Big data-based construction cost analysis system and method

Abstract

The present application relates to the field of engineering construction, in particular to a construction cost analysis system and method based on big data. Through obtaining multi-modal construction data in the system, the multi-modal construction data is sequentially subjected to data preprocessing and standardization, the text and numerical value in the initial multi-modal construction data are aligned, the aligned text numerical value data and image space-time data are used to construct a knowledge graph, an LSTM-GRU hybrid network and an HMM hidden Markov model are connected, a construction cost prediction model is established, a space-time attention mechanism is added to the model, the characteristic multi-modal construction data is input into the model for prediction, the model is iteratively trained through a PPO algorithm, and the materials and manpower in the construction project are uniformly scheduled and managed based on the minimum cost path and the risk plan library. The key space-time area can be dynamically focused, the spatial resolution of the prediction result is improved to the bid section level, and a basis is provided for accurate scheduling.

Description

Technical Field

[0001] This invention relates to the field of network security technology, and in particular to a construction cost analysis system and method based on big data. Background Technology

[0002] Traditional construction cost analysis, relying on manual experience and static historical data, suffers from the following drawbacks: Limited data dimensions: It is based solely on internal company cost records, lacking real-time integration of market dynamics and environmental factors. Furthermore, it fails to effectively collect relevant external data such as the construction costs of surrounding facilities and temporary site rental prices, making it difficult for cost analysis to comprehensively cover the complex economic environment in which the project actually operates. Significant lag: The cost model has a long update cycle and cannot adapt to unforeseen risks. After a risk occurs, the process from information collection and processing to integration into the cost analysis system often takes a considerable amount of time, resulting in severely delayed cost adjustment measures and potentially causing adverse consequences such as tight or even broken project cash flow. Moreover, it is difficult to provide differentiated forecasts for different construction stages, geographical regions, or subcontracted projects. Summary of the Invention

[0003] The purpose of this invention is to solve the above problems by designing a construction cost analysis system and method based on big data.

[0004] To achieve the above objectives, the technical solution of the present invention further includes the following modules in the above-mentioned big data-based construction cost analysis system:

[0005] A construction data acquisition module is used to acquire multimodal construction data in the system, and to perform data preprocessing and standardization on the multimodal construction data in sequence to obtain initial multimodal construction data;

[0006] The feature data alignment module is used to align the text and numerical values ​​in the initial multimodal construction data, and to align the images and spatiotemporal data. The aligned text and numerical data and image spatiotemporal data are used to construct a knowledge graph to obtain the feature multimodal construction data.

[0007] The prediction model building module is used to connect the LSTM-GRU hybrid network and the HMM hidden Markov model to establish an LSTM-GRU-HMM construction cost prediction model. A spatiotemporal attention mechanism is added to the model to obtain the target LSTM-GRU-HMM construction cost prediction model.

[0008] The cost risk prediction module is used to input the feature multimodal construction data into the target LSTM-GRU-HMM construction cost prediction model for prediction. Through the PPO algorithm, the model is trained iteratively to generate a Pareto optimal solution set, and the minimum cost path and risk contingency plan library are obtained.

[0009] The human resources and materials scheduling module is used to uniformly schedule and manage materials and human resources in construction projects based on the minimum cost path and risk contingency plan library.

[0010] Furthermore, in the aforementioned big data-based construction cost analysis system, the construction data acquisition module includes the following sub-modules:

[0011] The data acquisition submodule is used to access the enterprise's ERP system, BIM model and public database through API interface to obtain material inventory data, bill of quantities data and building material price index;

[0012] The image scanning submodule is used to acquire construction log images from the OCR scanner and extract text data from the construction log images;

[0013] The sensor submodule is used to collect site environmental data and human data using temperature and humidity sensors and vibration sensors.

[0014] The standardization submodule is used to perform outlier removal, missing value supplementation, and standardization processing on the material inventory data, bill of quantities data, building material price index, construction log images, text data, construction site environment data, and manpower data to obtain initial multimodal construction data.

[0015] Furthermore, in the aforementioned big data-based construction cost analysis system, the feature data alignment module includes the following sub-modules:

[0016] The extraction submodule is used to extract text name data from the initial multimodal construction data using an NLP natural language processing model, including at least material names, construction period nodes, and risk events; and to associate the text name data with numerical labels to establish a mapping relationship and obtain the first associated data.

[0017] The localization submodule is used to locate the image region in the initial multimodal construction data based on the YOLO object detection model, label the quality defect, obtain the text data in the quality defect label, and obtain the image data.

[0018] The matching submodule is used to match the date and location information in the image data with the 3D coordinates of the BIM model, fuse sensor data and environmental conditions in the image, generate spatiotemporal joint features, and obtain the second associated data.

[0019] A submodule is established to create static and dynamic rule bases for the first and second associated data, and to store heterogeneous data associations using the Neo4j graph database. Multi-hop relationships in the knowledge graph are encoded into feature vectors to obtain feature multimodal construction data.

[0020] Furthermore, in the aforementioned big data-based construction cost analysis system, the prediction model construction module includes the following sub-modules:

[0021] The connection submodule is used to connect the LSTM-GRU hybrid network and the HMM hidden Markov model to obtain the feature vectors of historical construction data in the system. ;

[0022] The extraction submodule is used to extract the feature vector using an LSTM layer. The long-term dependency features are used to output the first hidden state, and the feature vector is learned using a GRU layer. The short-term fluctuation pattern outputs the second hidden state;

[0023] The splicing submodule is used to splice the first hidden state and the second hidden state into a hybrid feature to obtain the initial construction feature data;

[0024] The market policy phase in the historical construction data is defined as the hidden state of the Hidden Markov Model (HMM). Then the state transition matrix of the model as follows:

[0025]

[0026] in, Indicates from state Transition to state The probability, Indicates time The hidden state;

[0027] The generation submodule is used to generate an observation probability distribution using the initial construction feature data, thereby obtaining the observation probability of the Hidden Markov Model (HMM):

[0028]

[0029]

[0030] in, Represents the observation vector. Representing state Generate observation vectors The probability, This represents the initial construction feature data. This represents the activation function. Represents the weight matrix. Indicates bias;

[0031] The computation submodule is used to solve for the optimal state sequence in the model. The macroscopic construction state label sequence data was obtained.

[0032] Furthermore, in the aforementioned big data-based construction cost analysis system, the prediction model construction module includes the following units:

[0033] An attention mechanism addition unit is used to add a spatiotemporal attention mechanism to the LSTM-GRU-HMM construction cost prediction model, and to dynamically allocate cost weights for different regions and construction stages in the macro-construction status label sequence data using the spatiotemporal attention mechanism.

