Construction estimating method, apparatus, and device

Abstract

The application discloses a construction estimation method, device and equipment, and relates to the technical field of engineering cost. The method comprises the following steps: obtaining a target construction mode and a first engineering quantity characteristic parameter corresponding to the target construction mode; determining a first characteristic vector corresponding to the target construction mode according to the first engineering quantity characteristic parameter and a target estimation rule library corresponding to the first engineering quantity characteristic parameter; processing the first characteristic vector by using a target integrated model to obtain construction price prediction values corresponding to N target construction modes; and determining a target construction price estimation value corresponding to the target construction mode according to the N construction price prediction values. According to the embodiment of the application, the accuracy of construction estimation can be improved.

Description

Technical Field

[0001] This application belongs to the field of engineering cost technology, specifically relating to a construction cost estimation method, apparatus and equipment. Background Technology

[0002] With the digital transformation of the construction industry, modular construction is becoming a widely adopted method in modern construction projects. Modular construction breaks down buildings into multiple standardized modules, which are prefabricated in factories and then transported to the construction site for rapid assembly. Statistics show that projects using modular construction can shorten the construction period by an average of 30%-50% and reduce the incidence of construction quality problems by approximately 40%. This construction method offers advantages such as significantly shortening the construction cycle, reducing on-site pollution, and improving construction quality and safety.

[0003] However, compared to traditional construction methods, modular construction involves multiple stages (such as prefabrication, transportation, and on-site assembly), leading to greater complexity in cost estimation. Modular construction projects typically include various construction modes such as rebar cage construction, rebar cage + steel cladding construction, composite slab construction for equipment room roofs, and composite slab construction for pipe racks. Each mode has significantly different cost structures and calculation methods. Traditional cost estimation methods rely on manual experience and simple calculation tools, which cannot effectively address the complexity of modular construction projects and cannot cope with the dynamic impact of factors such as market price fluctuations, changes in transportation distance, and differences in construction site environments on costs. Consequently, the accuracy of construction cost estimates needs improvement. Summary of the Invention

[0004] The technical problem to be solved by this application is to provide a construction cost estimation method, apparatus and equipment to address the above-mentioned shortcomings of the existing technology. Using this construction cost estimation method can improve the accuracy of construction cost estimation.

[0005] In a first aspect, embodiments of this application provide a construction cost estimation method, including: Obtain the target construction mode and its corresponding first quantity characteristic parameters; the target construction mode includes one of the following: steel cage construction mode, steel lining construction mode, and custom construction mode; the first quantity characteristic parameters are parameters used to characterize the quantity and cost composition of the target construction mode. Based on the first engineering quantity characteristic parameters and its corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined; the first feature vector is related to the core costs corresponding to the target construction mode; the target estimation rule base includes the target average ratio between the engineering quantities, the target average unit price of the core costs, and the target average proportion of the core costs to the total cost under the target construction mode. The first feature vector is processed using a target ensemble model to obtain the predicted construction price values ​​corresponding to N target construction modes. The target ensemble model includes N terms from the target ridge regression model, target decision tree model, target random forest model, target support vector regression model, and target neural network; N is a positive integer greater than or equal to 2. Based on N predicted construction prices, determine the estimated target construction price corresponding to the target construction mode.

[0006] In some embodiments of the first aspect, before determining the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and their corresponding target estimation rule base, the method further includes: Obtain the target type corresponding to the first engineering quantity characteristic parameter; Based on the target type, the target estimation rule library is determined from multiple preset estimation rule libraries. The preset estimation rule library includes the preset average ratio between various quantities of work, the preset average unit price of various core costs, and the preset average proportion of core costs to the total cost.

[0007] In some embodiments of the first aspect, before obtaining the target type corresponding to the first engineering quantity characteristic parameter, the method further includes: Obtain historical project quantity characteristic parameters; historical project quantity characteristic parameters are historical parameters used to characterize the project quantity and cost characteristics under different construction modes; Clustering algorithms are used to group the characteristic parameters of historical engineering quantities, resulting in at least two preset types; A preset estimation rule library is generated for each preset type and historical engineering quantity characteristic parameter to obtain multiple preset estimation rule libraries.

[0008] In some embodiments of the first aspect, a first feature vector corresponding to the target construction mode is determined based on the first engineering quantity characteristic parameters and their corresponding target estimation rule base, including: Based on the target estimation rule base, complete the first engineering quantity characteristic parameters to obtain the second engineering quantity characteristic parameters; Based on the second engineering quantity characteristic parameters and the target estimation rule base, determine the core costs corresponding to the target construction mode; All core costs are used to construct the first feature vector.

[0009] In some embodiments of the first aspect, before determining the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and their corresponding target estimation rule base, the method further includes: The first engineering quantity feature is preprocessed to obtain the preprocessed first engineering quantity feature parameters; the preprocessing includes at least one of the following: numerical type conversion, invalid data removal, missing data filling, and standardization. Based on the first engineering quantity characteristic parameters and their corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined, including: Based on the preprocessed first engineering quantity characteristic parameters and their corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined.

[0010] In some implementations of the first aspect, the target construction price estimate corresponding to the target construction mode is determined based on N predicted construction prices, including: The average of the N predicted construction prices is used as the target estimated construction price.

[0011] In some implementations of the first aspect, the target construction price estimate corresponding to the target construction mode is determined based on N predicted construction prices, including: The ratio method was used to verify the results based on the average of N construction price predictions. When the verification results are used to indicate that the verification has passed, the average value is determined as the target construction price estimate.

[0012] In some embodiments of the first aspect, after determining the estimated target construction price corresponding to the target construction mode based on N predicted construction prices, the method further includes: Visualize the estimated construction cost of the target project.

[0013] Based on the same inventive concept, in a second aspect, embodiments of this application also provide a construction cost estimation device, comprising: The first acquisition module is used to acquire the target construction mode and its corresponding first engineering quantity characteristic parameters; the target construction mode includes one of the steel cage construction mode, steel lining construction mode and custom construction mode; the first engineering quantity characteristic parameters are parameters used to characterize the engineering quantity and cost composition characteristics of the target construction mode. The first determining module is used to determine the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and its corresponding target estimation rule library; the first feature vector is related to the core costs corresponding to the target construction mode; the target estimation rule library includes the target average ratio between the engineering quantities, the target average unit price of the core costs, and the target average proportion of the core costs to the total cost under the target construction mode. The first processing module is used to process the first feature vector using a target ensemble model to obtain the predicted construction price values ​​corresponding to N target construction modes; the target ensemble model includes N terms from the target ridge regression model, target decision tree model, target random forest model, target support vector regression model, and target neural network; N is a positive integer greater than or equal to 2; The second determining module is used to determine the estimated target construction price corresponding to the target construction mode based on N predicted construction prices.

[0014] Based on the same inventive concept, in a third aspect, embodiments of this application provide an electronic device, the electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores one or more computer programs executable by the at least one processor, the one or more computer programs being executed by the at least one processor to enable the at least one processor to perform the above-described construction cost estimation method.

[0015] Based on the same inventive concept, in a fourth aspect, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the above-described construction cost estimation method.

[0016] Based on the same inventive concept, in a fifth aspect, embodiments of this application provide a computer program product that includes computer-readable code or a non-volatile computer-readable storage medium carrying computer-readable code. When the computer-readable code is run in the processor of an electronic device, the processor in the electronic device executes the above-described construction cost estimation method.

[0017] According to the construction cost estimation method, apparatus, and equipment provided in this application embodiment, the target construction mode and its corresponding first engineering quantity feature parameters are first obtained; then, based on the first engineering quantity feature parameters and its corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined; then, the first feature vector is processed using a target ensemble model to obtain N predicted construction prices corresponding to the target construction modes; and finally, based on the N predicted construction prices, the estimated target construction price corresponding to the target construction mode is determined. In other words, in this application embodiment, since the target ensemble model includes N terms from a target ridge regression model, a target decision tree model, a target random forest model, a target support vector regression model, and a target neural network; where N is a positive integer greater than or equal to 2, processing the first feature vector using the target ensemble model to obtain N predicted construction prices corresponding to the target construction modes can, on the one hand, integrate the advantages of at least two models to improve the accuracy of the predicted construction prices; on the other hand, by employing an advanced combination of machine learning algorithms, it can effectively cope with the complexity of modular construction projects and address the dynamic impact of factors such as market price fluctuations, changes in transportation distance, and differences in the construction site environment on costs, thereby improving the accuracy of the predicted construction prices. Furthermore, based on N predicted construction prices, the estimated target construction price corresponding to the target construction mode is determined, which improves the accuracy of the estimated target construction price. In summary, the embodiments of this application can improve the accuracy of construction cost estimation.

[0018] Furthermore, in the embodiments of this application, the algorithm can be flexibly adjusted according to different project needs and market environments to ensure that each engineering task can obtain an accurate cost estimate.

[0019] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate the present application and form part of the specification. They are used together with the embodiments of the present application to explain the application and do not constitute a limitation thereof. The above and other features and advantages will become more apparent to those skilled in the art from the detailed example embodiments described with reference to the accompanying drawings, in which: Figure 1 This illustration shows a flowchart of a construction cost estimation method provided in an embodiment of this application. Figure 2 This illustration shows another flowchart of the construction cost estimation method provided in the embodiments of this application; Figure 3 A schematic diagram illustrating a four-layer system architecture design for the construction cost estimation method provided in this application embodiment; Figure 4 A schematic diagram illustrating the system functional structure of the construction cost estimation method provided in an embodiment of this application is shown. Figure 5 This illustration shows a schematic diagram of a human-computer interaction interface for training a target ensemble model provided in an embodiment of this application. Figure 6 This illustration shows a schematic diagram of a target ensemble model training method provided in an embodiment of this application. Figure 7 This illustration shows a schematic diagram of feature clustering analysis provided in an embodiment of this application; Figures 8(a), 8(b), and 8(c) respectively illustrate a schematic diagram of the performance index of different regression models provided in the embodiments of this application; Figure 9 This illustration shows a schematic diagram of a system user interface design provided in an embodiment of this application; Figure 10 This illustration shows a comparative diagram of the performance metrics of various regression models provided in the embodiments of this application; Figure 11 This illustration shows another comparative diagram of the performance metrics of various regression models provided in the embodiments of this application; Figure 12 This invention provides a schematic diagram of the construction cost estimation device according to an embodiment of the present application. Figure 13This illustration shows a structural diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] To enable those skilled in the art to better understand the technical solutions of this application, exemplary embodiments of this application are described below in conjunction with the accompanying drawings, including various details of the embodiments of this application to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art should recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0022] Where there is no conflict, the various embodiments of this application and the features thereof may be combined with each other.

