Project cost AI measuring and calculating method and device based on big data and computer equipment

Abstract

The invention relates to an engineering cost AI measuring and calculating method and device based on big data and computer equipment, and is applied to measuring and calculating the cost of various engineering construction projects, and the method comprises the steps: carrying out the data preprocessing of the pre-collected initial data of the engineering construction projects, and obtaining the preprocessing data; inputting the preprocessed data into a pre-trained project cost model to obtain a measured project cost; the project cost model is a support vector regression model of a preset type, and the precise measurement and calculation of the cost of various engineering construction projects are effectively realized.

Description

Technical Field

[0001] The present invention relates to the field of communication technology, and in particular to a method, device and computer equipment for AI calculation of engineering cost based on big data. Background Art

[0002] With the rapid development of the global economy and the acceleration of urbanization, project cost management plays a vital role in all kinds of engineering construction. In recent years, my country has also been continuously promoting the reform of project cost management, gradually changing from the traditional "unified quantity and price" model to the "separation of quantity and price" model. This change aims to improve the efficiency and transparency of project cost management through a more flexible and scientific pricing method. However, despite certain progress, there are still many shortcomings in my country's project cost management. Especially in the early decision-making investment control management, due to the lack of scientific and systematic evaluation methods and sufficient data support, the accuracy and effectiveness of investment decisions are seriously affected. In addition, the lack of accumulation and analysis capabilities of project cost data has also restricted the further development of my country's project cost management in the direction of scientific, information and digitalization.

[0003] In the research on engineering cost estimation methods, especially for some engineering construction projects with insufficient data accumulation, such as school projects, hospital projects, gymnasium projects and other sub-field projects, they have their own unique complexity and diversity. On the one hand, the construction volume of such projects is relatively small, and data collection is more difficult, which belongs to the category of small sample data; on the other hand, the construction content of such projects covers multiple single buildings such as teaching buildings, dormitory buildings, gymnasiums, libraries, inpatient buildings, outpatient buildings, as well as supporting facilities such as roads and greening. The construction period is long, many professions are involved, and the cost estimation is difficult. Therefore, traditional engineering cost estimation methods, such as linear regression models, are difficult to accurately reflect the actual situation of such engineering costs due to their weak fitting ability for nonlinear relationships. Although fuzzy mathematics can better handle fuzzy information, its membership determination is highly subjective, which can easily lead to deviations in the estimation results. Although high-precision prediction models such as case reasoning and artificial neural networks have high prediction accuracy, they require a large number of data samples to support them. This undoubtedly increases the difficulty and cost of estimation for some of the above-mentioned sub-field projects where data collection is difficult.

[0004] To sum up, in view of the difficulties and shortcomings in cost estimation of the above-mentioned types of engineering construction projects, it is particularly important to develop a new AI estimation method for engineering cost based on big data. Summary of the invention

[0005] The main purpose of the present invention is to provide an AI calculation method, device and computer equipment for engineering cost based on big data, so as to solve the problem that it is difficult to accurately calculate the cost of various engineering construction projects in the prior art.

[0006] To achieve the above-mentioned purpose, the first aspect of the present invention provides an AI calculation method for engineering cost based on big data, which is applied to the cost calculation of various engineering construction projects. The method comprises:

[0007] Preprocessing the pre-collected initial data of the engineering construction project to obtain preprocessed data;

[0008] The preprocessed data is input into a pre-trained project cost model to obtain an estimated project cost; the project cost model is a support vector regression model of a preset type.

[0009] Furthermore, the step of inputting the preprocessed data into a pre-trained engineering cost model to obtain an engineering cost estimation step includes:

[0010] The preprocessed data includes multiple groups of sub-preprocessed data, wherein one group of the sub-preprocessed data corresponds to a sub-project included in the engineering construction project; each group of the sub-preprocessed data is input into the pre-trained engineering cost model to obtain the estimated cost of the sub-project corresponding to each group of the sub-preprocessed data;

[0011] The estimated costs of each of the sub-projects are summarized to obtain the estimated cost of the project.

[0012] Furthermore, the step of preprocessing the pre-collected initial data of the engineering construction project to obtain preprocessed data includes:

[0013] Acquire initial data of the engineering construction project to be measured, wherein the initial data includes data of multiple preset types;

[0014] Inputting the initial data into a pre-trained neural network model, extracting project features of the engineering construction project through the coding layer of the neural network model, classifying the project features using a classifier, and determining project nature features;

[0015] Generate a sub-item sequence according to the property characteristics of the item through the sub-item generation layer of the neural network model, wherein the sub-item sequence includes a plurality of sub-items;

[0016] Generate corresponding sub-preprocessing data for each generated sub-item through the data decoding layer of the neural network model;

[0017] Determining whether each sub-item in the sub-item sequence meets a predetermined recursive stop condition;

[0018] If not, recursively generate the sub-items in the sub-item sequence that have not reached the predetermined recursive stop condition through the sub-item generation layer and the data decoding layer until the predetermined recursive stop condition is reached, thereby obtaining a multi-layer sub-item architecture and sub-preprocessed data corresponding to each sub-item;

[0019] The multi-layer sub-project architecture and the collection of sub-preprocessed data of all sub-projects are used as the preprocessed data.

[0020] Furthermore, the step of preprocessing the pre-collected initial data of the engineering construction project to obtain preprocessed data includes:

[0021] Acquire initial data of the engineering construction project to be measured, wherein the initial data includes data of multiple preset types;

[0022] The initial data is normalized to obtain the preprocessed data; wherein the normalization step includes: using the mapminmax function to perform linear transformation on each item of the initial data, and mapping each item of the initial data to a preset interval.

[0023] Furthermore, before the step of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data, the step includes:

[0024] Extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from historical data sets as training sample data;

[0025] Performing data normalization on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set;

[0026] Select radial basis function as kernel function, set penalty coefficient C and kernel function parameter g, and construct the initial support vector regression model;

[0027] The training set is input into the initial support vector regression model for model training until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model.

