A method and system for evaluating and analyzing prediction data based on construction engineering cost

Abstract

The application discloses a kind of prediction data evaluation analysis method and system based on construction engineering cost, it is related to intelligent data analysis technical field, including, the multi-source heterogeneous data of construction engineering cost is collected, and is fused by multi-head attention mechanism, generates integrated cost feature vector;Variational Bayes neural network is constructed, according to integrated cost feature vector, outputs cost prediction value and probability distribution, and calculates the confidence interval of cost prediction value;Through MR interaction, risk thermodynamic diagram, cost prediction value and the BIM model of construction engineering are combined, and cost visualization view is generated, according to actual construction condition change, by interacting with cost visualization view, real-time adjustment cost visualization view is generated, and cost prediction value is evaluated.The present application replaces the point estimation of traditional neural network by Gaussian variational distribution parameterization network weight, realizes the quantification of prediction uncertainty.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent data analysis, in particular to a prediction data evaluation and analysis method and system based on construction engineering cost. BACKGROUND

[0002] In recent years, with the continuous expansion and significant increase in complexity of construction engineering, the importance of engineering cost prediction in project management has become increasingly prominent. Traditional engineering cost prediction methods mainly rely on experience estimation, quota method or regression analysis, etc. static model, the core of which is to make linear deduction based on historical data and artificial experience.

[0003] With the improvement of informationization level of the construction industry, especially the development of Internet of Things (IoT), big data analysis and artificial intelligence technology, engineering cost prediction gradually evolves towards dynamic and intelligent direction. In order to improve the prediction accuracy and adaptability, researchers gradually introduce machine learning and deep learning technologies such as support vector machine (SVM), random forest (RF), long short-term memory network (LSTM) and convolutional neural network (CNN) to realize modeling analysis of historical cost data, material price fluctuations and construction progress and other multi-source information.

[0004] Although the above-mentioned technologies improve the accuracy and automation of engineering cost prediction to some extent, there are still many bottlenecks and challenges in actual application, for example, the prediction model only outputs a single numerical value, which cannot quantify the confidence interval and probability distribution of the prediction result, and it is difficult to meet the demand for risk assessment and uncertainty quantification in engineering practice. SUMMARY

[0005] In view of the above-mentioned existing problems, the present application is proposed.

[0006] Therefore, the present application provides a prediction data evaluation and analysis method based on construction engineering cost to solve the problem of difficult quantification of construction engineering cost prediction result uncertainty.

[0007] To solve the above technical problems, the present application provides the following technical solutions:

[0008] In a first aspect, the present application provides a prediction data evaluation and analysis method based on construction engineering cost, which comprises,

[0009] Collecting multi-source heterogeneous data of construction engineering cost, and fusing through multi-head attention mechanism to generate comprehensive cost feature vector;

[0010] Constructing a variational Bayesian neural network, outputting cost prediction value and probability distribution according to the comprehensive cost feature vector, and calculating the confidence interval of the cost prediction value;

[0011] Generate a risk heat map based on the probability distribution and confidence interval of the cost prediction value;

[0012] By MR interaction, combine the risk heat map, cost prediction value and BIM model of the construction project to generate a cost visualization view, interact with the cost visualization view according to the actual construction situation, adjust the cost visualization view in real time, and evaluate the cost prediction value.

[0013] As a preferred scheme of the prediction data evaluation and analysis method based on the construction project cost, the comprehensive cost feature vector is generated in the following specific steps,

[0014] Collect multi-source heterogeneous data through Internet of Things devices, public databases and network crawler tools;

[0015] Record the multi-source heterogeneous data in a unified timestamp format and store it as a structured table to obtain an original cost data set;

[0016] Clean and standardize the original cost data set to generate a standard cost data set;

[0017] Fuse the standardized cost data set through a multi-head attention mechanism to generate a preliminary cost feature matrix;

[0018] Reduce the dimension of the preliminary cost feature matrix through a variational autoencoder to generate a comprehensive cost feature vector.