[0023] As used herein, the term “and / or” includes any and all combinations of one or more related enumerated entries.

[0024] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application. As used herein, the singular forms “a” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will also be understood that when the terms “comprising” and / or “made of” are used in this specification, the presence of the stated feature, integral, step, operation, element, and / or component is specified, but the presence or addition of one or more other features, integrals, steps, operations, elements, components, and / or groups thereof is not excluded. Terms such as “connected” or “linked” are not limited to physical or mechanical connections but can include electrical connections, whether direct or indirect.

[0025] Unless otherwise specified, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art. It will also be understood that terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant art and this application, and will not be interpreted as having an idealized or overly formal meaning, unless expressly so defined herein.

[0026] It should be understood that the term "and / or" used in this document is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. Additionally, the character "" in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0027] Before introducing the construction cost estimation method provided in the embodiments of this application, the relevant technologies that may be involved will be explained.

[0028] Existing systems are mostly based on manual input and traditional data analysis methods, which have serious shortcomings in data integration capabilities, estimation accuracy, and computational efficiency. Specifically, the estimation accuracy is generally low, with prediction errors of traditional methods often exceeding 30%; there is a lack of unified support for various construction modes; real-time price forecasting and dynamic adjustment are not possible; and data processing capabilities are limited, making it difficult to integrate multi-dimensional engineering data.

[0029] Against this backdrop, this application's embodiments conduct an in-depth study of existing cost estimation techniques in modular construction projects, finding that existing technologies mainly suffer from the complexity of modular construction and the limitations of traditional estimation methods. Modular construction involves multiple stages, including factory prefabrication, transportation, and on-site assembly. The cost of each stage is affected by various factors such as raw material prices, transportation costs, and the construction site environment. Existing cost estimation systems lack the ability to effectively model such complex relationships among multiple stages and variables, making it difficult to accurately reflect the true cost composition of modular construction. Although some systems on the market have attempted to introduce artificial intelligence (AI) technology, most adopt single algorithm models, such as using only linear regression or simple neural networks, lacking multi-model integration and intelligent selection mechanisms. When faced with different types of construction projects, these systems often cannot adaptively select the optimal algorithm, resulting in limited prediction accuracy.

[0030] Specifically, existing cost estimation techniques and tools have several obvious shortcomings in practical applications.

[0031] 1) Insufficient algorithm simplicity and prediction accuracy

[0032] Traditional cost estimation methods primarily rely on quota calculations and simple statistical regression models, lacking support from advanced machine learning algorithms. Even existing systems employing AI technology often use single algorithm models, such as linear regression or simple neural networks, which cannot handle the complex nonlinear relationships in modular construction. Experiments show that the prediction error of a single algorithm model is typically between 25% and 40%, far from meeting the needs of accurate cost estimation. Existing systems lack multi-algorithm integration mechanisms, failing to automatically select the optimal algorithm based on different engineering characteristics and data features. For example, the most suitable algorithm models for reinforced cage construction and composite slab construction of pipe galleries may be completely different, but existing systems cannot achieve this intelligent algorithm selection.

[0033] 2) Weak data processing and integration capabilities

[0034] Existing systems are severely inadequate in data preprocessing, lacking effective data cleaning, standardization, and feature engineering mechanisms. Particularly when processing multi-dimensional engineering data, existing systems cannot effectively identify and handle outliers, missing values, and inconsistencies, leading to poor model training quality. Traditional systems lack intelligent analysis capabilities for historical project data, failing to identify project characteristic patterns through methods such as cluster analysis, making it difficult to apply targeted estimation rules to different types of projects. Existing systems also lack dynamic data update mechanisms, failing to integrate external factors such as market price changes and material cost fluctuations in real time.

[0035] To solve at least one of the above-mentioned technical problems, a construction cost estimation method, apparatus, and equipment are provided. Using this construction cost estimation method can improve the accuracy of construction cost estimation.

[0036] The construction cost estimation method provided in the embodiments of this application is described below.

[0037] The construction cost estimation method provided in this application is applicable to the construction cost estimation process of modular construction. This method can be executed by a construction cost estimation device and electronic equipment. The following description uses the execution of this construction cost estimation method by electronic equipment as an example.

[0038] like Figure 1 As shown, the construction cost estimation method provided in this application includes steps S110 to S140.

[0039] S110. Obtain the target construction mode and its corresponding first engineering quantity characteristic parameters; the target construction mode includes one of the following: steel cage construction mode, steel lining construction mode, and custom construction mode; the first engineering quantity characteristic parameters are parameters used to characterize the engineering quantity and cost composition characteristics of the target construction mode.

[0040] S120. Based on the first engineering quantity characteristic parameters and their corresponding target estimation rule base, determine the first feature vector corresponding to the target construction mode; the first feature vector is related to each core cost corresponding to the target construction mode; the target estimation rule base includes the target average ratio between each engineering quantity, the target average unit price of each core cost, and the target average proportion of core cost to total cost under the target construction mode.

[0041] S130. The first feature vector is processed using a target ensemble model to obtain the predicted construction price values ​​corresponding to N target construction modes. The target ensemble model includes N terms from the target ridge regression model, target decision tree model, target random forest model, target support vector regression model, and target neural network. N is a positive integer greater than or equal to 2.

[0042] S140. Based on N predicted construction prices, determine the estimated target construction price corresponding to the target construction mode.

[0043] According to the construction cost estimation method provided in this application embodiment, the target construction mode and its corresponding first engineering quantity feature parameters are first obtained; then, based on the first engineering quantity feature parameters and its corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined; then, the first feature vector is processed using a target ensemble model to obtain N predicted construction prices corresponding to the target construction modes; and finally, based on the N predicted construction prices, the estimated target construction price corresponding to the target construction mode is determined. In other words, in this application embodiment, since the target ensemble model includes N terms from a target ridge regression model, a target decision tree model, a target random forest model, a target support vector regression model, and a target neural network; where N is a positive integer greater than or equal to 2, processing the first feature vector using the target ensemble model to obtain N predicted construction prices corresponding to the target construction modes can, on the one hand, combine the advantages of at least two models to improve the accuracy of the predicted construction prices; on the other hand, by employing an advanced combination of machine learning algorithms, it can effectively cope with the complexity of modular construction projects and address the dynamic impact of factors such as market price fluctuations, changes in transportation distance, and differences in the construction site environment on costs, thereby improving the accuracy of the predicted construction prices. Furthermore, based on N predicted construction prices, the estimated target construction price corresponding to the target construction mode is determined, which improves the accuracy of the estimated target construction price. In summary, the embodiments of this application can improve the accuracy of construction cost estimation.

[0044] Furthermore, in the embodiments of this application, the algorithm can be flexibly adjusted according to different project needs and market environments to ensure that each engineering task can obtain an accurate cost estimate.

[0045] The specific implementation methods for each of the above steps are described below.

[0046] In step S110, for example, the rebar cage construction mode may be a mode specifically designed for modular construction projects of reinforced concrete structures. The first quantity characteristic parameter corresponding to the rebar cage construction mode may include at least one of the relevant quantity parameters such as tower crane rental, rebar production, and the number of slings and sleeves.

[0047] For example, the steel lining construction mode can be a mode applicable to the design of projects that include steel lining structures. The first quantity characteristic parameter corresponding to the steel lining construction mode may include at least one of the key quantities such as assembly site, fabrication jig, steel supports, buttresses, and walkway slabs.

[0048] It should be noted that, in order to meet the needs of special or non-standardized projects, the system provides a highly flexible custom construction mode, allowing users to define various construction parameters and quantities of the project according to the actual situation, so as to realize personalized cost estimation for specific engineering scenarios, which greatly enhances the system's versatility and adaptability.

[0049] For example, in response to a user's input operation via an electronic device's interactive interface, the target construction mode and its corresponding first quantity characteristic parameters can be obtained. The output operation can be the user's input of the target construction mode and its corresponding first quantity characteristic parameters.

[0050] In step S120, the core costs corresponding to different target construction modes are different. Specifically, the core costs corresponding to the rebar cage construction mode may include tower crane rental fees, rebar production costs, sling costs, and sleeve quantity costs, etc. The target average unit price corresponding to the rebar cage construction mode includes average unit tower crane rental cost, average unit rebar production line cost, average unit sling cost, and average unit sleeve cost. The core costs corresponding to the steel lining construction mode may include assembly site costs, jig fabrication costs, steel support costs, buttress column costs, and walkway costs, etc. The target average unit price corresponding to the steel lining construction mode includes average unit assembly site cost, average unit steel support cost, average unit steel support cost, and average unit walkway cost, etc. The core costs corresponding to the custom construction mode can be set according to user needs and are not limited here.

[0051] In some implementations, a first feature vector corresponding to the target construction mode is determined based on the first engineering quantity characteristic parameters and their corresponding target estimation rule base, including: Based on the target estimation rule base, complete the first engineering quantity characteristic parameters to obtain the second engineering quantity characteristic parameters; Based on the second engineering quantity characteristic parameters and the target estimation rule base, determine the core costs corresponding to the target construction mode; All core costs are used to construct the first feature vector.

[0052] For example, when a user selects the rebar cage construction mode on the interactive interface of an electronic device and only enters the total tonnage of the rebar, all other necessary quantities of work (such as the quantity of tower crane rental, the quantity of slings, the quantity of sleeves, etc.) can be automatically estimated based on the rules in the "Cluster 0" rule base (i.e. the target estimation rule base), such as the "average ratio of the number of sleeves to the total tonnage of the rebar" and the "average ratio of the tower crane rental quantity to the total tonnage of the rebar". The second quantity of work characteristic parameters are then summarized.