[0028] Furthermore, the step of inputting the preprocessed data into a pre-trained engineering cost model to obtain a cost estimation step includes:

[0029] Inputting the preprocessed data into a support vector regression model optimized by a pre-trained particle swarm algorithm to obtain a first estimated cost;

[0030] Inputting the preprocessed data into a support vector regression model optimized by a pre-trained sparrow search algorithm to obtain a second estimated cost;

[0031] The first estimated cost and the second estimated cost are integrated according to a preset algorithm to obtain a final estimated cost.

[0032] Furthermore, the preset type of support vector regression model includes a support vector regression model optimized by a particle swarm algorithm, and the step of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data includes:

[0033] Extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from historical data sets as training sample data;

[0034] Performing data normalization on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set;

[0035] Select radial basis function as kernel function, set first related parameters, including individual learning factor c1, group learning factor c2, maximum number of iterations, population size, inertia weight, and optimization range of penalty coefficient C and kernel function parameter g, and construct an initial particle swarm algorithm optimized support vector regression model;

[0036] Initialize the particle swarm, randomly generate a group of particles, where the speed and position of each particle represent a set of penalty coefficients C and kernel function parameters g, and randomly set the initial speed and position of each particle as the starting point for optimizing the penalty coefficients C and kernel function parameters g;

[0037] Inputting the training set into the initial support vector regression model optimized by the particle swarm algorithm to calculate the fitness value of each particle;

[0038] For each particle, the current fitness value is compared with the historical best fitness value to obtain the individual optimal value of each particle;

[0039] According to the individual optimal values ​​of all particles, the group optimal value of the entire particle swarm is determined;

[0040] According to the individual optimal value of each particle and the optimal value of the group, the speed update formula of the particle swarm algorithm is adopted to update the speed and position of each particle;

[0041] Determine whether the preset termination condition is met;

[0042] If yes, then output the optimal penalty coefficient C and the optimal kernel function parameter g corresponding to the optimal value of the group;

[0043] The optimal penalty coefficient C and the optimal kernel function parameter g are assigned to the initial support vector regression model optimized by the particle swarm algorithm for model training until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model optimized by the particle swarm algorithm.

[0044] Furthermore, the preset type of support vector regression model includes a support vector regression model optimized by a sparrow search algorithm, and the step of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data includes:

[0045] Extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from historical data sets as training sample data;

[0046] Performing data normalization on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set;

[0047] Select the radial basis function as the kernel function and set the second related parameters, which include the number of sparrow populations n, the maximum number of iterations iter max , the proportion of discoverers PD, the proportion of followers SD, the safety value ST, the warning value R2 of random number generation, and the optimization range of the penalty coefficient C and the kernel function parameter g, to construct the initial support vector regression model optimized by the sparrow search algorithm;

[0048] Randomly generate the initial position of each sparrow in the sparrow population, where the position of each sparrow represents a set of penalty coefficients C and kernel function parameters g;

[0049] The sparrow search algorithm is used to calculate the fitness value of each sparrow, and the fitness values ​​are sorted according to size to obtain the current optimal individual position of the sparrow and the best fitness value;

[0050] The discoverer position is updated according to the discoverer position update formula, the follower position is updated according to the follower position update formula, and the sentinel position is updated according to the sentinel position update formula;

[0051] Update the fitness value of each sparrow, and update the optimal fitness value according to the fitness values ​​of all sparrows;

[0052] Determine whether the preset termination condition is met;

[0053] If yes, then the sparrow position corresponding to the latest optimal fitness value is output as the optimal penalty coefficient C and the optimal kernel function parameter g;

[0054] The optimal penalty coefficient C and the optimal kernel function parameter g are assigned to the initial support vector regression model optimized by the sparrow search algorithm for model training until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model optimized by the sparrow search algorithm.

[0055] Furthermore, the step of integrating the first estimated cost and the second estimated cost according to a preset algorithm to obtain a final estimated cost includes:

[0056] A weighted average method is used as a preset algorithm, wherein the weights corresponding to the first estimated cost and the second estimated cost are allocated according to the mean square error or mean absolute error of the test set in the support vector regression model optimized by the particle swarm algorithm and the support vector regression model optimized by the sparrow search algorithm;

[0057] A weighted average of the first estimated cost and the second estimated cost is calculated to obtain a final estimated cost.

[0058] The second aspect of the present invention provides an AI calculation device for engineering cost based on big data, which is applied to the cost calculation of various engineering construction projects. The device comprises:

[0059] The data processing module is used to pre-process the pre-collected initial data of the engineering construction project to obtain pre-processed data;

[0060] The cost estimation module is used to input the pre-processed data into a pre-trained engineering cost model to obtain the estimated engineering cost; the engineering cost model is a preset type of support vector regression model.

[0061] A third aspect of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above-mentioned AI calculation method for engineering cost based on big data are implemented.

[0062] The AI ​​calculation method for engineering cost based on big data provided by the present invention utilizes big data to widely collect and integrate various types of data related to various engineering construction projects, and ensures the quality and consistency of the input data through data preprocessing steps; then the preprocessed data is input into the pre-trained engineering cost model, realizing accurate engineering cost calculation based on big data drive. The engineering cost model adopted by the present invention is a preset type of support vector regression (SVR) model, which has good nonlinear fitting ability and good generalization performance, and the SVR model has good adaptability to small sample data, effectively realizing accurate calculation of the cost of various engineering construction projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 It is a flow chart of an AI calculation method for engineering cost based on big data in one embodiment of the present invention;

[0064] Figure 2 It is a schematic block diagram of the structure of an AI calculation device for engineering cost based on big data in one embodiment of the present invention;

[0065] Figure 3 It is a schematic block diagram of the structure of a computer device according to an embodiment of the present invention.