[0019] As a preferred scheme of the prediction data evaluation and analysis method based on the construction project cost, the comprehensive cost feature vector is generated in the following specific steps,

[0020] Initialize the network structure of the variational Bayesian neural network as an input layer and a plurality of fully connected layers;

[0021] Assign a Gaussian variational distribution to the weight parameters of each fully connected layer of the variational Bayesian neural network and initialize it;

[0022] Train the variational Bayesian neural network by minimizing the loss function.

[0023] As a preferred scheme of the prediction data evaluation and analysis method based on the construction project cost, the Gaussian variational distribution of the weight parameters of each fully connected layer is assigned in the following specific steps,

[0024] The weight parameters include the weight matrix and bias vector of each fully connected layer of the variational Bayesian neural network;

[0025] Based on variational inference, Gaussian variational distributions are assigned to the weight matrix and bias terms of each fully connected layer to generate the mean weight matrix, standard deviation weight matrix, mean bias vector, and standard deviation bias vector.

[0026] Use the Xavier initialization method to initialize the mean weight matrix and standard deviation weight matrix;

[0027] The mean bias vector and standard deviation bias vector are initialized using a uniform distribution.

[0028] As a preferred embodiment of the prediction data evaluation and analysis method for construction project cost based on the present invention, the specific steps for training the variational Bayesian neural network are as follows:

[0029] Based on the lower bound of marginal likelihood, the loss function of variational Bayesian neural networks is defined.

[0030] Based on the loss function, the partial derivative of the loss function with respect to the Gaussian variational distribution of each weight parameter is obtained through backpropagation.

[0031] Update the weight parameters and their Gaussian variational distribution using the Adam optimizer.

[0032] As a preferred embodiment of the prediction data evaluation and analysis method for construction project cost based on the present invention, the specific steps for outputting the predicted cost value and probability distribution are as follows:

[0033] The comprehensive cost feature vector is input into the completed variational Bayesian neural network, and the mean and standard deviation branches of the last fully connected layer output the mean and standard deviation of the cost prediction value, respectively.

[0034] The average of the projected costs will be used as the projected cost value.

[0035] Based on the mean and standard deviation of the cost forecast, the probability distribution density function is calculated using the Gaussian distribution assumption, and a numerical representation of the probability distribution is generated using Monte Carlo sampling.

[0036] As a preferred embodiment of the prediction data evaluation and analysis method for construction project cost based on the present invention, the specific steps for generating the cost visualization view are as follows:

[0037] Constructing BIM models for building construction projects;

[0038] Based on the mapping relationship between record numbers of multi-source heterogeneous data and construction numbers in the BIM model, the estimated cost is associated with the BIM model of the building project.

[0039] extracting a spatial coordinate system of the BIM model of the construction project, calculating a bounding box of the BIM model of the construction project, and defining the center of the BIM model of the construction project as the center of the bounding box;

[0040] covering the risk heat map in a semi-transparent grid form to the horizontal plane of the BIM model of the construction project;

[0041] fixing the height as the height of the building center, and generating the cost visualization view.

[0042] In a second aspect, the present application provides a construction project cost prediction data evaluation and analysis system, comprising a feature vector module, a cost prediction module, a risk map module and an interactive evaluation module,

[0043] The feature vector module is used to collect multi-source heterogeneous data of the construction project cost, and to generate a comprehensive cost feature vector by fusing through a multi-head attention mechanism.

[0044] The cost prediction module is used to construct a variational Bayesian neural network, to output a cost prediction value and a probability distribution according to the comprehensive cost feature vector, and to calculate a confidence interval of the cost prediction value.

[0045] The risk map module is used to generate a risk heat map based on the probability distribution and the confidence interval of the cost prediction value.