[0053] It should be noted that for the quantities of work already provided by the user (i.e., the first quantity characteristic parameter), their values ​​are used directly; for the quantities of work not provided by the user (i.e., the second quantity characteristic parameter), the average ratio in the target estimation rule base can be used for intelligent estimation.

[0054] For example, still taking the rebar cage construction mode as an example, the target average ratio may include the average ratio of the number of sleeves to the total tonnage of rebar (i.e., the average number of sleeves per ton of rebar), the average ratio of tower crane rental work volume to the total tonnage of rebar (i.e., the average tower crane rental work volume per ton of rebar), and the average ratio of the number of slings to the total tonnage of rebar (i.e., the average number of slings per ton of rebar), etc. For example, if the user provides "total tonnage of rebar", other unprovided work volumes are estimated in the following ways: tower crane rental work volume = total tonnage of rebar × average tower crane rental work volume per ton of rebar, number of slings = total tonnage of rebar × corresponding average ratio (i.e., the average ratio of tower crane rental work volume to total tonnage of rebar), number of sleeves = total tonnage of rebar × corresponding average ratio (i.e., the average ratio of sleeve number to total tonnage of rebar); as another example, if the user does not provide "total tonnage of rebar", it can be estimated in reverse: total tonnage of rebar = tower crane rental work volume ÷ average tonnage of rebar per unit of tower crane rental work volume.

[0055] For example, once all quantity parameters (i.e., the second quantity characteristic parameters) are ready, the core costs can be calculated based on the "target average unit cost" in the target estimation rule base. Taking the rebar cage construction mode as an example again, the tower crane rental cost = estimated tower crane rental quantity × average unit tower crane rental cost; the rebar production line cost = total tonnage of rebar input by the user × average unit rebar production line cost; and so on, to calculate the required core costs.

[0056] In some implementations, before determining the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and their corresponding target estimation rule base, the method further includes: The first engineering quantity feature is preprocessed to obtain the preprocessed first engineering quantity feature parameters; the preprocessing includes at least one of the following: numerical type conversion, invalid data removal, missing data filling, and standardization. Based on the first engineering quantity characteristic parameters and their corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined, including: Based on the preprocessed first engineering quantity characteristic parameters and their corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined.

[0057] In this embodiment, by performing preprocessing such as numerical type conversion, invalid data removal, missing data filling, and standardization on the first engineering quantity characteristic parameters, and then determining the first feature vector corresponding to the target construction mode based on the preprocessed first engineering quantity characteristic parameters and their corresponding target estimation rule base, the standardization, consistency, and high quality of the generated first feature vector can be ensured, thereby improving the accuracy of construction estimation.

[0058] For example, using the `to_numeric` method of the pandas library, the numerical feature columns (such as various costs, quantities, and total project price) in the first quantity feature parameter are strictly converted to numeric types. For non-numeric entries that cannot be converted, the system will forcibly convert them to NaN (invalid value). Next, the converted first quantity feature parameter is identified and invalid project records are removed, i.e., rows where the total project price is null or less than or equal to zero, ensuring that the subsequently generated first feature vector is meaningful. Subsequently, for the remaining NaN in the feature columns or missing columns found during data loading, the system will uniformly fill them with the value "0", which ensures that the subsequently generated first feature vector is complete and clean. The system will also perform preliminary feature engineering. For example, in the steel lining construction mode, the system will automatically calculate and generate a new feature—"the proportion of actual costs of the six major factors to the total cost"—based on various core costs and the total project price, providing a basis for subsequent ratio-based prediction.

[0059] Furthermore, in order to eliminate the influence of differences in units, StandardScaler can be used for data standardization.

[0060] In some implementations, before determining the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and their corresponding target estimation rule base, the method further includes: Obtain the target type corresponding to the first engineering quantity characteristic parameter; Based on the target type, the target estimation rule library is determined from multiple preset estimation rule libraries. The preset estimation rule library includes the preset average ratio between various quantities of work, the preset average unit price of various core costs, and the preset average proportion of core costs to the total cost.

[0061] For example, the most core engineering quantity (such as the total tonnage of steel bars) is extracted from the first engineering quantity feature parameter as a key feature, and it is compared with a pre-trained K-Means clustering model to determine the target type corresponding to the first engineering quantity feature parameter.

[0062] For example, the preset estimation rule library corresponding to the target type among multiple preset estimation rule libraries is determined as the target estimation rule library.

[0063] For example, if the first quantity feature parameter cannot be explicitly matched to any preset type, that is, there is no target type corresponding to the first quantity feature parameter, then the global average rule base is used as the target rule base to ensure the robustness of subsequent prediction processes. The global average rule base is the estimation rule base corresponding to historical quantity feature parameters.

[0064] In some implementations, before obtaining the target type corresponding to the first engineering quantity characteristic parameter, the method further includes: Obtain historical project quantity characteristic parameters; historical project quantity characteristic parameters are historical parameters used to characterize the project quantity and cost characteristics under different construction modes; Clustering algorithms are used to group the characteristic parameters of historical engineering quantities, resulting in at least two preset types; A preset estimation rule library is generated for each preset type and historical engineering quantity characteristic parameter to obtain multiple preset estimation rule libraries.

[0065] For example, historical engineering quantity characteristic parameters may include historical parameters of relevant engineering quantities such as tower crane rental, steel bar production, and the number of slings and sleeves in the steel cage construction mode; historical key engineering quantities such as assembly site, jig fabrication, steel supports, buttresses, and walkway slabs in the steel lining construction mode; and other custom historical parameters in the custom mode.

[0066] For example, the K-Means clustering algorithm can be used to intelligently cluster historical engineering quantity characteristic parameters. Taking the rebar cage construction mode as an example, the total tonnage of rebar and the number of sleeves, two key features that best reflect the project scale and complexity, are selected as the clustering basis. The historical engineering quantity characteristic parameters are divided into a first preset type and a second preset type (i.e., at least two preset types). The first type represents projects with relatively low rebar and sleeve usage, usually corresponding to projects with smaller scale or relatively simple structures; the second type represents projects with significantly high rebar and sleeve usage, which is usually associated with larger-scale or more complex engineering structures with denser rebar. After clustering, an independent estimation rule base (i.e., a preset estimation rule base) is calculated and generated for each preset type and historical engineering quantity characteristic parameter (as a global rule). This estimation rule base includes key parameters such as "the average ratio between various engineering quantities," "the average unit price of various costs," and "the average proportion of core costs to the total cost." Price information comes from historical project data (i.e., historical engineering quantity characteristic parameters). This project clustering based on actual usage provides core data support for subsequent prediction models to adopt the most accurate and suitable estimation rules for projects of different sizes.

[0067] In step S130, for example, the target ensemble model includes a target ridge regression model, a target decision tree model, a target random forest model, a target support vector regression model, and a target neural network. At this point, there are corresponding construction price predictions for five target construction modes.

[0068] In other words, the target ensemble model employs multiple algorithms, including Linear Regression, Random Forest Regressor, Support Vector Regression (SVR), Decision Tree Regressor, and Multi-Level Processing (MLP) Regressor. Each algorithm has unique advantages in processing different types of engineering data. For example, Linear Regression is suitable for simple linear relationships, providing fast and easily interpretable predictions; Random Forest Regression and SVR can effectively handle nonlinear data and exhibit strong robustness and adaptability when facing complex data; Decision Tree Regression clearly displays the decision-making process through a tree structure; and Multi-Level Processing (MLP) Regression can capture complex nonlinear patterns in the data, making it particularly suitable for large-scale and complex engineering data.

[0069] For example, the first feature vector is simultaneously input into all pre-trained machine learning regression models (including ridge regression, decision tree, random forest, support vector regression, and neural network). Each model outputs an independent project total price prediction (i.e., construction price prediction).

[0070] The training process of the target ensemble model is described below.

[0071] 1) Internal data aggregation and updating within the database

[0072] The data processing flow of this application embodiment aims to ensure the quality, consistency, and accuracy of the input data, which is the cornerstone of subsequent high-precision model training. This process is not a simple file import, but an automated data pipeline that includes multiple stages such as internal database data aggregation, data cleaning and verification, feature engineering, and standardization. The system's data comes from a structured MySQL database, mainly involving two types of core data tables: the "Price Benchmark" table (e.g., Price Benchmark 1, Price Benchmark 2) and the "Key Factors" table (e.g., Key Factor 1, Key Factor 2). Before model training, a key data aggregation preprocessing step is first executed, which is implemented by the `process_and_update_databases` function. The system reads detailed material quantity data from the "Price Benchmark" table, filters out core parameters (e.g., "reinforcement tonnage") according to the construction mode, and summarizes and sums the quantities of these parameters by project. Subsequently, the calculated total quantity data is updated back to the corresponding project's "Key Factors" table; for example, the summarized total reinforcement tonnage is updated to the "Reinforcement Tonnage" column of the "Key Factor 1" table. Through this automated process, the system integrates scattered detailed data into highly structured key feature data that can be directly used for machine learning, completing the first step of data collection and preparation.

[0073] 2) Data loading and deep preprocessing

[0074] After data aggregation is complete, the core CostPredictionSystem class begins loading data and performing deeper preprocessing, a process automatically completed in the background by the _preprocess_data method. Data is loaded from the updated "Key Factors" table, and the pandas library's to_numeric method is used to strictly convert all numerical feature columns (such as various costs, quantities, and total project price). For non-numerical entries that cannot be converted, the system will force them to NaN (invalid value). The system first identifies and removes invalid project records, i.e., rows where the total project price is null or less than or equal to zero, ensuring that the model only learns from meaningful historical data. Subsequently, for any remaining NaNs in the feature columns or missing columns discovered during data loading, the system will uniformly fill them with "0" values, ensuring that the feature matrix input to the machine learning model is complete and clean. The system also performs preliminary feature engineering. For example, in the "steel lining" mode, the system will automatically calculate and generate a new feature—"the proportion of actual costs of the six major factors to the total cost"—based on the core costs and the total project price, providing a basis for subsequent ratio-based predictions.