[0066] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0067] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0068] Reference Figure 1 The embodiment of the present invention provides an AI calculation method for engineering cost based on big data, which is applied to the cost calculation of various engineering construction projects. The method includes:

[0069] S1. Preprocessing the pre-collected initial data of the engineering construction project to obtain preprocessed data;

[0070] The method of the present application can be generally applied to cost estimation of various types of engineering construction projects, and in particular, is particularly suitable for engineering cost estimation in the construction field with a small amount of initial data samples.

[0071] In a specific embodiment, the step S1 of preprocessing the pre-collected initial data of the engineering construction project to obtain preprocessed data includes:

[0072] S101, obtaining initial data of the engineering construction project to be measured, wherein the initial data includes data of multiple preset types;

[0073] In this step S101, in a specific embodiment, the initial data includes basic layer data, optimization layer data, and capability layer data; wherein the basic layer data is the basic data item that has a greater impact on the project cost and directly determines the basic cost framework of the project. The optimization layer data mainly reflects the complexity of the building's exterior and interior decoration. The capability layer data reflects the indirect correlation between the level of equipment installation in the building and the project cost.

[0074] Taking the cost estimation of a school project as an example to illustrate the types of initial data, the basic layer data includes the scale of school operation (such as the number of students, the number of faculty and staff), the proportion of special rooms (such as the proportion of rooms with special functions such as laboratories, libraries, gymnasiums, etc.), geological conditions (such as soil type, groundwater level, etc.), total building area, building height, number of underground floors, underground building area, number of above-ground floors, above-ground building area, building height, structure type (such as reinforced concrete structure, steel structure, etc.), foundation type (such as pile foundation, raft foundation, etc.) and seismic resistance level; the optimization layer data includes exterior wall decoration (such as stone, ceramic tile, paint, etc.) and interior wall decoration (such as wallpaper, latex paint, etc.); the capability layer data includes the installation level of high-voltage equipment (such as power distribution system, lighting system, etc.), the installation level of low-voltage intelligent equipment (such as security system, network system, etc.), the installation level of ventilation and air-conditioning equipment, the project cost index (related to market material price fluctuations) and the project management level (affecting construction efficiency and cost control).

[0075] The determination of the type of the above initial data not only directly affects the complexity and efficiency of subsequent data preprocessing, but also determines the accuracy and reliability of the prediction results of the engineering cost model. It is necessary to screen out the above data items that have a significant impact on the engineering cost from the massive information. In particular, by introducing the capacity layer data, the non-traditional engineering cost influencing factors are taken into consideration, thereby improving the comprehensiveness and accuracy of the engineering cost prediction. This innovative attempt improves the accuracy of the entire measurement method.

[0076] S102, normalizing the initial data to obtain the preprocessed data; wherein the normalization step includes: using the mapminmax function to perform linear transformation on each item of the initial data, and mapping each item of the initial data to a preset interval.

[0077] In this step S102, in order to eliminate the influence of different dimensions and magnitudes on the model prediction results, the collected data needs to be normalized. This embodiment uses the mapminmax function to perform linear transformation, and maps all data items of the base layer, optimization layer and capability layer to a preset interval (e.g., [0, 1] or [-1, 1]) to improve the stability and prediction accuracy of the model.

[0078] The following is an example of the data normalization process in a project, taking the "total building area" (base layer data) and "power equipment installation level" (capacity layer data) as examples to illustrate the normalization process.

[0079] The total construction area (unit: square meters) is a continuous numerical data, which represents the total area of ​​all buildings in the project. In actual projects, the total construction area may range from thousands to tens of thousands or even hundreds of thousands of square meters, with a large magnitude difference.

[0080] For total building area (continuous numeric data):

[0081] Assume that the minimum value of the total building area in the original data set is min_area, and the maximum value is max_area. For any total building area value x, its normalized value x_norm is calculated as follows:

[0082] x_norm=(x-min_area) / (max_area-min_area)

[0083] Thus, the total building area value will be mapped to the [0,1] interval.

[0084] The installation level of power equipment (rating system) is a relatively subjective evaluation indicator used to measure the installation quality and level of power equipment in the project. Assume that a five-level rating system (1-5 points) is adopted, where 1 point represents the lowest level and 5 points represents the highest level.

[0085] It can be assumed that the lowest score (1 point) corresponds to a normalized value of 0, and the highest score (5 points) corresponds to a normalized value of 1. Then, for any score value y (1≤y≤5), its normalized value y_norm can be simplified to: y_norm=(y-1) / (5-1)=(y-1) / 4.

[0086] S2. Input the preprocessed data into a pre-trained project cost model to obtain an estimated project cost; the project cost model is a support vector regression model of a preset type.

[0087] The above-mentioned preset type of support vector regression model includes at least one of a support vector regression (SVR) model, a particle swarm algorithm optimized support vector regression (PSO-SVR) model, and a sparrow search algorithm optimized support vector regression (SSA-SVR) model, or at least two of the above models are selected from the above models for integration.

[0088] In one embodiment, the step S2 of inputting the preprocessed data into a pretrained engineering cost model to obtain the engineering cost estimation includes:

[0089] S211, grouping the preprocessed data to obtain a plurality of groups of sub-preprocessed data, wherein one group of the sub-preprocessed data corresponds to a sub-project included in the engineering construction project;

[0090] S212, inputting each group of the sub-preprocessed data into the pre-trained engineering cost model to obtain the estimated cost of the sub-project corresponding to each group of the sub-preprocessed data;

[0091] S213. Add the estimated costs of each of the sub-projects to obtain the estimated project cost.

[0092] In this embodiment, the above sub-project can be a building, an area, etc. in the entire project, and multiple sub-projects constitute a complete engineering construction project. Each group of sub-preprocessing data corresponds to each group of sub-projects. It can be understood that the data type of each group of sub-preprocessing data can be the same or different, and can be determined according to the characteristics of the specific sub-project. The specific method for determining the data type refers to the method in step S101.