[0034] The prediction weighting unit is used to calculate the importance weight of historical time steps in the macro-construction state label sequence data to the current prediction using a time attention mechanism. ;

[0035]

[0036]

[0037] in, This represents the output of the scoring function in the attention mechanism. This represents the weight matrix in the temporal attention mechanism, applied to the concatenated features. Perform a linear transformation. This represents macro-level construction status label sequence data. Indicates the characteristics of the current moment. This represents the bias term in the time attention mechanism;

[0038] The weight calculation unit is used to utilize spatial attention based on the regional characteristics in the macroscopic construction state label sequence data. Calculate spatial weights ;

[0039]

[0040] in, Index representing a spatial region, It is the total number of spatial regions. Indicates the first Energy value of each spatial region;

[0041] The model obtains units, which are used to combine attention weights with hybrid features to generate the final prediction input. After training, the target LSTM-GRU-HMM construction cost prediction model is obtained.

[0042] Furthermore, in the aforementioned big data-based construction cost analysis system, the cost risk prediction module includes the following units:

[0043] The feature-based multimodal construction data is input into the target LSTM-GRU-HMM construction cost prediction model for prediction.

[0044] The Pareto optimal solution set is generated by training the model iteratively using the PPO algorithm. The optimization of the PPO algorithm includes at least a policy network for outputting the action probability distribution. Value networks are used to evaluate the value of a state. ;

[0045] Importance sampling ratio is:

[0046]

[0047] Pruning objective function:

[0048]

[0049] in, Represents the dominance function. Indicates an immediate reward. Indicates the discount factor. Indicates the clipping threshold. This represents the probability distribution of actions under the old policy. Indicates an action, Indicates state, Expressing expectations, Indicates the current strategy. Indicates the old strategy;

[0050] Define the Pareto dominance relationship, generate a diverse set of solutions by adjusting the weights, select non-dominated solutions to form the Pareto front, and obtain the minimum cost path and risk contingency plan library.

[0051] Furthermore, in the aforementioned big data-based construction cost analysis system, the human and material scheduling module includes the following units:

[0052] The task decomposition unit is used to decompose the construction task into sub-tasks such as material procurement, manpower allocation and equipment scheduling based on the key nodes in the minimum cost path, and generate a spatiotemporal constraint diagram based on the BIM model to identify resource conflict areas.

[0053] The resource allocation unit is used to mark tasks with a cost weight >15% in the path as critical path tasks, prioritize resource allocation, and reserve 20% resource redundancy for high-risk tasks in the contingency plan library.

[0054] The risk assessment unit is used to match a three-level response plan from the contingency plan library based on the type and level of the risk event. The plan includes at least a level 1 response, a level 2 response, and a level 3 response. The level 1 response is used to directly initiate a preset action, the level 2 response is used to generate multiple options, and the level 3 response is used to trigger a manual intervention process.

[0055] Furthermore, in the above-mentioned construction cost analysis method based on big data, the construction cost analysis method includes the following steps:

[0056] The system acquires multimodal construction data, and performs data preprocessing and standardization sequentially on the multimodal construction data to obtain initial multimodal construction data;

[0057] The text and numerical values ​​in the initial multimodal construction data are aligned, and the images and spatiotemporal data are aligned. The aligned text and numerical data and image spatiotemporal data are used to construct a knowledge graph to obtain feature multimodal construction data.

[0058] By connecting the LSTM-GRU hybrid network and the HMM hidden Markov model, an LSTM-GRU-HMM construction cost prediction model is established. A spatiotemporal attention mechanism is added to the model to obtain the target LSTM-GRU-HMM construction cost prediction model.

[0059] The feature multimodal construction data is input into the target LSTM-GRU-HMM construction cost prediction model for prediction. The Pareto optimal solution set is generated by iteratively training the model through the PPO algorithm, and the minimum cost path and risk contingency plan library are obtained.

[0060] Based on the minimum cost path and risk contingency plan library, materials and manpower in the construction project are uniformly scheduled and managed.

[0061] Furthermore, in the above-mentioned construction cost analysis method based on big data, the step of acquiring multimodal construction data from the system, and sequentially preprocessing and standardizing the multimodal construction data to obtain initial multimodal construction data includes:

[0062] By connecting to the enterprise's ERP system, BIM model and public database through API interface, material inventory data, bill of quantities data and building material price index can be obtained;

[0063] Acquire construction log images from an OCR scanner and extract text data from the construction log images;

[0064] Temperature and humidity sensors and vibration sensors are used to collect site environmental data and manpower data in the construction site.

[0065] The material inventory data, bill of quantities data, building material price index, construction log images, text data, site environment data, and manpower data are processed by outlier removal, missing value supplementation, and standardization to obtain initial multimodal construction data.

[0066] Furthermore, in the aforementioned construction cost analysis method based on big data, the method is characterized by aligning the text and numerical values ​​in the initial multimodal construction data, and aligning the images and spatiotemporal data, and constructing a knowledge graph from the aligned text and numerical data and image spatiotemporal data to obtain feature-based multimodal construction data, including:

[0067] The text name data in the initial multimodal construction data is extracted using an NLP natural language processing model, including at least material names, construction period nodes, and risk events; the text name data and numerical labels are associated to establish a mapping relationship and obtain the first associated data;

[0068] Based on the YOLO object detection model, image regions in the initial multimodal construction data are located, quality defect labels are marked, and text data in the quality defect labels are obtained to obtain image data;

[0069] The date and location information in the image data are matched with the 3D coordinates of the BIM model, and the sensor data and the environmental state in the image are fused to generate spatiotemporal joint features, thus obtaining the second associated data.

[0070] Establish static and dynamic rule bases for the first and second associated data, and use the Neo4j graph database to store heterogeneous data associations. Encode multi-hop relationships in the knowledge graph into feature vectors to obtain feature multimodal construction data.