[0075] 3) Data standardization

[0076] To eliminate the impact of dimensional differences between different features on model training, the system employs StandardScaler for data standardization at specific stages. Before performing K-Means clustering analysis, the system standardizes the core features used for clustering (such as "total tonnage of rebar" and "number of sleeves") to ensure that each feature has the same weight when calculating distance, thereby improving the accuracy of clustering. When training machine learning models, the system incorporates standardization as part of its pipeline. For algorithms sensitive to data scale, such as Ridge Regression (RidgeCV), Support Vector Regression (SVR), and Multi-Level Perimeter Regressor (MLPRegressor), data is automatically standardized before entering model training, which significantly improves the model's convergence speed and prediction accuracy. Through this series of automated, multi-stage data processing steps, the embodiments of this application ensure the standardization, consistency, and high quality of input data, laying a solid foundation for subsequent multi-model parallel training and high-precision intelligent prediction.

[0077] To ensure the model can be trained effectively and has good generalization ability, the data is first standardized using StandardScaler. Standardization transforms the values ​​of all features into the same scale range, avoiding the adverse effects of feature scale differences on model training. This is especially important for models that are sensitive to data scale, such as support vector regression and neural networks.

[0078] During multi-model training, the `train_test_split` function splits the data into training and test sets, ensuring that models are trained and validated on different datasets. Furthermore, the system employs cross-validation techniques (such as GridSearchCV) to optimize the hyperparameters of each model, finding the optimal hyperparameter combination through grid search, thereby improving model performance and prediction accuracy.

[0079] Five different machine learning algorithms, each with specific advantages and applicable scenarios, are integrated into this system to achieve high-precision construction cost prediction. These algorithms include Ridge Regression, Decision Tree, Random Forest, Support Vector Machine, and Neural Network. During training, they all follow a unified supervised learning logic: using the actual costs of key components in historical projects (such as tower crane and rebar production line costs) as input features, and the total project price as the target label. The prediction model is built by learning the complex mapping relationship between component costs and the total price. To address the data characteristics of different construction modes and provide diverse prediction perspectives, the system specifically integrates the following five algorithms: Ridge regression (RidgeCV) uses L2 regularization to handle multicollinearity and automatically selects the optimal regularization parameter α through cross-validation. Its objective function is:

[0080] in, This is the actual value. It is an eigenvalue. It is the regression coefficient. is the regularization strength parameter, N is the number of samples, and p is the number of features (i.e., the number of independent variables). This is the loss function for Ridge Regression. It is particularly suitable for situations where multicollinearity exists among features. Decision Tree Regression constructs tree-like decision rules by recursively splitting the dataset, selecting features and split points that minimize the mean squared error of child nodes, and has good interpretability. Random Forest Regression constructs multiple decision trees and averages their predictions, providing better generalization ability and robustness than a single decision tree. Support Vector Regression is based on an ε-insensitive loss function, searching for the optimal regression hyperplane, and is suitable for handling high-dimensional data and nonlinear relationships. Multilayer Perceptron Regression uses a feedforward neural network, trained through a backpropagation algorithm, and can capture complex nonlinear relationships, making it particularly suitable for processing large-scale complex data.

[0081] In step S140, in some embodiments, the estimated target construction price corresponding to the target construction mode is determined based on N predicted construction prices, including: The average of the N predicted construction prices is used as the target estimated construction price.

[0082] In this way, the advantages of at least two models can be combined to improve the accuracy of construction price predictions.

[0083] In other implementations, the target construction price estimate corresponding to the target construction mode is determined based on N predicted construction price values, including: The ratio method was used to verify the results based on the average of N construction price predictions. When the verification results are used to indicate that the verification has passed, the average value is determined as the target construction price estimate.

[0084] For example, the calculation formula for ratio method verification includes: Ratio Method Predicted Total Price = (Sum of all core costs) / (Average percentage of each core cost in the historical target estimation rule base). Here, the average percentage represents the average percentage of the sum of all core costs to the total project cost in historical projects (i.e., historical quantity characteristic parameters). This result provides users with a traditional estimation reference based on historical cost structures.

[0085] For example, if both the integrated average projected total price (i.e., the target construction price estimate) and the ratio-based projected total price are within a preset range, the verification result is determined to indicate that the verification has passed. The preset range can be set according to actual conditions and is not limited here. If at least one of the target construction price estimate and the ratio-based projected total price is not within the preset range, the verification result is determined to indicate that the verification has failed.

[0086] Furthermore, if the verification result is used to indicate that the verification failed, it is usually handled as follows: 1) Adjust the model or ratio: Optimize the prediction model or adjust the parameters of the ratio method according to the actual situation; 2) Data inspection: Check if there are any problems with the input data (i.e., the first engineering quantity characteristic parameters).

[0087] In some implementations, after determining the estimated target construction price corresponding to the target construction mode based on N predicted construction prices, the method further includes: Visualize the estimated construction cost of the target project.

[0088] This allows users to intuitively obtain the estimated cost of the target construction project.

[0089] To better understand the construction cost estimation method provided in the embodiments of this application, the following description is provided in conjunction with specific implementation methods.

[0090] The construction cost estimation method provided in this application can be executed through a modular construction price prediction system based on multi-algorithm integration. This system adopts a four-layer architecture design, combining advanced artificial intelligence algorithms and big data processing technology to achieve high-precision, intelligent cost estimation. Through the integration of multiple models, the embodiments of this application can combine the advantages of each model to provide optimized prediction results under different scenarios. Whenever new data is input into the system, the system automatically compares the performance of multiple models and selects the most suitable one for price prediction. This not only improves the accuracy of prediction but also allows for flexible adjustment of the algorithm under different project needs and market environments, ensuring accurate cost estimation for each engineering task. The overall solution is as follows: Figure 2As shown, the process begins with "data input and preprocessing," followed by "artificial intelligence model prediction" to "generate prediction results," and finally "result output." The process includes a data collection phase, namely "collecting construction project data," followed by "selecting a model (regression, random forest, neural network)," and then "user input of project parameters" to "output prediction results." In terms of data processing, there are steps of "data cleaning and outlier removal," followed by "model training and optimization," and then "interacting with the AI ​​model" for "chart and report display." In addition, there is a "data standardization" phase, followed by "cross-validation and hyperparameter tuning," and then "real-time feedback and adjustment" to "export results (PDF / Excel)." Finally, after "preparing data input for the AI ​​model," "generating prediction results," followed by "displaying result charts," and finally "supporting batch processing of multiple projects." The entire process is connected by arrows, presenting the complete process from data preparation to result output. The figure illustrates the construction task cost prediction process implemented through a multi-model ensemble algorithm, further strengthening the advantages of this application's embodiment in improving cost estimation accuracy and response speed.

[0091] The system adopts a system overall architecture design.

[0092] The system adopts a layered architecture design, such as Figure 3 As shown, the overall structure comprises four layers: presentation layer, application layer, domain layer, and infrastructure layer. The presentation layer, implemented using the Python Dash framework, is responsible for direct user interaction, providing a user interface, data visualization, and report display functions, and supporting interactive data exploration and the generation of standardized technical and economic documents. The application layer implements core business functions, including intelligent cost estimation models, component indicator library management, and rapid investment estimation. It integrates multiple artificial intelligence algorithms to provide accurate cost estimates, maintains and updates indicator data for various construction components, and supports rapid investment assessment in the early stages of projects. The domain layer carries the implementation of core business logic and domain rules, including AI algorithm services, data processing services, and a business rule engine. It provides model training and optimization functions, is responsible for data cleaning, transformation, and analysis, and maintains and executes various business rules to ensure the accuracy of calculation results. The infrastructure layer provides underlying technical support and basic service capabilities, including a MySQL database for storing core business data and historical records, a caching service for high-speed data access, file storage for managing various documents and reports, and a message queue for handling asynchronous tasks to improve system concurrency.

[0093] System functional structure as follows Figure 4As shown, the front-end interface includes basic indicator calculation, composite indicator calculation, and comprehensive indicator calculation. Users can input data and customize calculation rules to obtain the required indicators. The system manages a multi-level indicator database, covering indicator storage, updates, and version control. The intelligent analysis algorithm model library supports model training and maintenance. Users can manage roles and permissions and query operation logs to audit system activities. The data acquisition and management service is responsible for defining and cleaning data sources. The back-end services include user management, indicator calculation, intelligent analysis algorithms, and automatic data import. The user management service handles registration, authentication, and access control. The indicator calculation service performs calculations of basic, composite, and comprehensive indicators based on user input and calculation rules to support front-end requirements. The intelligent analysis algorithm service is responsible for model training, storage, and updates to ensure algorithm effectiveness. The automatic data import suite supports data access from multiple sources, including structured and unstructured data, covering general data parsing, cleaning, and quality control. Finally, the parsed data is imported into the multi-level indicator database to improve retrieval efficiency.

[0094] Furthermore, the embodiments of this application also possess the ability to continuously learn and optimize, adaptively adjusting and optimizing the prediction model by continuously inputting new project data. Thus, over time, the system's predictive ability will continuously improve, more accurately reflecting market price trends and construction cost fluctuations. The technical solution of the embodiments of this application mainly includes the following steps: 1. Data Collection and Preprocessing 1) Internal data aggregation and updating within the database The data processing flow of this application embodiment aims to ensure the quality, consistency, and accuracy of the input data, which is the cornerstone of subsequent high-precision model training. This process is not a simple file import, but an automated data pipeline that includes multiple stages such as internal database data aggregation, data cleaning and verification, feature engineering, and standardization. The system's data comes from a structured MySQL database, mainly involving two types of core data tables: the "Price Benchmark" table (e.g., Price Benchmark 1, Price Benchmark 2) and the "Key Factors" table (e.g., Key Factor 1, Key Factor 2). Before model training, the system first performs a crucial data aggregation preprocessing step, which is implemented by the `process_and_update_databases` function. The system reads detailed material quantity data from the "Price Benchmark" table, filters out core parameters (e.g., "reinforcement tonnage") according to the construction mode, and summarizes and sums the quantities of these parameters by project. Subsequently, the system updates the calculated total quantity data back to the corresponding project's "Key Factors" table; for example, it updates the total reinforcement tonnage column of the "Key Factor 1" table with the summarized total reinforcement tonnage. Through this automated process, the system integrates scattered detailed data into highly structured key feature data that can be directly used for machine learning, completing the first step of data collection and preparation.