[0093] This embodiment decomposes a large engineering construction project into multiple sub-projects, calculates the cost of each sub-project respectively through the above-mentioned engineering cost model, and finally sums the calculated cost of each sub-project, that is, accurately calculates the cost of the entire engineering construction project. The entire engineering construction project is grouped and the sub-preprocessing data corresponding to each sub-project is input in a refined manner, so that the engineering cost model can perform more accurate calculations based on the characteristics of each sub-project, thereby improving the accuracy of the cost calculation of the entire engineering construction project.

[0094] For a specific embodiment of cost estimation of decomposing a large engineering construction project into multiple sub-projects, the pre-processed data can be set manually for each sub-project one by one, or can be obtained in the following manner. The step S1 of pre-processing the pre-collected initial data of the engineering construction project to obtain the pre-processed data includes:

[0095] S111, obtaining initial data of the engineering construction project to be measured, wherein the initial data includes data of multiple preset types;

[0096] S112, inputting the initial data into a pre-trained neural network model, extracting project features of the engineering construction project through the coding layer of the neural network model, classifying the project features using a classifier, and determining project nature features;

[0097] S113, generating a sub-item sequence according to the property characteristics of the item through the sub-item generation layer of the neural network model, wherein the sub-item sequence includes a plurality of sub-items;

[0098] S114, generating corresponding sub-preprocessing data for each generated sub-item through the data decoding layer of the neural network model;

[0099] S115, determining whether each sub-item in the sub-item sequence reaches a predetermined recursive stop condition;

[0100] S116, if not, recursively generate the sub-items in the sub-item sequence that have not reached the predetermined recursive stop condition through the sub-item generation layer and the data decoding layer until the predetermined recursive stop condition is reached, thereby obtaining a multi-layer sub-item architecture and sub-preprocessed data corresponding to each sub-item;

[0101] S117: Use the multi-layer sub-project architecture and the collection of sub-preprocessed data of all sub-projects as the preprocessed data.

[0102] In this embodiment, in the above step S111, the initial data is the data corresponding to the entire engineering construction project, which can be collected or calculated manually, including but not limited to basic information of the engineering construction project (such as project name, location, construction unit), design parameters, and the design parameters may include, for example, basic layer data, optimization layer data, and capability layer data; among them, the basic layer data is the basic data item that has a greater impact on the project cost, and directly determines the basic cost framework of the project; the optimization layer data mainly reflects the complexity of the building's appearance and interior decoration; the capability layer data reflects the indirect relationship between the level of equipment installation in the building and the project cost.

[0103] In the above step S112, the collected initial data is input into a pre-trained neural network model. The model performs deep learning on the input data through its encoding layer to extract key features of the project, such as structural complexity, technical difficulty, and the impact of geographical location on cost. Subsequently, a classifier is used to intelligently classify these features to determine the nature of the project, such as whether it is a complex high-rise building, whether it contains special construction processes, etc. These nature features will guide the generation of subsequent sub-projects.

[0104] In the above step S113, based on the project nature characteristics, the sub-project generation layer of the neural network model will automatically decompose the large project into multiple sub-projects according to the preset rules or learned patterns. Specifically, the sub-project generation layer learns the intrinsic representation of the input data through a deep learning algorithm, such as a recurrent neural network (RNN), a long short-term memory network (LSTM), a gated recurrent unit (GRU) or a more advanced Transformer model. Further, the sub-project generation layer can also integrate a preset rule engine for applying some preset rules when decomposing sub-projects to automatically decompose large engineering construction projects into multiple sub-projects. The division of sub-projects can be based on functional areas (such as civil engineering, water and electricity, HVAC, decoration, etc.), construction stages (such as foundation engineering, main structure, decoration engineering, etc.) or building structures (such as teaching buildings, playgrounds, gymnasiums, libraries, etc.). The generated sub-project sequence forms a preliminary decomposition structure of the engineering construction project.

[0105] In the above step S114, for each generated sub-item, the data decoding layer of the neural network model further processes to generate corresponding sub-item preprocessing data. These data include the design parameters of the sub-item (such as the aforementioned basic layer data, optimization layer data, and capability layer data).

[0106] The above steps S115-S116 are used for recursive generation and hierarchical construction. For some sub-projects, which are still relatively complex and need to be further refined, the system will check whether each sub-project meets the recursive stop condition (such as the sub-project scale is less than the preset threshold, the sub-project level reaches the preset level depth, etc.). If the conditions are not met, the sub-project will be repeatedly executed with steps S113 and S114, that is, the sub-project generation layer and the data decoding layer will be recursively generated again until the preset recursive stop condition is reached (such as the sub-project scale is less than the preset threshold, the sub-project level reaches the preset level depth, etc.), so as to realize the multi-layer architecture construction of the engineering construction project and automatically generate the corresponding pre-processed data of the sub-projects in each layer of the architecture.

[0107] The above step S117 is used to generate a complete data set including a multi-layer sub-project architecture and its corresponding pre-processed data. This data set not only shows the multi-layer architecture of the engineering construction project, but also provides the pre-processed data of each sub-project, so that each sub-project can be finely calculated, thereby significantly improving the accuracy and efficiency of the estimation of large-scale engineering construction projects.

[0108] Specifically, the training process of the support vector regression (SVR) model is as follows. In a specific embodiment, before the step S1 of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data, it includes:

[0109] S301, extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from historical data sets as training sample data;

[0110] In this step S301, the historical data set is a record set of a certain type of engineering construction project in the past period of time, including various attributes of the project and key information such as the final cost. The information in the historical data set can be obtained through visits and surveys by cost consulting units, cost information related websites, and project site field visits. The initial data is a basic data set extracted from the historical data set for training the support vector regression model. In a specific embodiment, the initial data includes basic layer data, optimization layer data, and capability layer data; illustratively, for school projects, the basic layer data includes school scale, proportion of special rooms, geological conditions, total building area, building floor height, number of underground floors, underground building area, number of above-ground floors, above-ground building area, building height, structure type, foundation type, and seismic resistance level; the optimization layer data includes exterior wall decoration and interior wall decoration; the capability layer data includes the installation level of strong current equipment, the installation level of weak current intelligent equipment, the installation level of ventilation and air conditioning equipment, the engineering cost index, and the project management level.

