the commercial complex project, including "main structure reinforcement engineering" and "curtain wall installation engineering," the above real-time recalculation process is executed, summing all updated sub-project costs to obtain the updated direct project cost, which is 78.5 million yuan. Based on the updated direct project cost of 78.5 million yuan, indirect costs are calculated at a preset rate of 10%, resulting in 7.85 million yuan; profits are calculated at a preset profit rate of 8%, resulting in 6.28 million yuan; and taxes are calculated at the current national tax rate of 9%, resulting in 7.065 million yuan. Adding the updated direct engineering costs, indirect costs, profits, and taxes together, we get 78.5 million yuan plus 7.85 million yuan plus 6.28 million yuan plus 7.065 million yuan, which equals 99.695 million yuan. This amount is the real-time dynamic total cost assessment value.

[0082] The real-time cost synthesis engine's process of item-by-item cost aggregation and risk coefficient integration includes risk assessment and integration steps. The risk assessment and integration steps analyze the historical volatility and current volatility of each element in the dynamic cost impact factor data. For example, the monthly volatility standard deviation of "HRB400E rebar" price over the past 24 months is 3.5%, while the current monthly price volatility has reached 5.2%. Cost elements with volatility exceeding the risk threshold are identified. The risk threshold is set at 1.5 times the historical volatility standard deviation, i.e., 5.25%. The current "HRB400E rebar" price volatility of 5.2% does not exceed the threshold, but the current volatility of "exterior wall aluminum panels" price reaches 7.8%, and its historical volatility standard deviation is 4.1%. Therefore, "exterior wall aluminum panels" are marked as a high-risk cost element. For the high-risk cost element of "exterior wall aluminum panels," based on its historical maximum fluctuation of 20% over the past three years, the maximum potential additional cost increase it may cause to the "unitized glass curtain wall installation" sub-project is calculated. The estimated cost of aluminum panels in this sub-project is 2 million yuan, and the maximum additional cost increase is 2 million yuan multiplied by 20%, which equals 400,000 yuan. Similar maximum additional cost increases calculated for all sub-projects, including the risk increases for materials such as steel structures and cables, are randomly aggregated 100,000 times using a Monte Carlo simulation probability distribution model to simulate combinations of adverse fluctuations in various risk elements simultaneously. The 95th percentile value of the aggregated result is calculated to be 2.85 million yuan. This 2.85 million yuan is the overall cost risk reserve for the project to be evaluated.

[0083] The report output module outputs the real-time dynamic total cost assessment value of the project to be evaluated and the dynamic cost benchmarks of all sub-projects as a structured cost assessment report. The module populates a pre-defined report template with fixed chapters and fields with the real-time dynamic total cost assessment value of 99.695 million yuan, a list of dynamic cost benchmarks for each sub-project, a list of identified high-risk cost elements ("exterior wall aluminum panels, low-voltage cables"), and an overall cost risk reserve of 2.85 million yuan. In the "Main Material Price Trend Analysis" section of the report template, a line chart is displayed showing the historical price trends of materials such as "PO42.5 cement" and "HRB400E rebar" over the past 24 months, their current market spot prices, and the price forecasts for the next 6 months predicted by the benchmark generation model. In the "Project Funding Planning" section of the report template, a monthly rolling project cash flow forecast table is generated. This cash flow forecast table is based on the dynamic cost benchmarks of each sub-project and the start and end times of each task in the project's main schedule. The table columns include month, planned output value, expected cash outflow, and cumulative cash flow. The report output module will generate a complete structured cost assessment report containing text, charts, and tables, convert it into a PDF document for storage, and output it to the user through the system interface.

[0084] In some embodiments, the aggregation of the maximum additional cost increment can be simulated using a triangular distribution or a normal distribution model. The formula for the overall risk reserve in Monte Carlo simulation involves multiple random sampling calculations, the core of which is calculating the total random increment. Distribution: Where: symbol This represents the stochastic total cost increment caused by all high-risk cost factors in a single simulation, denoted by [symbol]. Indicates the number of identified high-risk cost elements, symbol Representing each identified individual high-risk cost element, symbol Let represent a random number that follows a uniform distribution between 0 and 1, used to simulate the . Whether a risk has been triggered and its severity, symbol Indicates the first The historical maximum additional cost increment corresponding to each high-risk cost element. This can be understood through extensive simulations. By taking the highest quantile value, the overall cost risk reserve covering most risk scenarios can be obtained.

