Engineering cost information service platform

Abstract

The invention relates to the technical field of computers, in particular to a project cost information service platform which comprises a multi-source data acquisition subsystem, a big data storage and cleaning platform, an AI dynamic prediction engine, a BIM analysis module, a cost calculation and analysis module and a user service interface. Through the dual-mode design of the crawler unit and the API interface unit of the multi-source data acquisition subsystem, the data acquisition frequency is improved to once acquisition every 15 minutes, and the time difference between the cost data and the actual price of the market is ensured not to exceed 30 minutes. Meanwhile, the big data storage and cleaning platform integrates multi-source data such as real-time market offers, government guidance prices, historical contracts and macroeconomic indexes through standardized processing and association mapping, data islands are broken, a cost data set covering full dimensions is constructed, and the precision of cost management and control is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to an engineering cost information service platform. Background Technology

[0002] As a vital pillar of the national economy, the construction industry is expanding in scale and its project complexity is significantly increasing. Throughout the entire lifecycle of a construction project, cost management is a crucial link in ensuring economic benefits and controlling investment risks. It involves not only multiple stages such as budget preparation, cost control, and settlement review, but also the dynamic tracking and analysis of multi-dimensional information including material price fluctuations, changes in labor costs, and adjustments to policies and regulations.

[0003] Traditional engineering cost information service systems typically update data quarterly or annually. However, the construction market experiences frequent fluctuations in material prices and labor costs due to factors such as market supply and demand, raw material price fluctuations, and policy adjustments. This results in a significant disconnect between the data obtained by users and the actual market, leading to low accuracy in budget preparation and a substantial increase in the risk of project cost overruns. Therefore, it is necessary to design an engineering cost information service platform. Summary of the Invention

[0004] In order to overcome the shortcomings of the existing technology, the purpose of this invention is to provide an engineering cost information service platform.

[0005] The technical solution adopted in this invention is as follows: an engineering cost information service platform, comprising a multi-source data acquisition subsystem, a big data storage and cleaning platform, an AI dynamic prediction engine, a BIM parsing module, a cost calculation and analysis module, and a user service interface; the output end of the multi-source data acquisition subsystem is connected to the input end of the big data storage and cleaning platform, the output end of the big data storage and cleaning platform is connected to the input ends of the AI ​​dynamic prediction engine, the BIM parsing module, and the cost calculation and analysis module, respectively, the output ends of the AI ​​dynamic prediction engine and the BIM parsing module are both connected to the input end of the cost calculation and analysis module, and the output end of the cost calculation and analysis module is connected to the user service interface; The multi-source data acquisition subsystem is used to collect multi-source heterogeneous valence correlation data in real time, providing raw data support for the platform; The big data storage and cleaning platform is used to store, clean and integrate the collected data, build a unified cost database, and provide standardized data support. The AI ​​dynamic prediction engine is used to dynamically predict future cost trends based on data from a unified cost database and external influencing factors. The BIM parsing module is used to parse the BIM model, automatically extract the bill of quantities, and provide basic data for automatic pricing. The cost calculation and analysis module is used to realize automated cost calculation, cost accounting, risk analysis and multi-dimensional comparison; The user service interface is used to receive user requests, output service results, and support integration with external systems.

[0006] As a further description of the above technical solution: The multi-source data acquisition subsystem includes a crawler unit, an API interface unit, and a data preprocessing unit. The crawler unit is built on the Python Scrapy framework to create a distributed crawler cluster, which crawls real-time price quotes and logistics cost data from building materials e-commerce platforms and logistics platforms every 15 minutes. The API interface unit connects to government e-commerce platforms, industry association data centers, enterprise databases, and authoritative statistical agencies through RESTful APIs, synchronizing government guidance prices, historical contract data, macroeconomic indices, and policy documents. The data preprocessing unit is used to perform format conversion, field completion, and data fingerprint generation on the raw data.

[0007] As a further description of the above technical solution: The heterogeneous pricing-related data collected by the multi-source data acquisition subsystem includes real-time market price data, government-guided prices, historical contract data, logistics cost data, macroeconomic indices, policy documents, and BIM model files.

[0008] As a further description of the above technical solution: The big data storage and cleaning platform adopts a Hadoop distributed architecture, including a distributed storage module and a data cleaning module. The distributed storage module uses HDFS to classify and store structured, semi-structured, and unstructured data. The data cleaning module processes data according to the processes of deduplication, outlier filtering, missing value filling, standardization, and association mapping to build a high-dimensional unified cost database.