[0111] S302, performing data normalization processing on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set;

[0112] In step S302, the normalization process is the same as the method described above, and will not be repeated here. After normalization, the data is divided according to a preset ratio (e.g., 8:2), where 80% of the sample data is used as a training set for training and building a model; and 20% of the sample data is used as a test set for testing the prediction effect of the model.

[0113] S303, selecting a radial basis function as a kernel function, setting a penalty coefficient C and a kernel function parameter g, and constructing an initial support vector regression model;

[0114] The kernel function is an important part of the SVR model, which can map data from the input space to a high-dimensional feature space, so that the original linearly inseparable data can be segmented by a hyperplane in the new space. The formula of the above radial basis function (RBF) is:

[0115] K(x i ,x j ) = exp(-γ||x i -x j || 2 )

[0116] Among them, x i and x jare two points in the input space, representing samples or observations in the data set. γ is a hyperparameter of the RBF kernel function, controlling the “width” or “smoothness” of the function.

[0117] The penalty coefficient C determines the penalty strength of the model. The larger the C, the greater the penalty effect on the model, the weaker the generalization ability of the model, and the more prone to overfitting; the smaller the C, the smaller the penalty effect on the model, the higher the error tolerance of the model, and the more prone to underfitting. In order to ensure good generalization performance of the model and avoid the risk of overfitting, the value of C must be reasonable and moderate.

[0118] The kernel function parameter g is used to control the complexity and fitting ability of the model. In the RBF kernel function, g defines the width of the function curve, which affects the degree of fit of the model to the data. g affects the scope of the kernel function. If the g parameter is too small, the scope of the kernel function will become larger, making the model simpler and prone to underfitting. On the contrary, if the g parameter is too large, the scope of the kernel function will become smaller, making the model too complex and prone to overfitting. Therefore, it is necessary to set the kernel function parameter g reasonably to ensure the prediction performance and generalization ability of the model.

[0119] RBF as a kernel function can effectively handle nonlinear relationships and is suitable for complex problems of cost estimation of various engineering construction projects affected by multiple factors. The complexity and generalization ability of the model can be controlled by adjusting the penalty coefficient C and the kernel function parameter g.

[0120] The initial support vector regression model selected in this embodiment is Epsilon-SVR, which is used to solve the regression problem and minimize the tolerance error by setting an insensitive loss function. The loss function is used to optimize the objective function of the algorithm. By minimizing the loss function, the parameters of the model can be optimized and the prediction effect of the model can be improved. Usually, the default value of the loss function p is 0.1.

[0121] S304: Input the training set into the initial support vector regression model for model training until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model.

[0122] The preprocessed and divided training set is input into the initial SVR model, and the model parameters are continuously adjusted through an iterative optimization algorithm (such as the sequential minimum optimization algorithm SMO) to minimize the training error. The training process will continue until the preset convergence condition is met (for example, the error is less than a certain threshold) or the predetermined number of iterations is reached. During the training process, the model will gradually learn the complex relationship between the input sample x and the output variable y.

[0123] Specifically, SVR transforms the input samples into a high-dimensional space through a nonlinear mapping algorithm and constructs a linear fitting regression function. The formula is as follows:

[0124] f(x)=ωφ(x)+b

[0125] Where ω is the weight vector and b is the bias constant.

[0126] For the sample (x, y), the SVR model tolerates the error ∈ between the model output f(x) and the actual output y, and constructs a "separation band" with a width of 2∈ on both sides of the linear function f(x). When the training sample falls outside the separation band, the loss is calculated. When it falls within the separation band, the model prediction is considered accurate and no loss is calculated. At this time, the SVR optimization problem can be expressed as:

[0127]

[0128] where ||ω|| 2 / 2 represents the regularization term, C is the penalty factor, l ∈ It represents the ∈-insensitive loss function, and m is the number of samples.

[0129] Introducing the slack variable ξ i and The SVR expression can be written as:

[0130]

[0131]

[0132] By introducing the Lagrangian function, we can get the dual problem of SVR:

[0133]

[0134]

[0135] where a i and is the Lagrange multiplier, K(x i , x j ) is the kernel function.

[0136] Solving the problem yields the regression prediction value:

[0137]

[0138] In a specific embodiment, the preset type of support vector regression model includes a support vector regression model optimized by a particle swarm algorithm, and the step S1 of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data includes:

[0139] S310, extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from historical data sets as training sample data;

[0140] The specific process of step S310 is the same as step S301 and will not be repeated here.

[0141] S311, performing data normalization processing on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set;

[0142] The specific process of step S311 is the same as step S302 and will not be repeated here.

[0143] S312, selecting a radial basis function as a kernel function, setting first related parameters, wherein the first related parameters include an individual learning factor, a group learning factor, a maximum number of iterations, a population size, an inertia weight, and an optimization range of a penalty coefficient C and a kernel function parameter g, and constructing an initial particle swarm algorithm optimized support vector regression model;

[0144] S313, initializing a particle swarm, randomly generating a group of particles, wherein the speed and position of each particle represent a set of penalty coefficients C and kernel function parameters g, and randomly setting the initial speed and position of each particle as the starting point for optimizing the penalty coefficients C and kernel function parameters g;

[0145] S314, inputting the training set into the initial support vector regression model optimized by the particle swarm algorithm, and calculating the fitness value of each particle;

[0146] S315, for each particle, compare the current fitness value with the historical best fitness value to obtain the individual optimal value of each particle;

[0147] S316, determining the group optimal value of the entire particle swarm according to the individual optimal values ​​of all particles;

[0148] S317, according to the individual optimal value of each particle and the optimal value of the group, the speed update formula of the particle swarm algorithm is used to update the speed and position of each particle; the speed update formula of the particle swarm algorithm is:

[0149]

[0150]

[0151] Where W is the inertia weight, which is a non-negative number that controls the development and search capabilities of the algorithm; c1 is the individual learning factor, c2 is the group learning factor, which determines the ability to follow the historical optimal solution; U1 and U2 are random numbers; pbest is the individual optimal value; gbest is the group optimal value; fit(i) is the fitness value; the population size N, Xi is the initial position of the particle, V i is the initial velocity of the particle, and t is the number of iterative optimizations.