[0085] It is understandable that the calculation rates for indirect costs, profits, and taxes can be preset and configured based on enterprise quotas or local standards. Optionally, the generation of the cash flow forecast needs to consider the payment ratio and payment period of the project funds. Multiplying the planned monthly output value by the agreed payment ratio and lags it by a certain period is necessary to obtain a more accurate expected cash outflow sequence. In some embodiments, preset volatility thresholds can be set separately for different categories of cost impact factors. For example, a more lenient threshold, such as 8%, can be set for steel with frequent price fluctuations, while a stricter threshold, such as 3%, can be set for local materials with relatively stable prices. Optionally, the identification of high-risk cost elements can be achieved by introducing machine learning models. This involves analyzing multi-dimensional market sentiment data to predict potential price fluctuation risks, rather than relying solely on historical volatility statistics. The generation of the real-time dynamic total cost assessment value is a periodic or event-triggered automated process. Whenever the dynamic cost impact factor data is significantly updated, the real-time cost synthesis engine automatically initiates a new round of recalculation and assessment.

[0086] See Figure 5This chart analyzes the price trends of key materials, showing the price movements of PO42.5 cement and HRB400E rebar over the past 30 months. The two materials differ significantly in price magnitude, with rebar prices exceeding 10 times that of cement, making it a more sensitive variable in project costs. The continued rise in rebar prices will directly increase the cost of the main structural engineering, necessitating close monitoring of market fluctuations and the implementation of hedging measures. Cement prices, on the other hand, are relatively stable, making them a reliable benchmark in cost assessments with low forecasting error risk. The upward trend and potential fluctuations in rebar prices are the main source of project cost risk and must be included in the risk reserve calculation model. Based on the predicted trends, it is advisable to lock in bulk material procurement contracts when historical prices are relatively low to mitigate the impact of future cost increases.

[0087] The above describes the engineering project cost assessment system based on dynamic data prediction as described in this invention.

[0088] Understandably, once the entire project is completed, it will possess TB-level, or even PB-level, cost information data exceeding 10 billion. Through the data acquisition module and benchmark generation module of this invention, users can quickly retrieve the valuable data they need from the massive amount of information and quickly support the rapid application of specific data.

[0089] With the gradual implementation of various types of highly challenging projects, my country will integrate more and more engineering data and related assessment data, structuring and standardizing them to realize a search engine service model. This model will provide cost management personnel with more valuable, specialized, and in-depth information and related services, and will also have a vertical industry focus.

[0090] Vertical search engines can perform in-depth processing of structured and restructured data extracted from web pages, providing professional search services. Furthermore, vertical search can provide intelligent processing functions on demand, such as automatic classification, automatic clustering, automatic indexing, automatic deduplication, and text mining. Therefore, through the implementation of this invention, this system introduces its own vertical search engine technology based on project feature information to quickly provide data requesters with the most valuable data they are interested in and want within the data warehouse. The vertical search engine allows data requesters to filter and locate project and cost information within one minute using keywords such as project name, compilation year, route, specialty, stage, region, and warning value; it also allows data requesters to quickly search and query using information such as process, construction method, engineering characteristics, and material and equipment information, making the cost information search engine as convenient as using Baidu or Google.

[0091] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A project cost assessment system based on dynamic data prediction, characterized in that, include: The data acquisition module is used to collect the historical cost database of the target project, perform structured parsing on the historical cost database to form a cost feature association map, and use a dynamic similarity weighted cost data extraction method to extract the corresponding historical cost component data from the historical cost database to form an initial cost reference data pool. The benchmark generation module is used to input the initial cost reference data pool into the dynamic cost benchmark generation model for processing. The dynamic cost benchmark generation model is based on time series analysis and market factor correction algorithm to generate dynamic cost benchmarks for sub-projects applicable to the project to be evaluated. The factor acquisition module is used to collect dynamic cost impact factor data in the current market environment in real time. The cost synthesis module is used to input the dynamic cost benchmark of the sub-project and the dynamic cost impact factor data into the real-time cost synthesis engine. The real-time cost synthesis engine performs cost superposition and risk coefficient integration item by item to calculate and generate the real-time dynamic total cost assessment value of the project to be evaluated. In the data acquisition module, the process of forming an initial cost reference data pool involves the following steps: S10: Collect the historical cost database of the target project. The historical cost database contains a complete cost list of completed projects and their corresponding project feature information. S11: Perform structured parsing on the complete cost list in the historical cost database, extract the cost components of each sub-item project, and establish a mapping relationship between the cost components of each sub-item project and its corresponding project feature information to form a project cost feature association map. S12: Obtain the initial project feature set of the project to be evaluated, perform pattern matching between the initial project feature set of the project to be evaluated and the engineering cost feature association map, and identify the set of historical sub-projects similar to the project to be evaluated. S13: Based on the aforementioned set of similar historical sub-projects, a dynamic similarity-weighted cost data extraction method is used to extract the corresponding historical cost component data from the historical cost database to form an initial cost reference data pool.