[0046] The interactive evaluation module is used to combine the risk heat map, the cost prediction value and the BIM model of the construction project through MR interaction, to generate a cost visualization view, to adjust the cost visualization view in real time and to evaluate the cost prediction value according to the actual construction situation.

[0047] In a third aspect, the present application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the computer program is executed by the processor to implement any step of the construction project cost prediction data evaluation and analysis method according to the first aspect of the present application.

[0048] In a fourth aspect, the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement any step of the construction project cost prediction data evaluation and analysis method according to the first aspect of the present application.

[0049] The present application has the beneficial effects that: the present application replaces the point estimation of the traditional neural network by parameterizing the network weight of the Gaussian variational distribution, realizes the quantification of the prediction uncertainty, calculates the risk score by the weighted ratio of the standard deviation and the confidence interval width, realizes the multi-dimensional risk aggregation, the intuitive display of the heat map can be used for resource priority allocation, the feature vector is adjusted by gesture interaction of the HoloLens 2, dynamic scene deduction is realized, the user can observe and adjust the effect in real time, and rapid decision-making is assisted. BRIEF DESCRIPTION OF DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0051] Fig. 1 The flowchart of the prediction data evaluation and analysis method based on the construction engineering cost.

[0052] Fig. 2 The module diagram of the prediction data evaluation and analysis system based on the construction engineering cost.

[0053] Fig. 3 The schematic diagram of the variational Bayesian neural network structure.

[0054] Fig. 4 The flowchart of the MR interactive visualization generation. DETAILED DESCRIPTION

[0055] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the drawings of the specification.

[0056] In the following description, many specific details are set forth in order to provide a thorough understanding of the present application, but the present application can also be implemented in other ways different from those described herein, and those skilled in the art can make similar generalizations without departing from the connotation of the present application, therefore the present application is not limited to the specific embodiments disclosed below.

[0057] Secondly, the "one embodiment" or "embodiment" referred to herein means that the specific features, structures or characteristics can be included in at least one implementation of the present application. "In one embodiment" appearing in different places in the specification does not mean the same embodiment, nor is it an independent or alternative embodiment.

[0058] REFERENCE Figs. 1-4For an embodiment of the present application, the embodiment provides a construction project cost prediction data evaluation analysis method, comprising the following steps:

[0059] S1, collect multi-source heterogeneous data of construction project cost, and fuse through multi-head attention mechanism to generate comprehensive cost feature vector.

[0060] Among them, the multi-source heterogeneous data includes real-time construction data, historical cost data and external market data;

[0061] Collect real-time construction data through Internet of Things devices.

[0062] Specifically, track the daily consumption of various materials, such as daily cement consumption, through radio frequency identification tags; record worker attendance data, such as the number of workers of various types and the time length, through infrared sensors; monitor the running state of construction equipment, such as the daily running time of excavators, through embedded sensors;

[0063] Collect historical cost data from public databases.

[0064] Specifically, access industry sharing platforms to obtain cost data of the same type of projects across developers, such as material unit prices of various projects and labor costs of various types of workers;

[0065] Use web crawler tools to collect external market data.

[0066] Specifically, the real-time price of building materials, such as the daily price of cement, can be extracted from building material market platforms through the Scrapy framework (a kind of web crawler tool), and macroeconomic indicators, such as annual inflation rate, can be obtained through public API interfaces;

[0067] Record the multi-source heterogeneous data in a unified timestamp format and store it as a structured table to obtain an original cost data set;

[0068] Clean the original cost data set, that is, remove outliers and standardize to generate a standard cost data set;

[0069] Specifically, calculate the anomaly score of each original cost data according to the anomaly score formula, set the anomaly score threshold according to the anomaly score distribution of all original cost data, and determine the outliers when the anomaly score of the original cost data is less than the anomaly score threshold, and remove them. Interpolate the original cost data after removing outliers, apply Z-score standardization method, and standardize the original cost data set after removing outliers and interpolation to obtain the standard cost data set;