[0152] S318, determining whether a preset termination condition is reached; the preset termination condition may be, for example, a maximum number of iterations, a fitness value reaching a preset threshold, etc.;

[0153] S319: If yes, output the optimal penalty coefficient C and the optimal kernel function parameter g corresponding to the group optimal value;

[0154] S320, assigning the optimal penalty coefficient C and the optimal kernel function parameter g to the initial support vector regression model optimized by the particle swarm algorithm for model training, until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model optimized by the particle swarm algorithm.

[0155] This embodiment automatically searches for the optimal penalty coefficient C and kernel function parameter g through the particle swarm algorithm, avoiding the tedious manual parameter adjustment process in the traditional method and improving the efficiency of model training. The optimized parameters can better balance the complexity of the model and the training error, and improve the prediction accuracy and generalization ability of the SVR model.

[0156] After the model training is completed, it also includes importing the test set, using the svmpredict function to perform simulation prediction of the PSO-SVR model, outputting the prediction results, and analyzing the results based on the evaluation indicators.

[0157] In a specific embodiment, the preset type of support vector regression model includes a support vector regression model optimized by a sparrow search algorithm, and the step of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data includes:

[0158] S330, extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from the historical data set as training sample data;

[0159] The specific process of step S330 is the same as step S301 and will not be repeated here.

[0160] S331, performing data normalization processing on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set;

[0161] The specific process of step S331 is the same as step S302 and will not be repeated here.

[0162] S332, select the radial basis function as the kernel function, set the second related parameters, the second related parameters include the number of sparrow populations n, the maximum number of iterations iter m ax, discoverer ratio PD, follower ratio SD, safety value ST, warning value R2 of random number generation, and optimization range of penalty coefficient C and kernel function parameter g, construct the initial support vector regression model optimized by sparrow search algorithm;

[0163] S333, randomly generating the initial position of each sparrow in the sparrow population, wherein the position of each sparrow represents a set of penalty coefficients C and kernel function parameters g;

[0164] S334, using a sparrow search algorithm to calculate the fitness value of each sparrow, and sorting the fitness values ​​according to size, to obtain the current optimal individual position of the sparrow and the best fitness value;

[0165] S335, updating the discoverer position according to the discoverer position update formula, updating the follower position according to the follower position update formula, and updating the sentinel position according to the sentinel position update formula;

[0166] The formula for updating the discoverer's position is:

[0167]

[0168] Among them, X i，j is the individual position of the sparrow; t is the current iteration number; iter max represents the maximum number of iterations; a is a random number, and a∈(0,1]; Q is a random number that satisfies the normal distribution; L is a d-dimensional row vector whose elements are all 1; warning value R2∈(0,1]; safety value ST∈[0.5,1].

[0169] The follower position update formula is:

[0170]

[0171] Among them, X worst represents the current worst global position; n is the population size; X p is the position of the optimal finder; A + is a d-dimensional row vector whose elements are randomly assigned to 1 or -1 and satisfy A + =A T (AA T ) -1 .

[0172] The update formula of the sentinel position is:

[0173]

[0174] Among them, X best is the current global optimal position; β is the step size control coefficient, which is a random number that obeys N(0,1); K∈[-1,1] is a random number, f i 、f g and f w They are the current fitness value, the global optimal fitness value and the global worst fitness value respectively; E is a very small constant to avoid the denominator being 0.

[0175] S336, updating the fitness value of each sparrow, and updating the optimal fitness value according to the fitness values ​​of all sparrows;

[0176] S337, determining whether a preset termination condition is reached; the preset termination condition may be, for example, a maximum number of iterations, a fitness value reaching a preset threshold, etc.;

[0177] S338, if yes, output the sparrow position corresponding to the latest optimal fitness value as the optimal penalty coefficient C and the optimal kernel function parameter g;

[0178] S339, assigning the optimal penalty coefficient C and the optimal kernel function parameter g to the initial support vector regression model optimized by the sparrow search algorithm for model training, until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model optimized by the sparrow search algorithm.

[0179] This embodiment automatically searches for the optimal penalty coefficient C and kernel function parameter g through the sparrow search algorithm, avoiding the tedious manual parameter adjustment process in the traditional method and improving the efficiency of model training. The optimized parameters can better balance the complexity of the model and the training error, and improve the prediction accuracy and generalization ability of the SVR model.

[0180] After the model training is completed, it also includes importing the test set, using the svmpredict function to perform simulation prediction of the SSA-SVR model, outputting the prediction results, and analyzing the results based on the evaluation indicators.

[0181] In a specific embodiment, the step S2 of inputting the preprocessed data into a pre-trained engineering cost model to obtain a cost estimation includes:

[0182] S201, inputting the pre-processed data into a support vector regression model optimized by a pre-trained particle swarm algorithm to obtain a first estimated cost;

[0183] The specific process of step S201 is the same as steps S310-S320 in the aforementioned embodiment, and will not be repeated herein in this embodiment.

[0184] S202, inputting the pre-processed data into a support vector regression model optimized by a pre-trained sparrow search algorithm to obtain a second estimated cost;

[0185] The specific process of step S202 is the same as steps S330-S339 in the aforementioned embodiment, and will not be repeated herein in this embodiment.

[0186] S203: Integrate the first estimated cost and the second estimated cost according to a preset algorithm to obtain a final estimated cost.