2. The project cost assessment system based on dynamic data prediction as described in claim 1, characterized in that, The data acquisition module executes step S11: The complete cost list in the historical cost database is structured and parsed to extract the cost components of each sub-item of the project, and a mapping relationship is established between the cost components of each sub-item and its corresponding project feature information, forming a project cost feature association map. Specifically, this includes: S111: In accordance with the standard bill of quantities specifications, the complete cost list shall be broken down to the level of individual items. S112: For each sub-item of the project, further analyze its direct cost composition, which includes labor consumption, material consumption, machine shift consumption and their corresponding historical unit price data; S113: Extract geographical environment attributes, structural type attributes, building scale attributes, and construction technology complexity attributes from the project feature information; S114: Establish a multi-dimensional statistical correlation model between the labor consumption, material consumption, and machine shift consumption in the direct cost composition of each sub-item of the project and the geographical environment attribute, structural type attribute, building scale attribute, and construction technology complexity attribute, respectively. S115: Based on the multi-dimensional statistical correlation model of all sub-items of the project, construct the project cost feature correlation graph; the project cost feature correlation graph includes project feature attribute nodes and sub-item project nodes, and the sub-item project nodes are connected to the relevant project feature attribute nodes through edges, and the weight of the edges is set based on the corresponding correlation strength.

3. The project cost assessment system based on dynamic data prediction as described in claim 1, characterized in that, The data acquisition module executes step S12: It acquires the initial project feature set of the project to be evaluated, performs pattern matching between the initial project feature set and the engineering cost feature association map, and identifies a set of historical sub-projects similar to the project to be evaluated, specifically including: S121: Analyze the design drawings and technical specifications of the project to be evaluated, and extract the geographical environment characteristics, structural type characteristics, building scale characteristics and technical requirements characteristics of the project to be evaluated as the initial project feature set; S122 transforms each feature in the initial project feature set into a feature vector that can be recognized by the engineering cost feature association map; S123: In the engineering cost feature association map, calculate the multidimensional similarity between the feature vector of the project to be evaluated and the feature vector of each historical sub-project node in the map. S124: Set a multidimensional similarity threshold, filter out all historical sub-project nodes whose multidimensional similarity exceeds the threshold, and classify them into the set of similar historical sub-projects.

4. The project cost assessment system based on dynamic data prediction as described in claim 1, characterized in that, The data acquisition module executes step S13: Based on the set of similar historical sub-projects, a dynamic similarity-weighted cost data extraction method is used to extract corresponding historical cost component data from the historical cost database to form an initial cost reference data pool, specifically including: S131: For each historical sub-project in the set of similar historical sub-projects, calculate its weight coefficient based on its multidimensional similarity with the project to be evaluated. S132: Weight the historical unit price data corresponding to each historical sub-item project according to the weighting coefficients; S133: Summarize the weighted historical unit price data according to their corresponding labor, material, and machinery categories; S134: The summarized unit price data for each category are subjected to outlier removal and statistical smoothing to generate a reference unit price sequence divided by category, which constitutes the initial cost reference data pool.