[0070] Fuse the standardized cost data set through the multi-head attention mechanism to generate a preliminary cost feature matrix;

[0071] Specifically, the features (standardized, such as cement consumption, etc.) of each record are extracted from the standardized cost data set, all features are spliced into a matrix form in the order of record number, each row represents a feature vector of a record, and each column represents a feature, the extracted features are filled into the matrix in a fixed order using numerical mapping, a feature matrix is obtained, the feature matrix is converted into a tensor format through matrix operation, a multi-head attention mechanism is initialized, including setting the number of attention heads, the dimension of a single head, and the total dimension, for example, setting the number of attention heads to 8, the dimension of a single head to 64, and the total dimension to 512, calculating the query matrix, key matrix and value matrix through linear transformation in the input layer according to the feature matrix, and calculating the weight matrix of each attention head through scaled dot-product attention formula and scaled dot-product and softmax normalization, adding residual connection, splicing the weight matrices of all attention heads to obtain an attention weighted matrix, performing linear transformation on the attention weighted matrix based on the full connection layer weight matrix and bias vector in the full connection layer to obtain an intermediate feature matrix, using the ReLU activation function to judge each element of the intermediate feature matrix, if the element value is greater than 0, the original value is retained; if it is less than or equal to 0, it is set to 0, the feature expression ability is enhanced, a nonlinear feature matrix is generated, dropout regularization is applied to the nonlinear feature matrix to prevent overfitting, and a regularized feature matrix is generated, and a preliminary cost feature matrix is generated by applying layer normalization to the regularized feature matrix in the application layer;

[0072] The preliminary cost feature matrix is reduced in dimension by the variational autoencoder to generate a comprehensive cost feature vector;

[0073] The variational autoencoder includes an encoder and a decoder, the encoder is used to map the preliminary cost feature matrix to a latent space, and outputs a mean vector and a standard deviation vector, the latent vector is generated by reparameterizing the mean vector and the standard deviation vector, and the decoder reconstructs the latent vector by minimizing the reconstruction loss and the KL divergence to generate the comprehensive cost feature vector.

[0074] S2, constructing a variational Bayesian neural network, outputting a cost prediction value and a probability distribution according to the comprehensive cost feature vector, and calculating a confidence interval of the cost prediction value.

[0075] The network structure of the variational Bayesian neural network is initialized as three full connection layers, and a Gaussian variational distribution is assigned to the weight parameters of each full connection layer, wherein the weight parameters include a weight matrix and a bias vector;

[0076] Specifically, the variational Bayesian neural network comprises three full connection layers, the first and second full connection layers adopt ReLU activation functions, the third layer comprises two branches, including a mean branch and a standard deviation branch, a Gaussian variational distribution is assigned to a weight matrix of each full connection layer of the variational Bayesian neural network based on variational inference, a corresponding mean weight matrix and a standard deviation weight matrix are generated, the mean matrix is initialized by using a Xavier initialization method, and the standard deviation matrix is initialized as 0.1, a Gaussian variational distribution is assigned to a bias vector of each full connection layer of the variational Bayesian neural network, a corresponding mean bias vector and a standard deviation bias vector are generated, the mean bias vector is initialized by using a uniform distribution, and the standard deviation bias vector is initialized as a constant 0.1;

[0077] The comprehensive cost feature vector is input into the first full connection layer of the variational Bayesian neural network, the weight matrix and the bias vector of the first full connection layer are sampled from the Gaussian variational distribution by using a reparameterization method, a first layer output is obtained, the weight matrix and the bias vector of the second full connection layer are sampled in the same way as the first full connection layer, a second layer output is obtained, and the mean of the cost prediction value and the standard deviation of the cost prediction value are output in the mean branch and the standard deviation branch of the third full connection layer respectively, and the calculation formula is as follows:

[0078] ;

[0079] ;