[0071] Its beneficial effects are as follows: by acquiring multimodal construction data from the system, the multimodal construction data is preprocessed and standardized sequentially; the text and numerical values ​​in the initial multimodal construction data are aligned; a knowledge graph is constructed from the aligned text and numerical data and image spatiotemporal data; an LSTM-GRU hybrid network and an HMM hidden Markov model are connected to establish an LSTM-GRU-HMM construction cost prediction model; a spatiotemporal attention mechanism is added to the model; the feature multimodal construction data is input into the model for prediction; the Pareto optimal solution set is generated by iteratively training the model through the PPO algorithm; and the materials and manpower in the construction project are uniformly scheduled and managed based on the minimum cost path and risk contingency plan library. 1. Solves the problem of semantic fragmentation in traditional multimodal data; 2. Can dynamically focus on key spatiotemporal areas (such as high-cost and high-risk sections), improving the spatial resolution of prediction results to the section level, providing a basis for precise scheduling; 3. Based on historical data and real-time prediction, automatically generates emergency plans for more than 20 risk scenarios (such as rainstorms and material shortages), with a response speed several times faster than manual formulation; 4. Combines the minimum cost path with the risk plan library to achieve dynamic allocation of materials and manpower, improving resource utilization. Attached Figure Description

[0072] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0073] Figure 1 This is a schematic diagram of the first embodiment of the construction cost analysis system based on big data in this invention.

[0074] Figure 2 This is a schematic diagram of a second embodiment of the construction cost analysis system based on big data in this invention.

[0075] Figure 3 This is a schematic diagram of the third embodiment of the construction cost analysis system based on big data in this invention.

[0076] Figure 4 This is a schematic diagram of the first embodiment of the construction cost analysis method based on big data in this invention. Detailed Implementation

[0077] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0078] Those skilled in the art will understand that, unless specifically stated otherwise, the singular forms “a,” “an,” and “the” used herein may also include the plural forms. It should be further understood that the term “comprising” as used in this specification means the presence of the stated features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0079] The present invention will now be described in detail with reference to the accompanying drawings, such as... Figure 1 As shown, the construction cost analysis system based on big data includes the following modules:

[0080] 101. Construction data acquisition module, used to acquire multimodal construction data in the system, and perform data preprocessing and standardization on the multimodal construction data in sequence to obtain initial multimodal construction data;

[0081] Specifically, this embodiment also includes a data acquisition submodule, which is used to access the enterprise ERP system, BIM model and public database through API interface to obtain material inventory data, bill of quantities data and building material price index;

[0082] The image scanning submodule is used to acquire construction log images from the OCR scanner and extract text data from the construction log images;

[0083] The sensor submodule is used to collect site environmental data and human data using temperature and humidity sensors and vibration sensors.

[0084] The standardization submodule is used to perform outlier removal, missing value supplementation, and standardization processing on material inventory data, bill of quantities data, building material price index, construction log images, text data, construction site environmental data, and manpower data to obtain initial multimodal construction data.

[0085] 102. Feature data alignment module, used to align text and numerical data in the initial multimodal construction data, and to align images and spatiotemporal data. The aligned text and numerical data and image spatiotemporal data are used to construct a knowledge graph to obtain feature multimodal construction data.

[0086] Specifically, this embodiment also includes an extraction submodule, used to extract text name data from the initial multimodal construction data using an NLP natural language processing model, including at least material names, construction period nodes, and risk events; and to associate the text name data with numerical labels to establish a mapping relationship and obtain the first associated data.

[0087] The localization submodule is used to locate image regions in the initial multimodal construction data based on the YOLO object detection model, label quality defects, obtain text data from the quality defect labels, and obtain image data.

[0088] The matching submodule is used to match the date and location information in the image data with the 3D coordinates of the BIM model, fuse sensor data with the environmental status in the image, generate spatiotemporal joint features, and obtain the second associated data.

[0089] A submodule is established to create static and dynamic rule bases for the first and second associated data, and to store heterogeneous data associations using the Neo4j graph database. Multi-hop relationships in the knowledge graph are encoded into feature vectors to obtain feature multimodal construction data.

[0090] 103. Prediction model building module, used to connect LSTM-GRU hybrid network and HMM hidden Markov model to establish LSTM-GRU-HMM construction cost prediction model, add spatiotemporal attention mechanism to model to obtain target LSTM-GRU-HMM construction cost prediction model.

[0091] Specifically, this embodiment also includes a connection submodule, used to connect the LSTM-GRU hybrid network and the HMM hidden Markov model to obtain the feature vector of historical construction data in the system. .

[0092] The extraction submodule is used to extract feature vectors using LSTM layers. The long-term dependency features are used to output the first hidden state, and the feature vector is learned using a GRU layer. The short-term fluctuation pattern outputs the second hidden state;

[0093] The splicing submodule is used to splice the first hidden state and the second hidden state into a hybrid feature to obtain the initial construction feature data;

[0094] Define the market policy phase in historical construction data as the hidden state of the Hidden Markov Model (HMM). Then the state transition matrix of the model as follows:

[0095]

[0096] in, Indicates from state Transition to state The probability, Indicates time The hidden state;

[0097] The generation submodule is used to generate the observation probability distribution using the initial construction feature data, thus obtaining the observation probability of the Hidden Markov Model (HMM):

[0098]

[0099]

[0100] in, Represents the observation vector. Representing state Generate observation vectors The probability, This represents the initial construction feature data. This represents the activation function. Represents the weight matrix. Indicates bias;

[0101] The computation submodule is used to solve for the optimal state sequence in the model. The macroscopic construction state label sequence data was obtained.

[0102] An attention mechanism addition unit is used to add a spatiotemporal attention mechanism to the LSTM-GRU-HMM construction cost prediction model, and to dynamically allocate cost weights for different regions and construction stages in the macro-construction state label sequence data using the spatiotemporal attention mechanism.

[0103] The prediction weighting unit is used to calculate the importance weight of historical time steps in the macro-construction state label sequence data to the current prediction using a time attention mechanism. ;

[0104]

[0105]

[0106] in, This represents the output of the scoring function in the attention mechanism. This represents the weight matrix in the temporal attention mechanism, applied to the concatenated features. Perform a linear transformation. This represents macro-level construction status label sequence data. Indicates the characteristics of the current moment. This represents the bias term in the time attention mechanism;

[0107] Importance weight This allows the model to automatically identify and reinforce historical segments that are strongly correlated with the current prediction. For example, when predicting the concrete pouring cost of a certain section, if material price fluctuations or extreme weather events occurred during the same period in the historical data, the importance weight of the corresponding time step will be adjusted. This will significantly increase the contribution of key historical information (such as the price jump in steel bars 3 months ago) to more than 60% of the total prediction, while weakening the impact of irrelevant periods (such as data from a stable period 1 year ago). This dynamic focusing mechanism solves the problem of long-term dependency loss caused by fixed time windows in traditional RNNs.