[0009] As a further description of the above technical solution: The AI ​​dynamic prediction engine adopts a fusion architecture of ARIMA model and LSTM neural network model, extracts historical price series, seasonal features, macroeconomic features and policy features as input, supports cost trend prediction for the next one month, three months, six months and custom time windows, and outputs predicted price series, confidence interval, price fluctuation risk level and trend description.

[0010] As a further description of the above technical solution: The BIM parsing module supports uploading BIM model files in IFC, Revit, and Navisworks formats. It parses component parameters using geometric topology analysis algorithms and natural language processing technology, automatically calculates quantities and generates standardized bills of quantities according to the construction project quantity list pricing specifications and project area pricing rules, and supports matching and updating of pricing rule libraries for various regions and specialties across the country.

[0011] As a further description of the above technical solution: The cost calculation and analysis module includes automatic pricing, cost accounting, risk analysis, and multi-dimensional comparison functions. The automatic pricing function matches real-time price data, historical price data, and predicted price data with the bill of quantities or user-input quantities, and calculates the total cost according to pricing rules. The cost accounting function breaks down the cost components and identifies core cost control points. The risk analysis function calculates the range of cost changes and high-risk items based on the prediction results. The multi-dimensional comparison function supports cost difference analysis across regions, time periods, and multiple projects, and identifies the causes.

[0012] As a further description of the above technical solution: The user service interface includes a core API gateway, a user interface, and a report generation and data visualization unit. The core API gateway implements request routing, permission verification, load balancing, and external system integration. The user interface supports full operation on the web and lightweight query and alert notification on mobile devices. The report generation unit supports the generation of structured reports in PDF, Word, and Excel formats. The data visualization unit generates intuitive charts such as line charts, pie charts, and bar charts, and supports interaction and export.

[0013] The present invention has the following beneficial effects: 1. This invention utilizes a dual-mode design of the crawler unit and API interface unit within the multi-source data acquisition subsystem, increasing the data acquisition frequency to once every 15 minutes, ensuring that the time difference between cost data and actual market prices does not exceed 30 minutes. Simultaneously, the big data storage and cleaning platform integrates multi-source data such as real-time market quotations, government-guided prices, historical contracts, and macroeconomic indices through standardized processing and correlation mapping, breaking down data silos and constructing a comprehensive cost dataset, significantly improving the accuracy of cost control.

[0014] 2. This invention's AI dynamic prediction engine integrates the advantages of ARIMA and LSTM models, combining multi-dimensional feature inputs to improve prediction accuracy. Simultaneously, the platform supports customizable prediction time windows, enabling forward-looking cost trend predictions for long-term projects and helping project owners formulate contingency strategies in advance. For example, if steel prices are predicted to rise, long-term supply contracts can be secured with suppliers in advance; if labor costs are predicted to fluctuate less, construction schedules can be optimized. This function fills the gap in the predictive capabilities of traditional systems, effectively mitigating cost risks caused by price fluctuations and providing a scientific basis for project investment decisions.

[0015] 3. The BIM parsing module of this invention supports direct uploading of mainstream BIM model files, automating component parsing, quantity calculation, and bill of quantities generation. Combined with the automatic pricing function of the cost calculation and analysis module, it shortens the cost preparation process and improves work efficiency. Simultaneously, the automated process completely eliminates omissions and miscalculations caused by manual quantity extraction and pricing, reducing the error rate of the bill of quantities and significantly improving the accuracy and reliability of cost estimates.

[0016] 4. The cost calculation and analysis module of this invention provides cross-regional, cross-time, multi-project, and multi-material comparative analysis functions, which can help project owners accurately pinpoint the reasons for cost differences: for example, by comparing regions, it can be found that the cost of purchasing a certain material in another location is lower, allowing for optimization of the supply chain configuration; by breaking down the costs of individual projects, it can identify links where labor costs account for too high a proportion, and construction efficiency can be improved through technological upgrades. This feature meets the needs of modern construction engineering for refined management and helps enterprises achieve their goals of cost reduction and efficiency improvement. Attached Figure Description

[0017] Figure 1 This is an organizational chart of the platform of this invention; Figure 2 This is a flowchart of the method of the present invention. Detailed Implementation