[0080] wherein, is the mean of the cost prediction value, is the weight matrix of the mean branch of the cost prediction value in the third layer, is the second layer output, is the bias vector of the mean branch of the cost prediction value in the third layer, is the standard deviation of the cost prediction value, is the weight matrix of the standard deviation branch of the cost prediction value in the third layer, is the bias vector of the standard deviation branch of the cost prediction value in the third layer;

[0081] Based on an evidence lower bound (ELBO), a loss function of the variational Bayesian neural network is defined, and the calculation formula is as follows:

[0082] ;

[0083] ;

[0084] ;

[0085] wherein, is the loss value, The likelihood term measures the fit between the predicted cost and the actual cost, and its value ranges from [value range missing]. , Let KL divergence be a constrained Gaussian variational distribution, with a range of values. , For Gaussian variational distribution The expectation operator under the function, specifically for the function Regarding weight parameters Expected value Indicates that under a given weight parameter Below, real value is created. The conditional probability density function describes the variational Bayesian neural network based on the weight parameters. The degree of match between the generated cost forecast and the actual cost. Create value for the real. For the weight parameters Gaussian variational distribution Weight parameters The prior distribution of the standard normal distribution. For the first The real value created by multi-source heterogeneous records For the first The mean of the cost forecasts for multiple heterogeneous records. This is an index variable representing the number of multi-source heterogeneous data records in the comprehensive cost feature vector. The number of data records in the comprehensive cost feature vector. For the first The standard deviation of the corresponding cost prediction values ​​in each training iteration For the first Weight parameters The mean of the Gaussian variational distribution (mean weight matrix and mean bias vector). For the first Weight parameters The standard deviation of the Gaussian variational distribution (standard deviation weight matrix and standard deviation bias vector). The index variable for the number of weight parameters. The number of weight parameters;

[0086] It should be noted that, regarding the weight parameters This refers to the index of each multi-source heterogeneous data record in the comprehensive cost feature vector. For example, when... "Time" can refer to the daily consumption of cement among various materials consumed daily.

[0087] Calculate the partial derivative of the Gaussian variational distribution of each weight parameter with respect to the loss function, compute the gradient, and use the Adam optimizer to update the weight parameters and their Gaussian variational distributions.

[0088] After the training is completed, the comprehensive cost feature vector is input into the variational Bayesian neural network, the mean and standard deviation of the cost prediction value are output, and the mean of the cost prediction value is output as the cost prediction value;

[0089] Based on the mean and standard deviation of the cost prediction value, a Gaussian distribution assumption is adopted to calculate the probability distribution density function, and a numerical representation of the probability distribution is generated using Monte Carlo sampling;

[0090] It should be noted that the cost prediction value is not only the total cost of the construction project, but also includes the cost of each process, component, material, machinery and labor of the construction project;

[0091] Based on the probability distribution of the cost prediction value, the confidence interval of the cost prediction value is calculated, and the calculation formula is as follows:

[0092] ;

[0093] Wherein, is the confidence interval, indicating the confidence range of the cost prediction value, is the confidence coefficient, which can correspond to a 95% confidence level, based on the standard normal distribution, the value is 1.96, is the fluctuation correction index, the value range is ;

[0094] It should be noted that when , the upper limit of the confidence interval is obtained, and when , the lower limit of the confidence interval is obtained.

[0095] S3, based on the probability distribution and confidence interval of the cost prediction value, a risk heat map is generated.

[0096] According to the probability distribution of the cost prediction value and the upper and lower limits of the confidence interval, the weighted ratio of the standard deviation of the cost prediction value and the width of the confidence interval is calculated as the risk score, and the calculation formula is as follows:

[0097] ;

[0098] Wherein, is the risk score of each multi-source heterogeneous data record, is the upper limit of the confidence interval, is the lower limit of the confidence interval;

[0099] All risk scores are collected and normalized, and a linear color mapping is used by Matplotlib to map the normalized risk scores to a color gradient based on the Viridis color scheme, and a risk heat map is obtained.