[0108] Importance weight The distribution reflects the confidence level of historical data in the current prediction. When the model encounters data noise or outliers (such as cost recording bias caused by sensor malfunctions), the importance weights of relevant time steps are adjusted. It will automatically reduce the weight of the data to avoid transmitting incorrect information. For example, if the cost data for a certain week shows an abnormally low value due to equipment failure, the time attention mechanism will adjust its weight accordingly. By compressing the value to below 0.1 and simultaneously increasing the weights of preceding and following normal weeks (total > 0.9), abnormal disturbances are suppressed while maintaining temporal continuity. This ability to quantify uncertainty reduces the model's prediction error by 15%-20% in non-stationary construction environments (such as policy adjustment periods).

[0109] By adjusting the importance weight With its computational granularity, the model can simultaneously capture both short-term fluctuations and long-term trends. In daily-level forecasting, importance weights... The data will focus on the most recent 7 days (accounting for >70%) to reflect changes in construction progress; in monthly forecasts, the importance weighting will be... The data is then evenly distributed over the past three months, highlighting the impact of material procurement cycles. This multi-scale fusion enables the model to accurately capture daily unexpected events (such as downtime costs due to equipment failure) and accurately predict monthly funding needs (such as progress payments) in section-level cost forecasting. The spatial resolution is improved to specific construction areas (such as the main tower pouring area), resulting in an improvement of more than 30% compared to traditional methods.

[0110] The weighting calculation unit is used to utilize spatial attention based on regional characteristics in the macroscopic construction state label sequence data. Calculate spatial weights ;

[0111]

[0112] in, Index representing a spatial region, It is the total number of spatial regions. Indicates the first The energy value of each spatial region.

[0113] The model obtains units, which are used to combine attention weights with hybrid features to generate the final prediction input. After training, the target LSTM-GRU-HMM construction cost prediction model is obtained.

[0114] The LSTM-GRU-HMM construction cost prediction model integrates the long-term memory capability of LSTM, the computational efficiency of GRU, and the probabilistic reasoning advantages of HMM. Combined with a spatiotemporal attention mechanism, it significantly improves the ability to capture dynamic changes in construction costs. The following explanation covers the network structure, hyperparameter values, and training steps:

[0115] The network structure adopts a layered hybrid architecture. The bottom layer is an LSTM-GRU parallel module, the middle layer is a spatiotemporal attention mechanism, and the top layer is an HMM state transition layer. The LSTM module controls the long-term information flow through forget gates, input gates, and output gates, and the hidden layer size can be 128-256, which is suitable for handling long-term dependent factors in construction costs (such as material price cycles). The GRU module simplifies the calculation through update gates and reset gates, and the hidden layer size can be 64-128, which is suitable for capturing short-term fluctuations (such as changes in construction progress). The outputs of the two are concatenated or weighted and then input into the spatiotemporal attention layer. The spatiotemporal attention mechanism includes spatial attention (assigning weights to different cost items) and temporal attention (adjusting the contribution of different time steps). The output feature vector is then input into the HMM layer. The HMM layer defines the hidden states (such as cost increase, decrease, and stability) and observation probabilities. The optimal state sequence is decoded using the Viterbi algorithm, and the final cost prediction value is output.

[0116] The hyperparameter values ​​should balance performance and computational efficiency. The learning rate can be initially set to 1e-3, using a dynamic adjustment strategy (e.g., ReduceLROnPlateau), reducing it to 1e-4 if the validation loss does not decrease for 5 consecutive rounds. The dropout rate should be set to 0.2-0.3 between the LSTM and GRU input layers and 0.1-0.2 between hidden layers to prevent overfitting. The batch size should be selected based on the data volume: 32-64 for small datasets (e.g., single projects) and 128-256 for large datasets (e.g., regional projects). The sequence length should cover the cost change cycle: 10-30 for short-term predictions (e.g., monthly) and 50-100 for long-term predictions (e.g., annual). The number of iterations is controlled by early stopping, with a patience of 10 and a maximum of 200 iterations. For weight initialization, LSTM uses Xavier initialization for stable gradients, and GRU uses He initialization to accelerate convergence. The recommended optimizer is Adam, with β1=0.9 and β2=0.999, which is suitable for noisy data. If the data is stable, Nadam can be used instead. The number of hidden states in the HMM should be set according to the cost fluctuation pattern. For simple scenarios (such as fixed processes), use 3-5, and for complex scenarios (such as multiple overlapping processes), use 5-8.

[0117] The training process begins with data preprocessing. The construction cost sequence is normalized to [0,1] using Min-Max, and then divided into a training set (70%), a validation set (15%), and a test set (15%). The data is split according to sequence length to ensure that each sample contains a complete cost cycle.

[0118] The model is then built using a hybrid network implemented in PyTorch. The outputs of the LSTM-GRU module are concatenated and then reduced to feature vectors through a fully connected layer. The spatiotemporal attention layer calculates spatial weights (softmax assigns cost term weights) and temporal weights (adjusts time step contributions), outputting weighted features. The HMM layer initializes the state transition matrix (random or based on prior knowledge) and the observation probability matrix.

[0119] Next, the training configuration was performed, using a loss function combining MSE regression and HMM log-likelihood, with Adam as the optimizer and ReduceLROnPlateau as the learning rate scheduler. GPU acceleration was selected, and batch processing was grouped by batch size.

[0120] The training loop then proceeds, with forward propagation sequentially through the LSTM-GRU, attention layer, and HMM layer. After calculating the total loss, backpropagation updates the parameters. After each validation round, if the loss does not decrease, the learning rate is reduced. Early stopping is used to monitor the validation loss; training terminates if the loss does not decrease for 10 consecutive rounds.

[0121] Finally, optimization and evaluation were performed. Optuna was used to search for optimal hyperparameters (such as hidden layer size and learning rate), and the evaluation metrics were MAE, RMSE (regression), and accuracy (HMM state prediction). The model's generalization ability was verified on the test set, and the sources of error (such as data noise or model structure) were analyzed. Finally, the LSTM-GRU-HMM construction cost prediction model was obtained.

[0122] 104. Cost and risk prediction module: This module is used to input feature multimodal construction data into the target LSTM-GRU-HMM construction cost prediction model for prediction. Through the PPO algorithm, the model is trained iteratively to generate a Pareto optimal solution set, and obtain the minimum cost path and risk contingency plan library.