[0018] Reference Figure 1 This invention provides an engineering cost information service platform, comprising a multi-source data acquisition subsystem, a big data storage and cleaning platform, an AI dynamic prediction engine, a BIM parsing module, a cost calculation and analysis module, and a user service interface; the modules are linked through standardized data interfaces and communication protocols, with the specific connection relationships as follows: The output of the multi-source data acquisition subsystem is connected to the input of the big data storage and cleaning platform through an encrypted data channel, which is used to transmit the raw acquired data to the core processing layer. The output of the big data storage and cleaning platform is connected to the input of the AI ​​dynamic prediction engine, the input of the BIM parsing module, and the input of the cost calculation and analysis module, respectively, providing standardized and high-quality data support for each core functional module. The output of the AI ​​dynamic prediction engine and the output of the BIM parsing module are both connected to the input of the cost calculation and analysis module, providing predicted price data and basic engineering quantity data. The output of the cost calculation and analysis module is connected to the user service interface to provide the final cost results and analysis report to the user. The user service interface establishes connections with the multi-source data acquisition subsystem, the AI ​​dynamic prediction engine, and the BIM parsing module, supporting users to initiate custom requests and parameter configurations.

[0019] The specific functions and technical implementations of each module are as follows: Multi-source data acquisition subsystem: Its core function is to collect multi-source heterogeneous valence-related data in real time. It adopts a dual-mode design of crawler unit and API interface unit to ensure the breadth of data coverage and the timeliness of acquisition. The web crawler unit is built on the Python Scrapy framework to create a distributed web crawler cluster. It is configured with a dynamic IP pool and anti-crawling mechanism to monitor major domestic building materials e-commerce platforms (such as JD.com and Suning.com), official websites of large suppliers, and logistics service platforms (such as SF Express and Deppon Logistics) in real time. It crawls real-time price quotes, inventory status, and delivery cost data of key building materials such as steel, cement, concrete, and waterproofing materials. The collection frequency is set to once every 15 minutes to ensure the real-time nature of price data. API Interface Unit: Through standardized RESTful API interfaces, it connects with government cost management departments' e-government platforms, industry association data centers, and enterprise historical contract databases to synchronize government guidance prices, quota standards, historical engineering contract data (after anonymization), and settlement data daily. Simultaneously, it accesses API interfaces from authoritative institutions such as the National Bureau of Statistics and the People's Bank of China to collect macroeconomic indices such as CPI (Consumer Price Index), PPI (Producer Price Index), exchange rates, and interest rates in real time, as well as relevant policy documents on environmental protection and taxation (stored in text format). Data preprocessing unit: Performs preliminary format conversion on the collected raw data (such as unifying the price unit of different platforms to yuan / ton and yuan / square meter), completes fields (such as adding default identifiers for materials with missing specifications and models), and generates data fingerprints through MD5 encryption, laying the foundation for subsequent deduplication processing.

[0020] Big Data Storage and Cleaning Platform: Utilizing a Hadoop distributed architecture, it constructs an integrated data hub encompassing storage, processing, and scheduling. Specific functions include: Distributed storage module: Based on HDFS (Hadoop Distributed File System), data is classified and stored. Structured data (such as price values ​​and contract amounts) is stored in a MySQL cluster, semi-structured data (such as contract texts and policy clauses) is stored in MongoDB, and unstructured data (such as BIM model files and report attachments) is stored in a distributed file server. Load balancing and failover of data nodes are achieved through ZooKeeper. The data cleaning module consists of four steps: First, deduplication, which removes duplicate data based on data fingerprint comparison; second, outlier filtering, which uses box plots to identify and remove abnormal prices that deviate from the reasonable market range (such as data that are 3 times higher or 50% lower than the average price); third, missing value imputation, which uses the KNN algorithm to fill in missing price information based on data of similar materials and regions; and fourth, standardization, which converts data from different sources according to preset field specifications (such as material codes, regional classifications, and time formats) to generate a standardized data dictionary. Data association mapping: Based on key fields such as material code, project area, and time dimension, establish the association between real-time quotations, historical prices, macroeconomic indices, and policy data to build a high-dimensional cost dataset, supporting multi-dimensional data linkage query and analysis.