[0100] S4, combine the risk heat map and the cost prediction value with the BIM model of the construction project through MR interaction, generate a cost visualization view, and adjust the cost visualization view in real time and evaluate the cost prediction value according to the actual construction condition changes through interaction with the cost visualization view.

[0101] The BIM model of the construction project is constructed, including three-dimensional geometric information of the construction project, such as foundation engineering and main structure, and metadata, such as component number and material type;

[0102] Load the building information model BIM through Microsoft HoloLens 2, combine the risk heat map and the cost prediction value with the BIM model of the construction project, wherein Microsoft HoloLens 2 is a head-mounted display device developed by Microsoft, which aims to provide a mixed reality experience, users can see three-dimensional holograms, and can interact with three-dimensional holograms;

[0103] Specifically, based on the mapping relationship between the record number of the multi-source heterogeneous data and the construction number in the BIM model of the construction project, the cost prediction value is associated with the BIM model of the construction project, for example, the main material (reinforced concrete) cost prediction value 5050 million corresponds to the component "C001" (beam-slab-column), the spatial coordinate system of the BIM model of the construction project is extracted, the bounding box of the BIM model of the construction project is calculated, the spatial range is determined, the center of the BIM model of the construction project is defined as the center of the bounding box, the risk heat map is overlaid to the horizontal plane of the BIM model of the construction project in the form of a semi-transparent grid, the height is fixed as the height of the building center, and a cost visualization view is generated;

[0104] According to the actual construction condition changes, such as construction progress deviation, the cost visualization view is adjusted in real time through gesture recognition of HoloLens 2;

[0105] Specifically, based on the Mixed Reality Toolkit, a grab gesture is defined, for example, two thumbs and index fingers pinch for 0.5 seconds, a distance less than 0.05 meters is detected, and a selection operation is identified, a component in the building information model is selected using the gesture recognition function of the Microsoft HoloLens 2, an interactive interface is displayed floating on the right side of the user's field of view, and contains adjustable comprehensive cost characteristic vectors such as construction progress deviation, material price fluctuation, etc. Each comprehensive cost characteristic vector is represented by a slider, and the initial value corresponds to the real-time comprehensive cost characteristic vector. The user points to the specified comprehensive cost characteristic vector of the interactive interface, and single-handed pinch confirms the adjustment target. Single-handed dragging along the horizontal axis, HoloLens 2 tracks the gesture displacement and maps it to the change amount of the comprehensive cost characteristic vector. The hands are opened, and HoloLens 2 detects the release gesture to confirm the adjustment and update the real-time comprehensive cost characteristic vector. Based on the updated comprehensive cost characteristic vector, the variational Bayesian neural network is input, and a new cost visualization view is calculated in real time;

[0106] The adjustment effect of the cost prediction value and the confidence interval in the updated cost visualization view is analyzed.

[0107] Specifically, the change rate of the new cost prediction value and the original cost prediction value is calculated to obtain the cost change rate, the change rate of the new confidence interval width and the original confidence interval width is calculated to obtain the width change rate, the cost change rate threshold is set according to the mean and standard deviation of the new cost prediction value, and the width change rate threshold is set according to the mean and standard deviation of the new confidence interval width. When the cost change rate is greater than the cost change rate threshold, it indicates that attention should be paid to whether the cost prediction value appears abnormal condition after this adjustment, for example, the cost prediction value is more than 25% different from the same type and same volume of building engineering project, otherwise, it is not. When the width change rate is greater than the width change rate threshold, it indicates that this adjustment significantly increases the uncertainty of the cost prediction value, and the cost prediction value needs to be adjusted and predicted again, otherwise, it is not.