[0095] 2) Data loading and deep preprocessing

[0096] After data aggregation is complete, the core CostPredictionSystem class begins loading data and performing deeper preprocessing, a process automatically completed in the background by the _preprocess_data method. The system loads data from the updated "Key Factors" table and uses the pandas library's to_numeric method to perform strict numeric type conversion on all numerical feature columns (such as various costs, quantities, total project price, etc.). For non-numerical entries that cannot be converted, the system forcibly converts them to NaN (invalid value). The system first identifies and removes invalid project records, i.e., rows where the total project price is null or less than or equal to zero, ensuring that the model only learns from meaningful historical data. Subsequently, for any remaining NaN (invalid values) in the feature columns or missing columns discovered during data loading, the system uniformly fills them with "0" values, ensuring that the feature matrix input to the machine learning model is complete and clean. The system will also perform preliminary feature engineering. For example, in the "steel lining" mode, the system will automatically calculate and generate a new feature based on the core costs and the total project price - "the proportion of actual costs of the six major factors to the total cost", providing a basis for subsequent ratio method prediction.

[0097] 3) Data standardization

[0098] To eliminate the impact of dimensional differences between different features on model training, the system employs StandardScaler for data standardization at specific stages. Before performing K-Means clustering analysis, the system standardizes the core features used for clustering (such as "total tonnage of rebar" and "number of sleeves") to ensure that each feature has the same weight when calculating distance, thereby improving the accuracy of clustering. When training machine learning models, the system incorporates standardization as part of its pipeline. For algorithms sensitive to data scale, such as Ridge Regression (RidgeCV), Support Vector Regression (SVR), and Multi-Level Perimeter Regressor (MLPRegressor), data is automatically standardized before entering model training, which significantly improves the model's convergence speed and prediction accuracy. Through this series of automated, multi-stage data processing steps, the embodiments of this application ensure the standardization, consistency, and high quality of input data, laying a solid foundation for subsequent multi-model parallel training and high-precision intelligent prediction.

[0099] 2. Multi-model training and optimization

[0100] This application provides a complete and efficient modular construction rapid cost estimation system. Its core functions are presented through an intuitive human-computer interaction interface. Users can select from three core prediction modes based on specific project needs to initiate a targeted intelligent cost estimation process. For example... Figure 5 As shown, these three modes are: Reinforcing cage construction mode: Designed specifically for modular construction projects of reinforced concrete structures, users can input relevant engineering parameters such as tower crane rental, rebar production, and the quantity of slings and sleeves in this mode. The system can then call up a dedicated prediction model to perform accurate and intelligent analysis and prediction of the total cost of the reinforcing cage project.

[0101] Steel lining construction mode: Applicable to projects that include steel lining structures. Users can input key quantities such as assembly site, jig fabrication, steel supports, buttresses, and walkway slabs. The system will then initiate cost calculations and machine learning predictions based on the complexity of the steel lining construction, providing a comprehensive cost analysis report.

[0102] Customized construction mode: To meet special or non-standardized project needs, the system provides a highly flexible customized mode, allowing users to define various construction parameters and quantities of the project according to the actual situation, so as to realize personalized cost estimation for specific engineering scenarios, which greatly enhances the system's versatility and adaptability.

[0103] Internally, the system employs a variety of algorithms, including Linear Regression, Random Forest Regressor, Support Vector Regression (SVR), Decision Tree Regressor, and Multi-Level Processing (MLP) Regressor. Each algorithm has unique advantages in processing different types of engineering data. For example, Linear Regression is suitable for simple linear relationships, providing fast and easily interpretable predictions; Random Forest Regression and SVR can effectively handle nonlinear data and exhibit strong robustness and adaptability when facing complex data; Decision Tree Regression clearly displays the decision-making process through a tree structure; and Multi-Level Processing (MLP) Regression can capture complex nonlinear patterns in the data, making it particularly suitable for large-scale and complex engineering data.

[0104] To ensure the model can be trained effectively and has good generalization ability, the system first uses StandardScaler to standardize the data. Standardization transforms the values ​​of all features into the same scale range, avoiding the adverse effects of feature scale differences on model training. This is especially important for models that are sensitive to data scale, such as support vector regression and neural networks.

[0105] During multi-model training, the system uses the `train_test_split` function to split the data into training and test sets, ensuring that models are trained and validated on different datasets. Furthermore, the system employs cross-validation techniques (such as GridSearchCV) to optimize the hyperparameters of each model, finding the optimal hyperparameter combination through grid search, thereby improving model performance and prediction accuracy. The overall process is as follows: Figure 6 As shown. This is a schematic diagram of a machine learning model training architecture, illustrating the structure of a typical regression model training system: Top-level class: ModelTraining This is the core training class, which contains the following methods: StandardScaler() - Data Standardization Method train_test_split() - Method for splitting the training and test sets GridSearchCV() - Grid search cross-validation for hyperparameter tuning get_residuals() - A method for retrieving residuals The five underlying regression model classes (inheritance relationship) From left to right: Linear Regression fit() - Training the model predict() - prediction score() - rating intercept(float) - Intercept property RandomForestRegressor fit() - Training the model predict() - prediction feature_importances() - Feature Importance SVR (Support Vector Regression) fit() - Training the model predict() - prediction n_support(ndarray) - Number of support vectors DecisionTreeRegressor fit() - Training the model predict() - prediction visualize(decision.tree()) - Visualize decision trees MLPRegressor (Multilayer Perceptron Regression) fit() - Training the model predict() - prediction train.OnDefaultlayer() - Trains the default layer hidden.LayerSIZES() - Size of the hidden layer Architecture Description Figure 6 This demonstrates a strategy or factory pattern design, where the ModelTraining class acts as the manager, flexibly using five different regression algorithms. Arrows indicate inheritance or dependency relationships; all regression models can be called by the ModelTraining class.

[0106] 3. Core Algorithm Module Design

[0107] The core of the system is a price prediction module implemented based on the CostPredictionSystem class. This module integrates a complete prediction engine, with its key feature being an adaptive prediction mechanism based on project clustering. This mechanism ensures that projects of different sizes and types can be matched with the most suitable estimation rules, thereby significantly improving the accuracy and reliability of the predictions. Taking the steel cage construction mode as an example, the steel lining construction mode is also carried out in this way.

[0108] 1) Intelligent project grouping: the foundation for building a targeted rule base

[0109] During the model training phase, the system employs the K-Means clustering algorithm to intelligently group massive amounts of historical project data. Taking "reinforcing cage construction mode" as an example, the system selects "total tonnage of reinforcing bars" and "number of sleeves" as the two key features that best reflect the scale and complexity of the project as the basis for clustering.

[0110] like Figure 7 As shown, K-Means clustering analysis clearly reveals the distribution patterns of historical project data across the two dimensions of "total tonnage of rebar" and "number of sleeves." Using a K=2 clustering algorithm, historical projects were automatically divided into two distinct groups: Cluster 0 (i.e., the first preset type): represents projects with relatively small amounts of steel bars and sleeves, usually corresponding to projects with small scale or relatively simple structure.

[0111] Cluster 1 (the second preset type): This includes projects with significantly higher amounts of steel reinforcement and sleeves, which are usually associated with larger-scale or more complex engineering structures with denser steel reinforcement.

[0112] After clustering, the system calculates and generates an independent "estimation rule base" for each cluster and for all data (as global rules). This rule base includes key parameters such as "average ratios between various project quantities," "average unit prices of various costs," and "average proportions of core costs to total costs." Price information comes from historical project data. This project clustering based on actual usage provides core data support for subsequent prediction models to adopt the most accurate and suitable estimation rules for projects of different sizes.

[0113] 2) Predictive Execution Process: From User Input to Intelligent Analysis Report

[0114] When a user makes a new price prediction in a real-world application, the system will strictly follow the following automated process: Step 1: User Input and Data Collection The user first interacts with the interface (such as...) Figure 9 As shown, select a construction mode (e.g., "reinforcing cage construction mode" or "steel lining construction mode") and input the known engineering quantity parameters. Thanks to the system's intelligent estimation capabilities, users do not need to provide all parameters, but only need to input one or a few core engineering quantities (e.g., only inputting "total tonnage of reinforcing bars" as 22 tons). The system collects this core data through the collect_user_quantities function and starts the prediction engine.

[0115] Step 2: Intelligent Matching and Rule Set Selection

[0116] This is a crucial step in connecting project clustering with price forecasting. The system calls the `_match_cluster_rules` method to match the core quantities of work input by the user (such as the total tonnage of 22 tons of steel reinforcement) with... Figure 7 The system matches the input value to the clustering model shown in the image. It determines which historical project cluster the input value belongs to (e.g., 22 tons belongs to cluster 0, which has lower usage). Once a match is found, the system automatically loads and applies the dedicated "estimation rule base" corresponding to cluster 0. If the user-input data features cannot explicitly match any cluster, the system will activate a backup "global average rule base" to ensure the robustness of the prediction process.

[0117] Step 3: Intelligent Engineering Quantity Completion

[0118] After determining the rule base to be used, the system enters the _estimate_quantities stage to intelligently complete any quantities not provided by the user. For example, if the user only enters "total tonnage of rebar," the system will automatically estimate all other necessary quantities (such as tower crane rental quantities, sling quantities, and sleeve quantities) based on rules in the "Cluster 0" rule base, such as "average ratio of sleeve quantity to total rebar tonnage" and "average ratio of tower crane rental quantity to total rebar tonnage." The results of this step will be clearly displayed in the "Intelligent Quantity Completion" section of the final report (e.g., ...). Figure 10 (As shown).