[0187] In a specific embodiment, the step S203 of integrating the first estimated cost and the second estimated cost according to a preset algorithm to obtain a final estimated cost includes:

[0188] S2031, using a weighted average method as a preset algorithm, wherein the weights corresponding to the first estimated cost and the second estimated cost are allocated according to the mean square error or mean absolute error of the test set in the support vector regression model optimized by the particle swarm algorithm and the support vector regression model optimized by the sparrow search algorithm;

[0189] Specifically, the performance of the PSO-SVR calculation model and the SSA-SVR calculation model are evaluated respectively using the test set data, which can be accomplished by calculating the mean square error (MSE) or mean absolute error (MAE) of the model on the test set. According to the calculated error value, weights are assigned to the two models. Exemplarily, the inverse of the error or a certain nonlinear transformation of the error (such as an exponential function) can be used as the basis for the weight to ensure that the model with a smaller error obtains a higher weight. The specific formula can be selected according to the specific application scenario to meet the requirement that the smaller the error, the greater the weight.

[0190] S2032. Calculate a weighted average of the first estimated cost and the second estimated cost to obtain a final estimated cost.

[0191] Specifically, the weights of the two models (respectively denoted as w and PSO and w SSA ) and their corresponding estimated costs (respectively denoted as C PSO and C SSA ).

[0192] Use the following formula to calculate the final estimated cost C final :

[0193] C final =w PSO ×C PSO +w SSA ×C SSA

[0194] Among them, w PSO +w SSA =1.

[0195] This embodiment can more reasonably utilize the measurement results of the two models, so that the model with better performance contributes more to the final measurement, thereby improving the accuracy of the measurement. Even if one of the models performs poorly in some cases, the other model can still have a positive impact on the final measurement results through its higher weight, thereby enhancing the robustness of the entire measurement system.

[0196] In a specific embodiment, the step S212 of inputting each group of the sub-preprocessed data into the pre-trained engineering cost model to obtain the estimated cost of the sub-project corresponding to each group of the sub-preprocessed data includes:

[0197] S2121, inputting the sub-preprocessed data corresponding to a sub-project into a support vector regression model optimized by a pre-trained particle swarm algorithm to obtain a first sub-estimated cost;

[0198] The specific process of this step S2121 is the same as steps S310-S320 in the aforementioned embodiment, and will not be repeated here in this embodiment.

[0199] S2122, inputting the sub-preprocessed data corresponding to the sub-project into a support vector regression model optimized by a pre-trained sparrow search algorithm to obtain a second sub-estimated cost;

[0200] The specific process of this step S2122 is the same as steps S330-S339 in the aforementioned embodiment, and will not be repeated here in this embodiment.

[0201] S2123. Integrate the first sub-estimated cost and the second sub-estimated cost according to a preset algorithm to obtain the estimated cost corresponding to the sub-project.

[0202] The specific process of step S2123 is the same as step S203 in the aforementioned embodiment, and will not be repeated here in this embodiment.

[0203] Reference Figure 2 In an embodiment of the present invention, there is also provided an engineering cost AI calculation device based on big data, which is applied to the cost calculation of various engineering construction projects. The device comprises:

[0204] The data processing module 10 is used to pre-process the pre-collected initial data of the engineering construction project to obtain pre-processed data;

[0205] The cost estimation module 20 is used to input the pre-processed data into a pre-trained engineering cost model to obtain an estimated engineering cost; the engineering cost model is a support vector regression model of a preset type.

[0206] For the specific implementation of each module in the above device embodiment, please refer to the above method embodiment, which will not be repeated here.

[0207] Reference Figure 3 In an embodiment of the present invention, a computer device is also provided. The computer device may be a server, and its internal structure may be as follows: Figure 3 As shown. The computer device includes a processor, a memory, a display screen, an input device, a network interface and a database connected via a system bus. Among them, the processor designed by the computer is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the corresponding data in this embodiment. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above-mentioned method of AI calculation of engineering cost based on big data is implemented.

[0208] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a portion of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device to which the solution of the present invention is applied.

[0209] An embodiment of the present invention further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0210] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media provided by the present invention and used in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double-speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM.

[0211] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, device, article or method including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, device, article or method. In the absence of further restrictions, an element defined by the sentence "includes a ..." does not exclude the presence of other identical elements in the process, device, article or method including the element.

[0212] The above description is only a preferred embodiment of the present invention, and does not limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for calculating engineering cost AI based on big data, characterized in that: Applied to cost estimation of various engineering construction projects, the method includes: Preprocessing the pre-collected initial data of the engineering construction project to obtain preprocessed data; The preprocessed data is input into a pre-trained project cost model to obtain an estimated project cost; the project cost model is a support vector regression model of a preset type.

2. The method for calculating construction cost based on AI of big data according to claim 1 is characterized in that: The step of inputting the pre-processed data into the pre-trained engineering cost model to obtain the step of estimating the engineering cost includes: The preprocessed data includes multiple groups of sub-preprocessed data, wherein one group of the sub-preprocessed data corresponds to a sub-project included in the engineering construction project; each group of the sub-preprocessed data is input into the pre-trained engineering cost model to obtain the estimated cost of the sub-project corresponding to each group of the sub-preprocessed data; The estimated costs of each of the sub-projects are summarized to obtain the estimated cost of the project.

3. The AI ​​calculation method for engineering cost based on big data according to claim 2 is characterized in that: The step of preprocessing the pre-collected initial data of the engineering construction project to obtain preprocessed data includes: Acquire initial data of the engineering construction project to be measured, wherein the initial data includes data of multiple preset types; Inputting the initial data into a pre-trained neural network model, extracting project features of the engineering construction project through the coding layer of the neural network model, classifying the project features using a classifier, and determining project nature features; Generate a sub-item sequence according to the property characteristics of the item through the sub-item generation layer of the neural network model, wherein the sub-item sequence includes a plurality of sub-items; Generate corresponding sub-preprocessed data for each generated sub-item through the data decoding layer of the neural network model; Determining whether each sub-item in the sub-item sequence reaches a predetermined recursive stop condition; If not, recursively generate the sub-items in the sub-item sequence that have not reached the predetermined recursive stop condition through the sub-item generation layer and the data decoding layer until the predetermined recursive stop condition is reached, thereby obtaining a multi-layer sub-item architecture and sub-preprocessed data corresponding to each sub-item; The multi-layer sub-project architecture and the collection of sub-preprocessed data of all sub-projects are used as the preprocessed data.