[0108] The embodiment also provides a prediction data evaluation and analysis system based on building engineering cost, comprising a feature vector module, a cost prediction module, a risk map module, and an interactive evaluation module,

[0109] The feature vector module is configured to collect multi-source heterogeneous data of construction engineering cost, and to generate a comprehensive cost feature vector by fusing the multi-head attention mechanism; the cost prediction module is configured to construct a variational Bayesian neural network, to output a cost prediction value and a probability distribution according to the comprehensive cost feature vector, and to calculate a confidence interval of the cost prediction value; the risk map module is configured to generate a risk heat map based on the probability distribution and the confidence interval of the cost prediction value; and the interactive evaluation module is configured to combine the risk heat map and the cost prediction value with a BIM model of the construction engineering by MR interaction, to generate a cost visualization view, to adjust the cost visualization view in real time by interacting with the cost visualization view according to actual construction condition changes, and to evaluate the cost prediction value.

[0110] The embodiment also provides a computer device suitable for the prediction data evaluation and analysis method based on construction engineering cost, which comprises a memory and a processor; the memory is configured to store computer executable instructions, and the processor is configured to execute the computer executable instructions to realize the prediction data evaluation and analysis method based on construction engineering cost proposed in the above embodiment.

[0111] The computer device can be a terminal, which comprises a processor, a memory, a communication interface, a display screen and an input device connected by a system bus. The processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is configured to communicate with external terminals in a wired or wireless manner. The wireless manner can be achieved by WIFI, operator network, NFC (near field communication) or other technologies. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer overlaid on the display screen, or a key, trackball or touchpad arranged on the shell of the computer device. In addition, the input device can be an external keyboard, touchpad or mouse, etc.

[0112] The embodiment also provides a storage medium on which a computer program is stored, the program being executed by a processor to implement the method for evaluating and analyzing prediction data based on construction engineering cost as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0113] To sum up, the application quantifies the prediction uncertainty by replacing the point estimation of the traditional neural network with the weight of the Gaussian variational distribution parameterization network, realizes the multi-dimensional risk aggregation by calculating the risk score through the weighted ratio of the standard deviation and the confidence interval width, and visually displays the heat map to facilitate the resource priority allocation, adjusts the feature vector through the gesture interaction of the HoloLens 2 to realize the dynamic scene deduction, and enables the user to observe and adjust the effect in real time to assist in rapid decision-making.

[0114] It should be noted that the above embodiments are only used to illustrate the technical solutions of the application rather than limit the application. Although the application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the application can be modified or replaced equivalently without departing from the spirit and scope of the technical solutions of the application, and all of them should be covered in the scope of the claims of the application.

Claims

1. A method for evaluating and analyzing predictive data based on construction project cost, characterized in that: include, Collect multi-source heterogeneous data on construction project costs and fuse them through a multi-head attention mechanism to generate a comprehensive cost feature vector; A variational Bayesian neural network is constructed to output the predicted cost and probability distribution based on the comprehensive cost feature vector, and the confidence interval of the predicted cost is calculated. A risk heatmap is generated based on the probability distribution and confidence interval of the cost forecast. Through MR interaction, risk heat maps, cost forecasts, and BIM models of building projects are combined to generate cost visualization views. Based on changes in actual construction conditions, the cost visualization views can be interacted with in real time to adjust the cost visualization views and evaluate cost forecasts.

2. The method for predicting data evaluation analysis based on construction engineering cost according to claim 1, characterized in that: The specific steps for generating the comprehensive cost feature vector are as follows: Collect multi-source heterogeneous data through IoT devices, public databases, and web crawler tools; Multi-source heterogeneous data are recorded in a unified timestamp format and stored as a structured table to obtain the original cost dataset; The original cost dataset is cleaned and standardized to generate a standard cost dataset; A preliminary cost feature matrix is ​​generated by fusing standardized cost datasets through a multi-head attention mechanism. The dimensionality of the preliminary cost feature matrix is ​​reduced by a variational autoencoder to generate a comprehensive cost feature vector.