[0119] Formula for average work volume ratio: 1. Average ratio of sleeve quantity to total tonnage of reinforcing bars: Average number of sleeves per ton of reinforcing bars 2. Average ratio of tower crane rental work volume to total steel reinforcement tonnage: Average tower crane rental work volume per ton of steel reinforcement. 3. Average ratio of lifting slings to total tonnage of reinforcing steel: Average number of lifting slings per ton of reinforcing steel. If the user provides the "total tonnage of steel bars", the other unprovided quantities are estimated as follows: Tower crane rental quantity = total tonnage of steel bars × average tower crane rental quantity per ton of steel bars, number of slings = total tonnage of steel bars × (corresponding average ratio), number of sleeves = total tonnage of steel bars × (corresponding average ratio). If the user does not provide the "total tonnage of steel bars", the system will estimate it in reverse: Total tonnage of steel bars = tower crane rental project volume ÷ average tonnage of steel bars per unit of tower crane rental project volume; Step 4: Core Cost Calculation Once all project quantity parameters are ready, the system will calculate the core costs based on the "average unit cost" in the rule base. For example: Tower crane rental cost = Estimated tower crane rental workload × Average unit tower crane rental cost Rebar production line cost = Total tonnage of rebar input by user × Average unit cost of rebar production line ... (and so on, to calculate the feature cost required for all ML models) Step 5: Multi-model ensemble prediction and ratio method validation The calculated core costs form a feature vector, which is simultaneously input into all pre-trained machine learning regression models (including ridge regression, decision trees, random forests, support vector regression, and neural networks). Each model outputs an independent project total price prediction. The system further calculates the arithmetic mean of these effective predictions to form a more robust "ensemble average prediction" total price. Simultaneously, the system uses the "average proportion of core costs to total cost" from the rule base for ratio-based prediction as cross-validation.

[0120] The algorithm module includes five different machine learning algorithms, each with specific advantages and applicable scenarios. To achieve high-precision construction cost prediction, this system integrates five machine learning algorithms: Ridge Regression, Decision Tree, Random Forest, Support Vector Machine, and Neural Network. During training, all five follow a unified supervised learning logic: using the actual costs of key components in historical projects (such as tower crane and rebar production line costs) as input features, and the total project price as the target label, they learn the complex mapping relationship between component costs and the total price to build a prediction model. To address the data characteristics of different construction modes and provide diverse prediction perspectives, the system specifically integrates the following five algorithms: Ridge regression (RidgeCV) uses L2 regularization to handle multicollinearity and automatically selects the optimal regularization parameter α through cross-validation. Its objective function is:

[0121] in, This is the actual value. It is an eigenvalue. It is the regression coefficient. This is the regularization strength parameter. It is particularly suitable for situations where multicollinearity exists among features. Decision tree regression constructs tree-like decision rules by recursively splitting the dataset, selecting features and split points that minimize the mean squared error of child nodes, and has good interpretability. Random forest regression constructs multiple decision trees and averages their predictions, providing better generalization ability and robustness than a single decision tree. Support vector regression (SVR) is based on an ε-insensitive loss function, searching for the optimal regression hyperplane, and is suitable for handling high-dimensional data and nonlinear relationships. Multilayer perceptron regression uses a feedforward neural network, trained through a backpropagation algorithm, and can capture complex nonlinear relationships, making it particularly suitable for processing large-scale complex data.

[0122] 4. Predictive Execution Mechanism and Algorithm Optimization

[0123] The prediction execution in this application employs a rigorous multi-stage automated processing mechanism to ensure intelligence, accuracy, and reliability throughout the entire process from user input to final report generation. This process is driven by the core callback function in the callback.py module and deeply integrates the complex algorithm logic of the CostPredictionSystem class in data.py.

[0124] Phase 1: User Input Processing and Hierarchical Data Collection

[0125] The prediction process begins with the user inputting known engineering quantity data through the interactive interface. The system's data collection mechanism (implemented by the `collect_user_quantities` function) has hierarchical processing capabilities, providing users with great flexibility. Prioritize the adoption of total project quantity by category: The system interface allows users to directly enter the total project quantity for a broad category, for example, the total project quantity under the category "Tower Crane Rental Fee". If the user provides this data, the system will prioritize it and use it as the final project quantity for that category.

[0126] Automatic aggregation of subdivided parameters: If the user does not provide the total quantity of work for a category, but instead fills in multiple subdivided parameters for that category, such as the quantities of “self-erecting tower crane (1500KNm)” and “heavy tower crane”, the system will automatically sum up the values ​​of these subdivided parameters to aggregate the total quantity of work for that category.

[0127] This design allows users to input both macro-level and granular data, and the system can intelligently process it into the standardized data format required for subsequent steps. At this stage, all relevant inputs, including key parameters such as "measures cost and engineering quantity," are collected.

[0128] Phase Two: Intelligent Rule Matching Based on Historical Project Clusters

[0129] After data collection is complete, the system enters the intelligent matching phase (executed by the `_match_cluster_rules` method). The core of this phase is to match the most suitable "estimation rule set" for the item to be predicted. These rule sets are pre-calculated and stored during the model training phase after performing K-Means clustering analysis on historical data, and are mainly divided into two categories: Cluster-Specific Rules: A dedicated rule library for different project groups identified through cluster analysis (e.g., "Cluster 0" represents small-scale projects, "Cluster 1" represents large-scale projects). Each library contains key statistical indicators unique to that type of project, such as "average ratios between various quantities of work," "average unit prices of various costs," and "average proportion of core costs to total cost."

[0130] Global Average Rules: A general rule base calculated based on all historical data. It serves as a backup plan when no specific project cluster can be matched, ensuring the robustness of the prediction process.

[0131] During actual prediction, the system extracts the most core engineering quantities input by the user (such as "total tonnage of steel bars") as key features, compares them with a pre-trained K-Means clustering model, and determines which historical project cluster the current project should belong to. Subsequently, the system automatically loads the dedicated rule base corresponding to that cluster, providing the most accurate reference benchmark for subsequent intelligent estimation.

[0132] Phase 3: Rule-based Intelligent Engineering Quantity Estimation

[0133] After determining the estimation rule set, the system enters the quantity estimation phase (executed by the _estimate_quantities method) to ensure that all quantity parameters necessary for cost calculation have valid values. This process fully demonstrates the system's intelligence. Directly use user input: For quantities of work already provided by the user, the system directly uses their values.

[0134] Intelligent estimation of unprovided items: For quantities of work not provided by the user, the system will intelligently estimate them using the "average ratio" in the loaded rule base. For example, in "reinforcing cage" mode, if the rule base already contains the "average ratio of sleeve quantity to total rebar tonnage," and the user only enters the "total rebar tonnage," the system will automatically estimate the sleeve quantity using the following formula: Estimated sleeve quantity = total tonnage of steel bars input by the user × ratio of (average sleeve quantity / average total tonnage of steel bars) in the rule base; Through this mechanism, even if the user provides only limited input, the system can intelligently complete a complete and reasonable bill of quantities, laying the foundation for the next step of cost calculation.

[0135] Phase Four: Modular Core Cost Calculation

[0136] This stage calculates the core cost features required for the machine learning model based on the complete bill of quantities generated in the previous step and the "average unit cost" in the rule base. This calculation process is modular, employing different cost combinations for different construction modes (such as rebar cages and steel linings). Taking the "rebar cage construction mode" as an example, the system will calculate four core costs: Tower crane rental cost = (estimated) tower crane rental volume × (average unit tower crane rental cost in the rule base); Rebar production line cost = (estimated) total tonnage of rebar × (average unit cost of rebar production line in the rule base); Sling cost = (estimated) number of slings × (average unit cost of slings in the rule base); Sleeve cost = (estimated) number of sleeves × (average unit sleeve cost in the rule base); These four costs constitute the input feature vector for predicting the total price of the "steel cage" model. They are not the total price of the project themselves, but rather key intermediate variables that reflect the cost composition of the project.

[0137] Phase 5: Multi-model ensemble prediction and cross-validation

[0138] Finally, the system enters the multi-model prediction stage. The core of this stage lies in using the four core costs calculated in the previous steps as feature vectors, which are then input into all the trained machine learning regression models. These models (such as ridge regression, decision trees, and random forests) independently predict the final project price by learning the complex nonlinear relationship between the "four core costs" and the final "total project price" in historical data.

[0139] Integrated forecasting: The system aggregates all the valid forecasts given by the models and calculates their arithmetic mean to generate a statistically more robust and reliable "integrated average forecast" total price.

[0140] Ratio Method Validation: Simultaneously, the system employs the traditional ratio method for total price prediction as an independent supplement and validation. The calculation formula is as follows: The ratio method predicts the total price as follows: (sum of the four core costs) / (average percentage of the four core costs in the historical rule base). The "average percentage" here represents the average percentage of the total project cost that these four core costs typically account for in historical projects. This result provides users with a traditional estimation reference based on historical cost structures.

[0141] Generally, if the combined average predicted total price and the ratio-based predicted total price are both within the preset range, the verification is considered successful.

[0142] If the verification fails, the usual procedure is as follows: Adjust the model or ratio: Optimize the prediction model or adjust the parameters of the ratio method according to the actual situation.

[0143] Data inspection: Check if there are any problems with the input data.

[0144] This system employs Leave-One-out Cross-Validation (LOOCV) to rigorously evaluate model performance, comprehensively measuring model quality through multiple evaluation metrics. As shown in Figures 8(a) to 8(c), the performance comparison results of various regression models under LOCV show that the Ridge Regression (RRV) model has the best overall performance, with a Mean Absolute Error (MAE) of 960,000 yuan, a Root Mean Square Error (RMSE) of 1,150,000 yuan, and an R² value as high as 0.96, indicating that its predictions are highly correlated with actual values ​​and have the smallest error. The Decision Tree model, as the second tier, has an MAE of approximately 2,600,000 yuan, an RMSE of approximately 4,180,000 yuan, and an R² value of 0.49. Random Forest and Ensemble Average perform similarly, with MAEs between 3,160,000 and 3,200,000 yuan and R² values ​​of 0.38 and 0.36, respectively. Support Vector Regression and Neural Networks, however, perform relatively poorly and are unsuitable for the current prediction task.