5. The project cost assessment system based on dynamic data prediction as described in claim 1, characterized in that, The benchmark generation module inputs the initial cost reference data pool into the dynamic cost benchmark generation model for processing. The dynamic cost benchmark generation model, based on time series analysis and market factor correction algorithms, generates dynamic cost benchmarks for the sub-items of the project to be evaluated, and performs the following steps: S20: For each category of reference unit price sequence in the initial cost reference data pool, establish a time series prediction model to predict the future unit price change trend line within the planned construction period of the project to be evaluated. S21: Obtain macro market index data, which includes industry inflation rate forecast data and regional economic development indicator data; S22: Fit and calibrate the future unit price change trend line with the macro market index data to generate a calibrated future benchmark unit price curve; S23: Combining the estimated quantity data of the project to be evaluated, the calibrated future benchmark unit price curve is converted into the comprehensive unit price of each sub-item, that is, the dynamic cost benchmark of the sub-item.

6. The engineering project cost assessment system based on dynamic data prediction as described in claim 1, characterized in that, The factor acquisition module is used to collect dynamic cost impact factor data in the current market environment in real time, and performs the following steps: S30: The dynamic cost impact factor data includes material price fluctuation data, labor cost index, and equipment rental rate change data; S31: An application programming interface for accessing multiple external data sources, including commodity trading platforms, human resource service agencies, and equipment supplier information systems; S32: Periodically obtain spot and futures price data of major building materials from the application programming interface of the commodity trading platform; S33: Periodically obtain wage guidance prices and social security contribution rate data for different types of work in different regions from the application programming interface of the human resources service agency; S34: Periodically obtain real-time rental quotes and maintenance cost data for various types of construction machinery from the application programming interface of the equipment supplier information system; S35: Perform data cleaning and format standardization on the obtained spot and futures price data, wage guidance price and social security contribution rate data, real-time rental quotation and maintenance cost data to form the aforementioned dynamic cost impact factor data in a structured form.

7. The engineering project cost assessment system based on dynamic data prediction as described in claim 1, characterized in that, The cost synthesis module is used to input the dynamic cost benchmark of the sub-project and the dynamic cost impact factor data into the real-time cost synthesis engine. The real-time cost synthesis engine performs cost aggregation and risk coefficient integration item by item to calculate and generate the real-time dynamic total cost assessment value of the project to be evaluated, and performs the following steps: S40: Compare the dynamic cost benchmark of each sub-item project with the latest dynamic cost impact factor data corresponding to the main materials, labor and machinery required for the sub-item project; S41: When the deviation between the dynamic cost impact factor data of a certain category and the predicted value used when generating the dynamic cost benchmark of the sub-item exceeds the preset fluctuation threshold, the original predicted value is replaced with the latest dynamic cost impact factor data, and the comprehensive unit price of the sub-item is recalculated in real time. S42: Perform real-time recalculation on all sub-projects of the project to be evaluated, and summarize to obtain the updated direct engineering costs; S43: Based on the updated direct engineering costs, calculate indirect costs, profits, and taxes according to preset rates; S44: Add the updated direct engineering costs, indirect costs, profits, and taxes to obtain the real-time dynamic total cost assessment value.

8. The engineering project cost assessment system based on dynamic data prediction as described in claim 7, characterized in that, The real-time cost synthesis engine performs cost aggregation and risk coefficient integration item by item, including risk assessment and integration. S45: Analyze the historical volatility and current volatility of each element in the dynamic cost impact factor data; S46: Identify cost elements whose volatility exceeds the risk threshold and mark them as high-risk cost elements; S47: For each high-risk cost element, calculate the maximum additional cost increment it causes to the corresponding sub-item cost based on its historical maximum fluctuation range; S48: Aggregate the maximum additional cost increments calculated from all sub-projects according to the probability distribution model to calculate the overall cost risk reserve value of the project to be evaluated.

9. The engineering project cost assessment system based on dynamic data prediction as described in claim 1, characterized in that, The system also includes: The report output module outputs the real-time dynamic total cost assessment value of the project to be evaluated and the dynamic cost benchmarks of all sub-projects as a structured cost assessment report. This module is used for: Fill the preset report template with the real-time dynamic total cost assessment value, the dynamic cost benchmark of each sub-project, the list of identified high-risk cost elements, and the overall cost risk reserve value. The report template displays historical trends, current values, and forecasts of major material prices in chart form. The report template generates a project cash flow forecast table that scrolls over time. The cash flow forecast table is generated based on the dynamic cost benchmark of the sub-project and the project schedule. The generated complete structured cost assessment report will be converted into a standard document format for storage and output.

Images