3. The method for predicting data evaluation analysis based on construction engineering cost according to claim 1, characterized in that: The specific steps for constructing the variational Bayesian neural network are as follows. The initial network structure of the variational Bayesian neural network consists of an input layer and several fully connected layers. Assign Gaussian variational distributions to the weight parameters of each fully connected layer in the variational Bayesian neural network and initialize them; A variational Bayesian neural network is trained by minimizing the loss function.

4. The method for predicting data evaluation analysis based on construction engineering cost according to claim 3, characterized in that: The Gaussian variational distribution for allocating and initializing the weight parameters is described in the following steps. The weight parameters include the weight matrix and bias vector of each fully connected layer in the variational Bayesian neural network. Based on variational inference, Gaussian variational distributions are assigned to the weight matrix and bias terms of each fully connected layer to generate the mean weight matrix, standard deviation weight matrix, mean bias vector, and standard deviation bias vector. Use the Xavier initialization method to initialize the mean weight matrix and standard deviation weight matrix; The mean bias vector and standard deviation bias vector are initialized using a uniform distribution.

5. The method for predicting data evaluation analysis based on construction engineering cost according to claim 3, characterized in that: The specific steps for training the variational Bayesian neural network are as follows. Based on the lower bound of marginal likelihood, the loss function of variational Bayesian neural networks is defined. Based on the loss function, the partial derivative of the loss function with respect to the Gaussian variational distribution of each weight parameter is obtained through backpropagation. Update the weight parameters and their Gaussian variational distribution using the Adam optimizer.

6. The method for predicting data evaluation analysis based on construction engineering cost according to claim 1, characterized in that: The specific steps for outputting the predicted cost and its probability distribution are as follows. The comprehensive cost feature vector is input into the completed variational Bayesian neural network, and the mean and standard deviation branches of the last fully connected layer output the mean and standard deviation of the cost prediction value, respectively. The average of the projected costs will be used as the projected cost value. Based on the mean and standard deviation of the cost forecast, the probability distribution density function is calculated using the Gaussian distribution assumption, and a numerical representation of the probability distribution is generated using Monte Carlo sampling.

7. The method for predicting data evaluation analysis based on construction engineering cost according to claim 1, characterized in that: The specific steps for generating the cost visualization view are as follows: Constructing BIM models for building construction projects; Based on the mapping relationship between the record number and the component number in the BIM model of multi-source heterogeneous data, the cost prediction value is associated with the BIM model of the construction project; The spatial coordinate system of the BIM model of the construction project is extracted, the bounding box of the BIM model of the construction project is calculated, and the center of the BIM model of the construction project is defined as the center of the bounding box; The risk heat map is overlaid on the horizontal plane of the BIM model of the construction project in the form of a semi-transparent grid; The height is fixed as the height of the building center, and the cost visualization view is generated.

8. A construction cost prediction data evaluation analysis system based on any one of the construction cost prediction data evaluation analysis methods according to claims 1 to 7. It includes a feature vector module, a cost prediction module, a risk map module, and an interactive evaluation module, The feature vector module is used to collect multi-source heterogeneous data of construction project cost, and to generate a comprehensive cost feature vector by fusing through a multi-head attention mechanism; The cost prediction module is used to build a variational Bayesian neural network, output a cost prediction value and a probability distribution according to the comprehensive cost feature vector, and calculate the confidence interval of the cost prediction value; The risk map module is used to generate a risk heat map based on the probability distribution and the confidence interval of the cost prediction value; The interactive evaluation module is used to combine the risk heat map, the cost prediction value and the BIM model of the construction project through MR interaction, generate a cost visualization view, adjust the cost visualization view in real time according to the actual construction situation, and evaluate the cost prediction value. 9.A computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that: The processor executes the computer program to realize the steps of the construction project cost prediction data evaluation analysis method according to any one of claims 1-7.

10. A computer readable storage medium having stored thereon a computer program, characterized in that: The computer program is executed by the processor to realize the steps of the construction project cost prediction data evaluation analysis method according to any one of claims 1-7.

Images