4. The AI ​​calculation method for engineering cost based on big data according to claim 1 is characterized in that: The step of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data includes: Extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from historical data sets as training sample data; Performing data normalization on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set; Select radial basis function as kernel function, set penalty coefficient C and kernel function parameter g, and construct the initial support vector regression model; The training set is input into the initial support vector regression model for model training until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model.

5. The AI ​​calculation method for engineering cost based on big data according to claim 1 is characterized in that: The step of inputting the pre-processed data into a pre-trained engineering cost model to obtain a cost estimate comprises: Inputting the preprocessed data into a support vector regression model optimized by a pre-trained particle swarm algorithm to obtain a first estimated cost; Inputting the preprocessed data into a support vector regression model optimized by a pre-trained sparrow search algorithm to obtain a second estimated cost; The first estimated cost and the second estimated cost are integrated according to a preset algorithm to obtain a final estimated cost.

6. The AI ​​calculation method for engineering cost based on big data according to claim 1 or 5 is characterized in that: The preset type of support vector regression model includes a support vector regression model optimized by a particle swarm algorithm. The step of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data includes: Extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from historical data sets as training sample data; Performing data normalization on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set; Select radial basis function as kernel function, set first related parameters, including individual learning factor c1, group learning factor c2, maximum number of iterations, population size, inertia weight, and optimization range of penalty coefficient C and kernel function parameter g, and construct an initial particle swarm algorithm optimized support vector regression model; Initialize the particle swarm, randomly generate a group of particles, where the speed and position of each particle represent a set of penalty coefficients C and kernel function parameters g, and randomly set the initial speed and position of each particle as the starting point for optimizing the penalty coefficients C and kernel function parameters g; Inputting the training set into the initial support vector regression model optimized by the particle swarm algorithm to calculate the fitness value of each particle; For each particle, the current fitness value is compared with the historical best fitness value to obtain the individual optimal value of each particle; According to the individual optimal values ​​of all particles, the group optimal value of the entire particle swarm is determined; According to the individual optimal value of each particle and the optimal value of the group, the speed update formula of the particle swarm algorithm is adopted to update the speed and position of each particle; Determine whether the preset termination condition is met; If yes, then output the optimal penalty coefficient C and the optimal kernel function parameter g corresponding to the optimal value of the group; The optimal penalty coefficient C and the optimal kernel function parameter g are assigned to the initial support vector regression model optimized by the particle swarm algorithm for model training until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model optimized by the particle swarm algorithm.

7. The AI ​​calculation method for engineering cost based on big data according to claim 1 or 5 is characterized in that: The preset type of support vector regression model includes a support vector regression model optimized by a sparrow search algorithm. The step of preprocessing the pre-collected initial data of the engineering construction project to obtain the preprocessed data includes: Extracting multiple groups of initial data of engineering construction projects and corresponding engineering costs from historical data sets as training sample data; Performing data normalization on the initial data, and dividing the normalized data into a training set and a test set, using the normalized initial data as input samples x in the training set, and using the project cost as output variable y in the training set; Select the radial basis function as the kernel function and set the second related parameters, which include the number of sparrow populations n, the maximum number of iterations iter max , the proportion of discoverers PD, the proportion of followers SD, the safety value ST, the warning value R2 of random number generation, and the optimization range of the penalty coefficient C and the kernel function parameter g, to construct the initial support vector regression model optimized by the sparrow search algorithm; Randomly generate the initial position of each sparrow in the sparrow population, where the position of each sparrow represents a set of penalty coefficients C and kernel function parameters g; The sparrow search algorithm is used to calculate the fitness value of each sparrow, and the fitness values ​​are sorted according to size to obtain the current optimal individual position of the sparrow and the best fitness value; The discoverer position is updated according to the discoverer position update formula, the follower position is updated according to the follower position update formula, and the sentinel position is updated according to the sentinel position update formula; Update the fitness value of each sparrow, and update the optimal fitness value according to the fitness values ​​of all sparrows; Determine whether the preset termination condition is met; If yes, then the sparrow position corresponding to the latest optimal fitness value is output as the optimal penalty coefficient C and the optimal kernel function parameter g; The optimal penalty coefficient C and the optimal kernel function parameter g are assigned to the initial support vector regression model optimized by the sparrow search algorithm for model training until a preset convergence condition or a predetermined number of iterations is met, thereby completing the training of the support vector regression model optimized by the sparrow search algorithm.

8. The AI ​​calculation method for engineering cost based on big data according to claim 5 is characterized in that: The step of integrating the first estimated cost and the second estimated cost according to a preset algorithm to obtain a final estimated cost includes: A weighted average method is used as a preset algorithm, wherein the weights corresponding to the first estimated cost and the second estimated cost are allocated according to the mean square error or mean absolute error of the test set in the support vector regression model optimized by the particle swarm algorithm and the support vector regression model optimized by the sparrow search algorithm; A weighted average of the first estimated cost and the second estimated cost is calculated to obtain a final estimated cost.

9. An AI calculation device for engineering cost based on big data, characterized in that: Applied to cost estimation of various engineering construction projects, the device comprises: The data processing module is used to pre-process the pre-collected initial data of the engineering construction project to obtain pre-processed data; The cost estimation module is used to input the pre-processed data into a pre-trained engineering cost model to obtain the estimated engineering cost; the engineering cost model is a preset type of support vector regression model.

10. A computer device comprising a memory and a processor, wherein a computer program is stored in the memory, wherein: When the processor executes the computer program, the steps of the AI ​​calculation method for engineering cost based on big data described in any one of claims 1 to 8 are implemented.

Images