[0145] Further analysis reveals that the ridge regression model's outstanding performance is primarily due to its effective handling of multicollinearity and the overfitting effect of the L2 regularization term. With relatively limited sample sizes, ridge regression introduces bias to significantly reduce variance, resulting in more stable predictive performance. In contrast, neural networks and support vector regression models are prone to overfitting in small sample environments, leading to insufficient generalization ability. While decision tree models offer good interpretability, single decision trees are susceptible to overfitting, and their performance is significantly inferior to ridge regression. Ensemble methods such as random forests, although theoretically capable of improving prediction accuracy, have not fully realized their advantages on the current dataset, possibly due to factors such as the number of features and sample distribution characteristics. Considering prediction accuracy, model stability, and computational complexity, ridge regression has been identified as the optimal prediction model, laying a solid foundation for future practical applications.

[0146] The system establishes a continuous learning and adaptive optimization mechanism, enabling it to continuously improve model performance based on new project data. Model optimization employs an incremental learning strategy, periodically retraining the model using newly collected project data. By comparing the performance of the new model with existing models on the validation set, the optimal model is automatically selected for deployment. The system also implements model version management, saving model parameters and performance metrics for different versions, and supporting model rollback and A / B testing. The dynamic optimization mechanism automatically adjusts model weights and prediction strategies based on the deviation between prediction results and actual data, ensuring the system consistently provides accurate predictions under ever-changing market conditions.

[0147] 5. Support for multiple construction modes and result display

[0148] like Figure 9 As shown, the system's user interface is simple and intuitive, providing users with a convenient operating experience. The interface adopts a modular layout, allowing users to select the appropriate construction mode based on project requirements, input relevant engineering quantity parameters, and the system will automatically perform calculations and display the results. Figure 10 , Figure 11 The prediction results.

[0149] The prediction results are displayed using a multi-dimensional strategy, providing users with comprehensive decision support information. During the result adjustment and final display phase, the prediction results are first adjusted for measure costs. The measure cost amount input by the user is added to all machine learning model prediction results and ratio method prediction results to ensure that the final total price prediction includes the directly specified measure costs. The system generates a detailed prediction result report, including a summary of user input clearly listing the main engineering quantity parameters input by the user; a rule matching source informing the user which historical project cluster's rules or global average rules this prediction is based on; and intelligent estimation details displaying the system's estimated missing engineering quantities or sub-item costs and their basis.

[0150] The core prediction results table details the total project price predicted by each individual machine learning algorithm, the ensemble average predicted total price of all machine learning algorithms, the ratio method used as a reference to validate the predicted total price, and the absolute and percentage deviations between each machine learning model's prediction and the ratio method's prediction, facilitating user comparison and analysis. The enhanced prediction summary provides a reliability assessment of the overall prediction results, compares the consistency between the ensemble machine learning prediction and the ratio method's prediction, and offers suggested ranges for the predicted total price, providing more valuable references for user decision-making.

[0151] The embodiments of this application have at least the following beneficial effects: 1) Prediction accuracy has been greatly improved Through multi-algorithm integration and intelligent clustering analysis, the system achieves a significant improvement in accuracy. The ridge regression model achieves an R² coefficient of determination of 0.961 in leave-one-out cross-validation, and the mean absolute error (MAE) is controlled within 961,015.85 yuan. The prediction accuracy meets the requirements of different stages: the initial estimate deviation is less than 5%. Compared with the 25%-40% prediction error of traditional methods, the embodiments of this application achieve a significant improvement in prediction accuracy, providing construction companies with a more reliable cost estimation tool.

[0152] 2) System overall performance optimization

[0153] Employing automated data processing and model training workflows, the system can complete cost estimations for complex projects within minutes, achieving efficiency improvements of tens of times compared to the days required by traditional manual methods. The system supports multiple modular construction modes and offers customizable configurations to adapt to the needs of different types of engineering projects. Its modular architecture provides excellent scalability, allowing for easy integration of new algorithms and functional modules. By providing functions such as multi-model prediction comparison, ratio method verification, and confidence analysis, the system offers users more comprehensive and reliable data support for decision-making. Intelligent report generation further enhances the system's usability. Accurate cost predictions help companies formulate reasonable bids during the bidding stage, effectively control costs during construction, significantly reduce cost estimation risks, and improve project profitability and market competitiveness.

[0154] 3) Construct a high-precision multi-algorithm integrated prediction model

[0155] This application aims to construct an integrated prediction system that incorporates multiple machine learning algorithms, including Ridge Regression (RidgeCV), Decision Tree Regressor (DecisionTreeRegressor), Random Forest Regressor (RandomForestRegressor), Support Vector Regression (SVR), and Multilayer Perceptron Regressor (MLP Regressor). It intelligently groups historical project data using the K-Means clustering algorithm, applying more targeted estimation rules and model parameters to different types of projects. The system employs Leave-One-out Cross-Validation (LOOCV) to evaluate model performance, ensuring that the prediction accuracy reaches: the initial estimated deviation of the calculation results is less than 5%. Experiments show that the Ridge Regression model has a mean absolute error (MAE) of 961,015.85 yuan and an R² coefficient of determination of 0.961 in the LOOCV evaluation.

[0156] 4) Achieve intelligent data processing and multi-source data integration

[0157] This application aims to build a powerful data processing engine capable of automatically handling data input in various formats (CSV, Excel, etc.), achieving intelligent data cleaning, standardization, and feature engineering. The system integrates a price benchmark database and a key factor database, ensuring data consistency and accuracy through a dynamic data preprocessing workflow. The system possesses multi-level data indicator management capabilities, supporting unified management of basic indicator libraries, composite indicator libraries, and comprehensive indicator libraries, enabling comprehensive estimation from individual component costs to overall project costs. It supports multiple modular construction modes and establishes a continuous learning and adaptive optimization mechanism.

[0158] This application's embodiments effectively solve the problems of low accuracy, poor efficiency, and weak adaptability in traditional modular construction cost estimation by introducing advanced artificial intelligence technology and big data processing methods, providing important technical support for the digital transformation of the construction industry.

[0159] It is understood that the various method embodiments mentioned above in this application can be combined with each other to form combined embodiments without violating the principle and logic. Due to space limitations, this application will not elaborate further. Those skilled in the art will understand that in the above methods of specific implementation, the specific execution order of each step should be determined by its function and possible internal logic.

[0160] Based on the same inventive concept, in a second aspect, embodiments of this application also provide a construction cost estimation device. For example... Figure 12 As shown, the construction cost estimation device provided in this application embodiment includes a first acquisition module 210, a first determination module 220, a first processing module 230, and a second determination module 240.

[0161] The first acquisition module 210 is used to acquire the target construction mode and its corresponding first engineering quantity characteristic parameters; the target construction mode includes one of the steel cage construction mode, steel lining construction mode and custom construction mode; the first engineering quantity characteristic parameters are parameters used to characterize the engineering quantity and cost composition characteristics of the target construction mode. The first determining module 220 is used to determine the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and its corresponding target estimation rule library; the first feature vector is related to the core costs corresponding to the target construction mode; the target estimation rule library includes the target average ratio between the engineering quantities, the target average unit price of the core costs, and the target average proportion of the core costs to the total cost under the target construction mode. The first processing module 230 is used to process the first feature vector using a target ensemble model to obtain the construction price prediction values ​​corresponding to N target construction modes; the target ensemble model includes N terms from the target ridge regression model, target decision tree model, target random forest model, target support vector regression model, and target neural network; N is a positive integer greater than or equal to 2; The second determining module 240 is used to determine the target construction price estimate corresponding to the target construction mode based on N construction price prediction values.

[0162] According to the construction cost estimation device provided in this application embodiment, the target construction mode and its corresponding first engineering quantity feature parameters are first obtained; then, based on the first engineering quantity feature parameters and its corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined; then, the first feature vector is processed using a target ensemble model to obtain N predicted construction prices corresponding to the target construction modes; and finally, based on the N predicted construction prices, the estimated target construction price corresponding to the target construction mode is determined. That is, in this application embodiment, since the target ensemble model includes N terms from a target ridge regression model, a target decision tree model, a target random forest model, a target support vector regression model, and a target neural network; where N is a positive integer greater than or equal to 2, processing the first feature vector using the target ensemble model to obtain N predicted construction prices corresponding to the target construction modes can, on the one hand, combine the advantages of at least two models to improve the accuracy of the predicted construction prices; on the other hand, by employing an advanced combination of machine learning algorithms, it can effectively cope with the complexity of modular construction projects and address the dynamic impact of factors such as market price fluctuations, changes in transportation distance, and differences in the construction site environment on costs, thereby improving the accuracy of the predicted construction prices. Furthermore, based on N predicted construction prices, the estimated target construction price corresponding to the target construction mode is determined, which improves the accuracy of the estimated target construction price. In summary, the embodiments of this application can improve the accuracy of construction cost estimation.

[0163] In some embodiments, the device further includes: The second acquisition module is used to acquire the target type corresponding to the first engineering quantity feature parameter; The third determination module is used to determine the target estimation rule library from multiple preset estimation rule libraries based on the target type. The preset estimation rule library includes the preset average ratio between various engineering quantities, the preset average unit price of various core costs, and the preset average proportion of core costs to the total cost.

[0164] In some embodiments, the device further includes: The third acquisition module is used to acquire historical engineering quantity characteristic parameters; historical engineering quantity characteristic parameters are historical parameters used to characterize the engineering quantity and cost characteristics under different construction modes; The clustering module is used to cluster historical engineering quantity characteristic parameters using a clustering algorithm to obtain at least two preset types; The generation module is used to generate a preset estimation rule library for each preset type and historical engineering quantity characteristic parameter, so as to obtain multiple preset estimation rule libraries.

[0165] In some implementations, the first determining module 220 is specifically used for: Based on the target estimation rule base, complete the first engineering quantity characteristic parameters to obtain the second engineering quantity characteristic parameters; Based on the second engineering quantity characteristic parameters and the target estimation rule library, determine the core costs corresponding to the target construction mode; All core costs are used to construct the first feature vector.