[0123] Specifically, this embodiment also includes inputting the feature multimodal construction data into the target LSTM-GRU-HMM construction cost prediction model for prediction;

[0124] The Pareto optimal solution set is generated through iterative training of the model using the PPO algorithm. The optimization of the PPO algorithm includes at least a policy network for outputting the action probability distribution. Value networks are used to evaluate the value of a state. ;

[0125] Importance sampling ratio is:

[0126]

[0127] Pruning objective function:

[0128]

[0129] in, Represents the dominance function. Indicates an immediate reward. Indicates the discount factor. Indicates the clipping threshold. This represents the probability distribution of actions under the old policy. Indicates an action, Indicates state, Expressing expectations, Indicates the current strategy. Indicates the old strategy;

[0130] Define the Pareto dominance relationship, generate a diverse set of solutions by adjusting the weights, select non-dominated solutions to form the Pareto front, and obtain the minimum cost path and risk contingency plan library.

[0131] In optimizing the LSTM-GRU-HMM construction cost prediction model based on the PPO algorithm, the reward function design needs to quantify multiple objectives such as cost and risk into calculable, immediate rewards to guide the model in generating a Pareto optimal solution set. Specifically, the cost objective can be quantified as the reciprocal of the deviation between the predicted cost and the actual cost, i.e.:

[0132]

[0133] Where Cpred is the model-predicted cost and Ctrue is the actual cost. This function maps the deviation to the interval (0,1], with smaller deviations resulting in higher rewards, encouraging accurate model predictions. The risk objective is defined by the negative logarithm of risk indicators (such as the probability of cost overruns and the probability of project delays), i.e.:

[0134]

[0135] Where Prisk is the risk probability evaluated by the model, and ϵ is a local minimum (e.g., 1e-5) to avoid the undefined logarithm. This design ensures that high-risk scenarios receive significant negative rewards, forcing the model to avoid risky paths. The overall reward is calculated using a weighted summation:

[0136]

[0137] The weights wcost and wrisk are determined through Pareto front analysis. Balancing cost minimization and risk control, the model is iteratively optimized through the PPO algorithm. In multiple rounds of training, the model gradually approaches the Pareto front, ultimately generating a minimum cost path and corresponding risk contingency plan library covering different risk-cost trade-offs.

[0138] 105. The Human Resources and Materials Scheduling Module is used to uniformly schedule and manage materials and human resources in construction projects based on the minimum cost path and risk contingency plan library.

[0139] Specifically, this embodiment also includes a task decomposition unit, which is used to decompose the construction task into sub-tasks of material procurement, manpower allocation and equipment scheduling according to the key nodes in the minimum cost path, generate a spatiotemporal constraint diagram based on the BIM model, and identify resource conflict areas.

[0140] The resource allocation unit is used to mark tasks with a cost weight >15% in the path as critical path tasks, prioritize resource allocation, and reserve 20% resource redundancy for high-risk tasks in the contingency plan library.

[0141] The risk assessment unit is used to match three-level response plans from the contingency plan library based on the type and level of the risk event. It includes at least level one, level two, and level three responses. Level one response is used to directly initiate preset actions, level two response is used to generate multiple options, and level three response is used to trigger manual intervention procedures.

[0142] Sub-step 1: Resource scheduling plan decomposition and priority setting;

[0143] Path parsing and task decomposition:

[0144] Based on the key node in the minimum cost path (prioritizing the procurement of local sand and gravel), the construction task is broken down into sub-tasks of material procurement (material types and quantities), manpower allocation (job types and shifts), and equipment scheduling (tower crane and excavator usage periods).

[0145] A spatiotemporal constraint diagram is generated based on the BIM model to identify resource conflict areas (concrete pouring in area A and rebar tying in area B need to share a tower crane).

[0146] Dynamic priority sorting:

[0147] Cost-sensitive tasks: Tasks with a cost weight >15% in the path (high-priced material procurement) are marked as "critical path tasks" and given priority in resource allocation;

[0148] Risk avoidance task: Reserve 20% of resources for high-risk tasks (construction during the rainy season) in advance in the contingency plan library.

[0149] Sub-step 2: Multi-level inventory and flexible workforce pool management;

[0150] Material Tiered Inventory Strategy:

[0151] Strategic inventory: Stock up on long-cycle materials (special steel) three months in advance based on forecasted demand;

[0152] Tactical inventory: For volatile materials (cement), a "Vendor Managed Inventory (VMI)" model is adopted, with weekly rolling replenishment;

[0153] Emergency inventory: Establish safety stock for materials (waterproof membrane) in the risk contingency plan, and automatically replenish the stock when a threshold is triggered.

[0154] Flexible manpower dispatch mechanism:

[0155] The core team is stable: technical workers (welders, surveyors) are stationed on-site for extended periods according to the project cycle;

[0156] Temporary manpower pool: General-purpose workers (manual laborers) are hired on a daily basis through the labor platform, and the pool is dynamically expanded or reduced according to the progress of the work.

[0157] Cross-project allocation: Share an enterprise-level human resource pool and draw personnel from low-priority projects during peak periods.

[0158] Sub-step 3: Real-time monitoring and dynamic scheduling engine;

[0159] Resource Status Dashboard:

[0160] The system provides a visual display of real-time resource utilization (tower crane utilization rate 82%), inventory warnings (steel reserves below the safety line), and manpower shortages (3 night shift bricklayers are short).

[0161] Dynamic scheduling rule engine:

[0162] Rule base: Preset "If a material is delayed by more than 3 days, activate the backup supplier and adjust the construction sequence";

[0163] Automatic triggering mechanism: When the sensor detects "continuous heavy rain > 6 hours", the "drainage plan" in the contingency plan library will be automatically pushed and outdoor operations will be suspended.

[0164] Sub-step 4: Risk contingency plan integration and emergency response;

[0165] Contingency plan matching and execution:

[0166] Based on the risk event type (supply chain disruption, extreme weather) and level (high / medium / low), match a three-level response plan from the contingency plan database:

[0167] Level 1 response: Directly initiate the preset action (call the strategy inventory);

[0168] Level 2 response: Generates multiple alternatives (adjusting processes or delaying construction) for decision-making;

[0169] Level 3 response: Triggering manual intervention process (reporting to headquarters for resource coordination).

[0170] Post-event review and contingency plan iteration:

[0171] Record the results of emergency response and optimize the contingency plan database through feedback loop (e.g., if a supplier repeatedly defaults, lower its priority).

[0172] Sub-step 5: Deployment of the multi-dimensional collaborative management platform;

[0173] Platform functional modules:

[0174] Resource Dashboard: Integrates cost forecasting, inventory, manpower and equipment data, and supports multi-dimensional drill-down analysis;

[0175] Mobile collaboration: Site administrators receive scheduling instructions and report execution status via an app;

[0176] Supplier Portal: Provides open interfaces for suppliers to check order status and report supply anomalies.