[0166] In some embodiments, the device further includes: The preprocessing module is used to preprocess the first engineering quantity feature to obtain the preprocessed first engineering quantity feature parameters; the preprocessing includes at least one of the following: numerical type conversion processing, invalid data removal processing, missing data filling processing, and standardization processing; The first determining module 220 is specifically used for: Based on the preprocessed first engineering quantity characteristic parameters and their corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined.

[0167] In some implementations, the second determining module 240 is specifically used for: The average of the N predicted construction prices is used as the target estimated construction price.

[0168] In some implementations, the second determining module 240 is specifically used for: The ratio method was used to verify the results based on the average of N construction price predictions. When the verification results are used to indicate that the verification has passed, the average value is determined as the target construction price estimate.

[0169] In some embodiments, the device further includes: The display module is used to visually present the estimated construction cost of the target project.

[0170] The construction cost estimation device provided in this application embodiment can be used to execute the construction cost estimation method, that is, it has the beneficial effects and implementation methods of the construction cost estimation method provided in this application embodiment. For details, please refer to the specific description of the construction cost estimation method in the above embodiment, which will not be repeated here.

[0171] Figure 13This is a block diagram of an electronic device provided in an embodiment of this application.

[0172] Reference Figure 13 This application provides an electronic device, which includes: at least one processor 701; at least one memory 702; and one or more I / O interfaces 703 connected between the processor 701 and the memory 702; wherein the memory 702 stores one or more computer programs that can be executed by at least one processor 701, and the one or more computer programs are executed by at least one processor 701 to enable at least one processor 701 to perform the above-described construction cost estimation method.

[0173] This application also provides a computer-readable storage medium storing a computer program thereon, wherein the computer program, when executed by a processor / processor core, implements the above-described construction cost estimation method. The computer-readable storage medium may be volatile or non-volatile.

[0174] This application also provides a computer program product, including computer-readable code, or a non-volatile computer-readable storage medium carrying computer-readable code. When the computer-readable code is run in the processor of an electronic device, the processor in the electronic device executes the above-described construction cost estimation method.

[0175] Those skilled in the art will understand that all or some of the steps, systems, and apparatuses disclosed above, and their functional modules / units, can be implemented as software, firmware, hardware, or suitable combinations thereof. In hardware implementations, the division between functional modules / units mentioned above does not necessarily correspond to the division of physical components; for example, a physical component may have multiple functions, or a function or step may be performed collaboratively by several physical components. Some or all physical components may be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit (ASIC). Such software can be distributed on a computer-readable storage medium, which may include computer storage media (or non-transitory media) and communication media (or transient media).

[0176] As is known to those skilled in the art, the term computer storage medium includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information, such as computer-readable program instructions, data structures, program modules, or other data. Computer storage media includes, but is not limited to, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), static random access memory (SRAM), flash memory or other memory technologies, portable compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, it is known to those skilled in the art that communication media typically contain computer-readable program instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0177] The computer-readable program instructions described herein can be downloaded from computer-readable storage media to various computing / processing devices, or downloaded via a network, such as the Internet, local area network, wide area network, and / or wireless network, to an external computer or external storage device. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and / or edge servers. A network adapter card or network interface in each computing / processing device receives the computer-readable program instructions from the network and forwards them to the computer-readable storage media in the respective computing / processing device.

[0178] The computer program instructions used to perform the operations of this application may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-dependent instructions, microcode, firmware instructions, status setting data, or source code or object code written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Smalltalk, C++, etc., and conventional procedural programming languages ​​such as the "C" language or similar programming languages. The computer-readable program instructions may be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or may be connected to an external computer (e.g., via the Internet using an Internet service provider). In some embodiments, electronic circuits, such as programmable logic circuits, field-programmable gate arrays (FPGAs), or programmable logic arrays (PLAs), are personalized by utilizing the status information of the computer-readable program instructions. These electronic circuits can execute the computer-readable program instructions to implement various aspects of this application.

[0179] The computer program product described herein can be implemented specifically through hardware, software, or a combination thereof. In one alternative embodiment, the computer program product is specifically embodied in a computer storage medium; in another alternative embodiment, the computer program product is specifically embodied in a software product, such as a software development kit (SDK), etc.

[0180] Various aspects of this application are described herein with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer-readable program instructions.

[0181] These computer-readable program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing apparatus to produce a machine such that, when executed by the processor of the computer or other programmable data processing apparatus, they create means for implementing the functions / actions specified in one or more blocks of the flowchart and / or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium that causes a computer, programmable data processing apparatus, and / or other device to operate in a particular manner; thus, the computer-readable medium storing the instructions comprises an article of manufacture that includes instructions for implementing aspects of the functions / actions specified in one or more blocks of the flowchart and / or block diagram.

[0182] Computer-readable program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, thereby causing the instructions executed on the computer, other programmable data processing apparatus, or other device to perform the functions / actions specified in one or more boxes of a flowchart and / or block diagram.

[0183] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of an instruction containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0184] Example embodiments have been disclosed herein, and while specific terminology has been used, it is for general illustrative purposes only and should not be construed as limiting. In some instances, it will be apparent to those skilled in the art that features, characteristics, and / or elements described in conjunction with particular embodiments may be used alone, or in combination with features, characteristics, and / or elements described in conjunction with other embodiments, unless otherwise expressly indicated. Therefore, those skilled in the art will understand that various changes in form and detail may be made without departing from the scope of this application as set forth by the appended claims.

Claims

1. A construction cost estimation method, characterized in that, include: Obtain the target construction mode and its corresponding first engineering quantity characteristic parameter; the target construction mode includes one of the following: steel cage construction mode, steel lining construction mode, and custom construction mode; the first engineering quantity characteristic parameter is a parameter used to characterize the engineering quantity and cost composition characteristics of the target construction mode; Based on the first engineering quantity characteristic parameters and its corresponding target estimation rule base, a first feature vector corresponding to the target construction mode is determined; the first feature vector is related to each core cost corresponding to the target construction mode; the target estimation rule base includes the target average ratio between each engineering quantity, the target average unit price of each core cost, and the target average proportion of the core cost to the total cost under the target construction mode. The first feature vector is processed using a target ensemble model to obtain N predicted construction prices corresponding to the target construction modes; the target ensemble model includes N terms from the target ridge regression model, target decision tree model, target random forest model, target support vector regression model, and target neural network; N is a positive integer greater than or equal to 2; Based on the N predicted construction prices, determine the estimated target construction price corresponding to the target construction mode.

2. The method according to claim 1, characterized in that, Before determining the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and its corresponding target estimation rule base, the method further includes: Obtain the target type corresponding to the first engineering quantity feature parameter; Based on the target type, a target estimation rule library is determined from multiple preset estimation rule libraries; the preset estimation rule library includes preset average ratios between various quantities of work, preset average unit prices of various core costs, and preset average proportions of core costs to total cost.

3. The method according to claim 2, characterized in that, Before obtaining the target type corresponding to the first engineering quantity feature parameter, the method further includes: Obtain historical engineering quantity characteristic parameters; these historical engineering quantity characteristic parameters are used to characterize the engineering quantity and cost characteristics under different construction modes. Clustering algorithms are used to group the historical engineering quantity characteristic parameters to obtain at least two preset types; A preset estimation rule library is generated for each of the preset types and the historical engineering quantity characteristic parameters to obtain multiple preset estimation rule libraries.

4. The method according to claim 1, characterized in that, The step of determining the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and its corresponding target estimation rule base includes: Based on the target estimation rule base, the first engineering quantity feature parameters are completed to obtain the second engineering quantity feature parameters; Based on the second engineering quantity characteristic parameters and the target estimation rule base, determine the core costs corresponding to the target construction mode; All of the aforementioned core costs are constructed into a first feature vector.

5. The method according to claim 1, characterized in that, Before determining the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and its corresponding target estimation rule base, the method further includes: The first quantity feature is preprocessed to obtain the preprocessed first quantity feature parameters; the preprocessing includes at least one of the following: numerical type conversion processing, invalid data removal processing, missing data filling processing, and standardization processing. The step of determining the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and its corresponding target estimation rule base includes: Based on the preprocessed first engineering quantity characteristic parameters and their corresponding target estimation rule base, the first feature vector corresponding to the target construction mode is determined.

6. The method according to claim 1, characterized in that, The step of determining the target construction price estimate corresponding to the target construction mode based on N predicted construction price values ​​includes: The average of the N predicted construction prices is determined as the target estimated construction price.

7. The method according to claim 1, characterized in that, The step of determining the target construction price estimate corresponding to the target construction mode based on N predicted construction price values ​​includes: Based on the average of the N predicted construction prices, a ratio method was used for verification to obtain the verification results. If the verification results indicate that the verification has passed, the average value is determined as the target construction price estimate.

8. The method according to claim 1, characterized in that, After determining the target construction price estimate corresponding to the target construction mode based on the N predicted construction price values, the method further includes: The estimated construction cost of the target project is visualized.

9. A construction cost estimation device, characterized in that, include: The first acquisition module is used to acquire the target construction mode and its corresponding first engineering quantity characteristic parameters; the target construction mode includes one of the following: steel cage construction mode, steel lining construction mode, and custom construction mode; the first engineering quantity characteristic parameters are parameters used to characterize the engineering quantity and cost composition characteristics of the target construction mode. The first determining module is used to determine the first feature vector corresponding to the target construction mode based on the first engineering quantity feature parameters and its corresponding target estimation rule library; the first feature vector is related to the core costs corresponding to the target construction mode; the target estimation rule library includes the target average ratio between the engineering quantities, the target average unit price of the core costs, and the target average proportion of the core costs to the total cost under the target construction mode. The first processing module is used to process the first feature vector using a target ensemble model to obtain N predicted construction prices corresponding to the target construction modes; the target ensemble model includes N terms from a target ridge regression model, a target decision tree model, a target random forest model, a target support vector regression model, and a target neural network; N is a positive integer greater than or equal to 2; The second determining module is used to determine the target construction price estimate corresponding to the target construction mode based on the N predicted construction price values.

10. An electronic device, characterized in that, include: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores one or more computer programs that can be executed by the at least one processor, the one or more computer programs being executed by the at least one processor to enable the at least one processor to perform the construction cost estimation method as described in any one of claims 1-8.

Images