[0177] Cross-departmental collaboration process:

[0178] Procurement-Construction Collaboration: The procurement department dynamically adjusts order delivery times based on the construction progress;

[0179] Finance-Engineering Reconciliation: Automatically generates reports on resource consumption and cost deviations, supporting rapid correction.

[0180] Its beneficial effects are as follows: 1. It solves the problem of semantic fragmentation in traditional multimodal data; 2. It can dynamically focus on key spatiotemporal areas (such as high-cost and high-risk sections), improving the spatial resolution of prediction results to the section level and providing a basis for precise scheduling; 3. Based on historical data and real-time prediction, it automatically generates emergency plans for more than 20 risk scenarios (such as rainstorms and material shortages), with a response speed several times faster than manual formulation; 4. By combining the minimum cost path with the risk plan library, it realizes the dynamic allocation of materials and manpower, improving resource utilization.

[0181] In this embodiment, please refer to Figure 2 The second embodiment of the construction cost analysis system based on big data in this invention includes a feature data alignment module comprising the following sub-modules:

[0182] The extraction submodule is used to extract text name data from the initial multimodal construction data using an NLP natural language processing model, including at least material names, project schedules, and risk events; it also associates the text name data with numerical labels to establish a mapping relationship and obtain the first associated data.

[0183] The localization submodule is used to locate image regions in the initial multimodal construction data based on the YOLO object detection model, label quality defects, obtain text data from the quality defect labels, and obtain image data.

[0184] The matching submodule is used to match the date and location information in the image data with the 3D coordinates of the BIM model, fuse sensor data with the environmental status in the image, generate spatiotemporal joint features, and obtain the second associated data.

[0185] A submodule is established to create static and dynamic rule bases for the first and second associated data, and to store heterogeneous data associations using the Neo4j graph database. Multi-hop relationships in the knowledge graph are encoded into feature vectors to obtain feature multimodal construction data.

[0186] The construction of knowledge graphs achieves the mapping from raw data to structured knowledge through multimodal data fusion. The following example, using a bridge engineering project, fully demonstrates a knowledge graph rule instance:

[0187] The raw data contains three categories:

[0188] Text data (NLP extraction): The construction log records "On May 10, 2023, the C50 concrete pouring of the main tower was completed, using 120 tons of HRB400 steel bars, and 3 honeycomb pitting defects were found";

[0189] Image data (YOLO positioning): Photos of the pouring site taken by drone, labeled "Honeycomb defect at the southeast corner of the main tower (L5m×W0.3m×D0.1m)" and timestamp "2023-05-10 14:20";

[0190] BIM model data: 3D coordinates of the main tower (X=120.5, Y=85.2, Z=68.3) and ambient temperature and humidity monitored by sensors (28℃, 65%RH).

[0191] Rule instance mapping process:

[0192] The NLP extraction submodule identifies the entities "HRB400 steel bar" (material name), "2023-05-10" (construction date milestone), and "honeycomb surface" (risk event) in the text, associates them with the numerical label "120 tons," and generates the first associated data.

[0193] (HRB400 steel bars) - [Consumption] -> (120 tons);

[0194] (2023-05-10) - [Risk Event] -> (Honeycomb Pockmarked Surface);

[0195] The YOLO localization submodule locates the coordinates of the defect area (X=120.8, Y=85.0, Z=68.0) in the image, extracts the text label "cellular defect", and generates image data.

[0196] (Southeast corner of the main tower) - [Defect Type] -> (Honeycomb Defect) - [Dimensions] -> (L5m × W0.3m × D0.1m);

[0197] Matching submodule: Matches the image date "2023-05-10" with the BIM model coordinates, integrates sensor temperature and humidity data, and generates second-level associated data.

[0198] (2023-05-10 14:20) - [Environmental Conditions] -> (Temperature 28℃ / Humidity 65%RH) - [Spatial Location] -> (X120.8, Y85.0, Z68.0);

[0199] Establish submodules: Build multi-hop relationships in Neo4j using static rules (e.g., "cellular defects → rework required") and dynamic rules (e.g., "temperature and humidity > 25℃ / 60% → concrete strength decreases by 5%).

[0200] (HRB400 steel reinforcement) - [used for] -> (main tower pouring) - [occurrence time] -> (2023-05-10);

[0201] (Main tower pouring) - [Defects occur] -> (Cellular defects) - [Associated risks] -> (Construction delay of 3 days);

[0202] (2023-05-10) - [Environmental Impact] -> (Concrete strength C45 → C42.75);

[0203] The final encoding is a feature vector [material consumption, defect level, environmental factors, risk probability], forming feature multimodal data that supports the spatiotemporal attention mechanism.

[0204] Its beneficial effect lies in solving the problem of semantic fragmentation in traditional multimodal data. In knowledge graph construction, entity relationship extraction technology is used to associate component information in engineering drawings with construction progress data, forming a structured knowledge network.

[0205] In this embodiment, please refer to Figure 3 The third embodiment of the construction cost analysis system based on big data in this invention includes a human resources and material scheduling module comprising the following units:

[0206] The task decomposition unit is used to decompose the construction task into sub-tasks such as material procurement, manpower allocation and equipment scheduling based on the key nodes in the minimum cost path, and generate a spatiotemporal constraint diagram based on the BIM model to identify resource conflict areas.

[0207] The resource allocation unit is used to mark tasks with a cost weight >15% in the path as critical path tasks, prioritize resource allocation, and reserve 20% resource redundancy for high-risk tasks in the contingency plan library.

[0208] The risk assessment unit is used to match three-level response plans from the contingency plan library based on the type and level of the risk event. It includes at least level one, level two, and level three responses. Level one response is used to directly initiate preset actions, level two response is used to generate multiple options, and level three response is used to trigger manual intervention procedures.

[0209] Its beneficial effect lies in achieving dynamic allocation of materials and manpower by combining the minimum cost path with a risk contingency plan library. When the demand for concrete in a certain section surges, the system automatically calls upon the inventory of nearby warehouses and adjusts transportation routes to reduce costs.

[0210] The above describes the construction cost analysis system based on big data provided by the embodiments of the present invention. The following describes the construction cost analysis method based on big data according to the embodiments of the present invention. Please refer to [link / reference]. Figure 4 One embodiment of the construction cost analysis method based on big data in this invention includes:

[0211] Step 401: Obtain the multimodal construction data from the system, and perform data preprocessing and standardization on the multimodal construction data in sequence to obtain the initial multimodal construction data;

[0212] Step 402: Align the text and numerical data in the initial multimodal construction data, and align the images and spatiotemporal data. Construct a knowledge graph from the aligned text and numerical data and image spatiotemporal data to obtain feature multimodal construction data.

[0213] Step 403: Connect the LSTM-GRU hybrid network and the HMM hidden Markov model to establish the LSTM-GRU-HMM construction cost prediction model. Add a spatiotemporal attention mechanism to the model to obtain the target LSTM-GRU-HMM construction cost prediction model.

[0214] Step 404: Input the feature multimodal construction data into the target LSTM-GRU-HMM construction cost prediction model for prediction. Iterate the training of the model through the PPO algorithm to generate the Pareto optimal solution set and obtain the minimum cost path and risk contingency plan library.

[0215] Step 405: Conduct unified scheduling and management of materials and manpower in the construction project based on the minimum cost path and risk contingency plan library.

[0216] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A construction cost analysis system based on big data, characterized by, The construction cost analysis system comprises the following modules: A construction data acquisition module is configured to acquire multi-modal construction data in the system, sequentially perform data preprocessing and standardization on the multi-modal construction data, and obtain initial multi-modal construction data; A feature data alignment module is configured to align text and numerical values in the initial multi-modal construction data, align images and space-time, construct a knowledge graph based on the aligned text and numerical value data and image and space-time data, and obtain feature multi-modal construction data; A prediction model construction module is configured to connect an LSTM-GRU hybrid network and an HMM hidden Markov model, establish an LSTM-GRU-HMM construction cost prediction model, add a space-time attention mechanism to the model, and obtain a target LSTM-GRU-HMM construction cost prediction model; The prediction model construction module comprises the following sub-modules: A connection module is configured to connect the LSTM-GRU hybrid network and the HMM hidden Markov model to obtain a feature vector of historical construction data in the system ; extracting a sub-module for extracting the feature vector using an LSTM layer outputting a first hidden state, learning the feature vector using a GRU layer outputting a second hidden state A splicing sub-module is configured to splice the first hidden state and the second hidden state into a hybrid feature, and obtain initial construction feature data; The market policy stage in the historical construction data is defined as the hidden state of the HMM The state transition matrix of the model is as follows: As follows: ； wherein, denotes the probability of transitioning from state to state , denotes the hidden state at time ; A generation sub-module is configured to generate an observation probability distribution using the initial construction feature data, and obtain an observation probability of an HMM hidden Markov model: ； ； wherein, represents an observation vector, represents a state generating an observation vector probability, represents initial construction feature data, represents an activation function, represents a weight matrix, represents a bias; A computing sub-module is configured to solve the optimal state sequence in the model to obtain the macro construction state label sequence data The prediction model construction module comprises the following units: An attention mechanism adding unit is configured to add a space-time attention mechanism to the LSTM-GRU-HMM construction cost prediction model, and dynamically allocate cost weights of different regions and construction stages in the macro construction label sequence data using the space-time attention mechanism; a prediction weight unit configured to calculate an importance weight of a historical time step in the macro construction state label sequence data for a current prediction using a temporal attention mechanism ; ； ； wherein, denotes the output of the scoring function in the attention mechanism, denotes the weight matrix in the temporal attention mechanism, which linearly transforms the concatenated features denotes the output of the scoring function in the attention mechanism, denotes the macro construction label sequence data, denotes the feature at the current time, denotes the bias term in the temporal attention mechanism; A weight calculation unit is configured to calculate spatial weights according to region features in the macro construction state label sequence data Computing spatial weights ; ； wherein, an index representing a spatial region, is a total number of spatial regions, represents an energy value of the th spatial region; A model obtaining unit is configured to combine the attention weight and the hybrid feature to generate a final prediction input, and obtain the target LSTM-GRU-HMM construction cost prediction model after training; A cost risk prediction module is configured to input the feature multi-modal construction data into the target LSTM-GRU-HMM construction cost prediction model for prediction, generate a Pareto optimal solution set through training of a PPO iteration model, and obtain a minimum cost path and a risk plan library; The cost risk prediction module comprises the following units: The feature multi-modal construction data is input into the target LSTM-GRU-HMM construction cost prediction model for prediction; Through training of a PPO iterative model, a Pareto optimal solution set is generated, the PPO optimization including at least a policy network for outputting an action probability distribution ; a value network for evaluating a state value ; The importance sampling ratio is: ； The clipping target function is: ； wherein, represents the advantage function, represents the immediate reward, represents the discount factor, represents the clipping threshold, represents the action probability distribution under the old policy, represents the action, represents the state, represents the expectation, represents the current policy, represents the old policy; A Pareto dominance relationship is defined, a diversity solution set is generated by adjusting weights, non-dominated solutions are screened to form a Pareto front, and a minimum cost path and a risk plan library are obtained; A human and material scheduling module is configured to uniformly schedule and manage materials and manpower in a construction project based on the minimum cost path and the risk plan library.

2. The big data based construction cost analysis system of claim 1, wherein, The construction data acquisition module comprises the following sub-modules: A data acquisition sub-module is configured to access an enterprise ERP system, a BIM model, and a public database through an API interface, acquire material inventory data, bill of quantities data, and building material price indexes; An image scanning sub-module is configured to acquire construction log images in an OCR scanner, and extract text data of the construction log images; A sensor sub-module is configured to collect site environment data and manpower data in a construction site using a temperature and humidity sensor and a vibration sensor; The standardized submodule is configured to perform outlier deletion, missing value supplementation and standardization processing on the material inventory data, the bill of quantities data, the building material price index, the construction log image, the text data, the construction site environment data and the human resource data, so as to obtain initial multi-modal construction data.

3. The big data based construction cost analysis system as claimed in claim 1, wherein, The feature data alignment module comprises the following submodules: The extraction submodule is configured to extract text name data in the initial multi-modal construction data by using an NLP natural language processing model, the text name data at least including material names, construction period nodes and risk events, and to establish a mapping relationship by associating the text name data with numerical value labels, so as to obtain first associated data; The positioning submodule is configured to position an image region in the initial multi-modal construction data based on a YOLO target detection model, label quality defect labels, and acquire text data in the quality defect labels, so as to obtain image data; The matching submodule is configured to match date and location information in the image data with 3D coordinates of a BIM model, fuse sensor data and environment states in the image, generate spatio-temporal joint features, and obtain second associated data; The establishment submodule is configured to establish static rule bases and dynamic rule bases of the first associated data and the second associated data, store heterogeneous data associations by using a Neo4j graph database, encode multi-hop relationships in a knowledge graph as feature vectors, and obtain feature multi-modal construction data.

4. The big data based construction cost analysis system of claim 1, wherein, The human resource and material scheduling module comprises the following units: The task decomposition unit is configured to decompose construction tasks into material procurement, human resource allocation and equipment scheduling subtasks according to key nodes in a minimum cost path, generate a spatio-temporal constraint graph based on a BIM model, and identify resource conflict areas; The resource allocation unit is configured to mark tasks with a cost weight > 15% in the path as key path tasks, preferentially allocate resources, and reserve 20% resource redundancy for high-risk tasks in the contingency library; The risk judgment unit is configured to match a three-level response scheme from the contingency library according to the type and level of the risk event, the three-level response scheme at least including a first-level response, a second-level response and a third-level response, the first-level response being used to directly start a preset action, the second-level response being used to generate a multi-scheme selection, and the third-level response being used to trigger a manual intervention process.

5. A construction cost analysis method based on big data, characterized by, The construction cost analysis method comprises the following steps: Obtain multi-modal construction data in a system, and sequentially perform data preprocessing and standardization on the multi-modal construction data to obtain initial multi-modal construction data; Align text and numerical values in the initial multi-modal construction data, and align images and space-time, construct a knowledge graph from the aligned text and numerical value data and image and space-time data, and obtain feature multi-modal construction data; Connect an LSTM-GRU hybrid network and an HMM hidden Markov model to establish an LSTM-GRU-HMM construction cost prediction model, add a spatio-temporal attention mechanism to the model, and obtain a target LSTM-GRU-HMM construction cost prediction model; The LSTM-GRU hybrid network and an HMM hidden Markov model are connected to obtain a feature vector of historical construction data in the system ; extracting the feature vectors using LSTM layers outputting a first hidden state using the long-term dependency features of the feature vectors outputting a second hidden state using the short-term fluctuation patterns of the feature vectors learned using GRU layers Concatenate the first hidden state and the second hidden state into a hybrid feature to obtain initial construction feature data; The market policy stage in the historical construction data is defined as the hidden state of the HMM The state transition matrix of the model is as follows: ​ ； wherein, denotes the probability of transitioning from state to state , denotes the hidden state at time ; Generate an observation probability distribution by using the initial construction feature data to obtain an observation probability of an HMM hidden Markov model: ； ； wherein, represents an observation vector, represents a state generating an observation vector probability, represents initial construction feature data, represents an activation function, represents a weight matrix, represents a bias; Solving the optimal state sequence in the model , to obtain the macro construction state label sequence data; The spatio-temporal attention mechanism is added to the LSTM-GRU-HMM construction cost prediction model, and the spatio-temporal attention mechanism is used to dynamically allocate the cost weights of different regions and construction stages in the macro construction label sequence data; calculating, using a temporal attention mechanism, an importance weight of a historical time step in the macro construction state label sequence data for a current prediction ; ； ； wherein, denotes the output of the scoring function in the attention mechanism, denotes the weight matrix in the temporal attention mechanism, which is applied to the concatenated features performs a linear transformation, denotes the macro construction label sequence data, denotes the features at the current time instant, denotes the bias term in the temporal attention mechanism; Utilizing spatial attention according to regional features in the macro construction state label sequence data Computing spatial weights ; ； wherein, an index representing a spatial region, is a total number of spatial regions, represents an energy value of the th spatial region; The attention weight and the mixed feature are combined to generate a final prediction input, and a target LSTM-GRU-HMM construction cost prediction model is obtained after training; The feature multi-modal construction data is input into the target LSTM-GRU-HMM construction cost prediction model for prediction, and a Pareto optimal solution set is generated through the training of the PPO iteration model, and a minimum cost path and a risk plan library are obtained; The feature multi-modal construction data is input into the target LSTM-GRU-HMM construction cost prediction model for prediction; Through training of a PPO iterative model, a Pareto optimal solution set is generated, the PPO optimization including at least a policy network for outputting an action probability distribution ; a value network for evaluating a state value ; The importance sampling ratio is: ； The target function is clipped: ； wherein, represents the advantage function, represents the immediate reward, represents the discount factor, represents the clipping threshold, represents the action probability distribution under the old policy, represents the action, represents the state, represents the expectation, represents the current policy, represents the old policy; The Pareto dominance relationship is defined, the diversity solution set is generated by adjusting the weight, the non-dominated solution is screened to form the Pareto front, and the minimum cost path and the risk plan library are obtained; Based on the minimum cost path and the risk plan library, the materials and manpower in the construction project are uniformly scheduled and managed.

6. The big data based construction cost analysis method of claim 5, wherein, The multi-modal construction data in the acquisition system is sequentially subjected to data preprocessing and standardization to obtain initial multi-modal construction data, including: Access the enterprise ERP system, BIM model and public database through the API interface to obtain material inventory data, bill of quantities data and building material price index; Obtain the construction log image in the OCR scanner and extract the text data of the construction log image; Use the temperature and humidity sensor and the vibration sensor to collect the site environment data and manpower data in the construction site; The material inventory data, bill of quantities data, building material price index, construction log image, text data, site environment data and manpower data are subjected to outlier deletion, missing value supplementation and standardization processing to obtain initial multi-modal construction data.

7. The big data based construction cost analysis method of claim 5, wherein, The text and numerical value in the initial multi-modal construction data are aligned, and the image and space-time are aligned, and the aligned text and numerical value data and image and space-time data are constructed into a knowledge graph to obtain feature multi-modal construction data, including: The text name data in the initial multi-modal construction data is extracted using the NLP natural language processing model, including at least material name, construction period node and risk event; the text name data and numerical value label are associated to establish a mapping relationship to obtain first associated data; The image region in the initial multi-modal construction data is located based on the YOLO target detection model, and the quality defect label is labeled to obtain the text data in the quality defect label to obtain image data; The date and location information in the image data are matched with the 3D coordinates of the BIM model, the sensor data are fused with the environment state in the image to generate spatio-temporal joint features, and second associated data are obtained; The static rule base and dynamic rule base of the first associated data and second associated data are established, and the Neo4j graph database is used to store heterogeneous data association, the multi-hop relationship in the knowledge graph is encoded into a feature vector to obtain feature multi-modal construction data.

Images