[0021] AI Dynamic Prediction Engine: Based on a time series prediction model, it constructs a closed-loop mechanism for training, prediction, and optimization to achieve dynamic prediction of cost trends. Model architecture: A fusion architecture of ARIMA model and LSTM neural network model is adopted, in which ARIMA model is used to capture the linear trend and periodic fluctuations of price data, and LSTM model is used to learn non-linear features (such as the indirect impact of macroeconomic policies on prices). Feature engineering: Extract historical price series (daily price data for the past 3 years), seasonal features (such as fluctuations in labor costs before and after the Spring Festival, and changes in building material transportation costs during the rainy season), macroeconomic features (CPI and PPI growth rates, exchange rate fluctuation rates), and policy features (such as environmental policy implementation indicators and tax adjustment coefficients) as input features for the model; Model training and updates: Based on historical datasets from a big data storage and cleaning platform, the model is trained regularly using a sliding window method (window size of 90 days), and the model parameters are updated every 7 days to ensure that the model adapts to market changes. At the same time, model evaluation metrics (MAE mean absolute error and RMSE root mean square error) are introduced to monitor prediction accuracy in real time. When the error exceeds a preset threshold (MAE>5%), retraining is automatically triggered. Forecasting function: Supports user-defined forecasting time windows, including preset forecasts for the next month, three months, and six months, as well as any time period set by the user (maximum not exceeding one year). Outputs the average forecast price, confidence interval (default 95% confidence level), price volatility risk level (low, medium, high), and trend description.

[0022] BIM Analysis Module: Focuses on the seamless integration of BIM models and cost calculations, enabling automated extraction of bill of quantities and generation of basic pricing data. File compatibility and verification: Supports uploading mainstream BIM model file formats such as IFC, Revit, and Navisworks. It has a built-in file format verification engine that automatically identifies file integrity (such as whether component attribute data is missing), prompts for damaged or incompatible files, and provides a format conversion tool. Component analysis and classification: Using geometric topology analysis algorithms and natural language processing technology, the system automatically identifies the component types (beams, columns, slabs, walls, foundations, etc.), geometric parameters (length, width, height, volume, area), and material information (such as steel bar type and concrete strength grade) in the BIM model, and classifies and codes the components according to the construction engineering quantity list pricing specifications. Automatic quantity calculation: Based on the parsed component parameters and preset quantity calculation rules (such as concrete components calculated by volume and walls calculated by area), the quantity statistics of each sub-item project are automatically completed, and a standardized quantity list is generated. The list includes core fields such as component code, name, specifications, unit of measurement, quantity, and material information. Pricing rule matching: Built-in pricing rule library for various regions and professions across the country (such as construction engineering), allowing users to select the corresponding pricing rule based on the project location and professional type, providing a basis for subsequent automatic pricing.

[0023] Cost Calculation and Analysis Module: As the core computing unit of the platform, it integrates multi-source data and core functions to achieve cost calculation, risk analysis, and multi-dimensional comparison. Automatic pricing function: Receives the bill of quantities output by the BIM parsing module, or the bill of quantities manually entered by the user, and automatically matches the real-time price data, historical price data, and predicted price data with the bill of quantities or the bill of quantities entered by the user. It calculates the direct engineering costs, measures costs, regulatory fees, taxes and other costs according to the selected pricing rules, and generates a complete cost result. Cost accounting function: Based on the cost results, automatically break down the cost composition of each sub-item of the project (such as the proportion of material cost, labor cost, and machinery usage fee) and identify cost control points; The risk analysis function combines AI-predicted price fluctuation trends to analyze the risk of project cost overruns, calculate risk exposure (such as the amount of cost increase that may result from a rise in steel prices in the next 3 months), and generate risk avoidance suggestions. Multi-dimensional comparison function: Supports cross-regional comparison (such as the cost difference of the same component between Beijing and Shanghai), cross-time comparison (such as the cost change of the same type of project in 2023 and 2024), multi-project comparison (such as the cost index comparison of different projects of the same company), and material price comparison (such as the comparison of quotations from different suppliers with the market average price). The algorithm calculates the difference rate and locates the reasons for the difference.

[0024] User service interface: Serving as the interaction hub between the platform and users, it provides diverse result outputs and request initiation methods. Core API Gateway: Built with Spring Cloud Gateway, it implements request routing, load balancing, access control, and interface rate limiting. It supports external systems (such as enterprise ERP systems and project management systems) to call the platform's core services through API interfaces, which support multiple data formats such as JSON and XML. User interface: Includes a web-based management platform and a mobile app. The web-based platform provides complete functions such as model uploading, parameter configuration, report generation, and data visualization, while the mobile app supports lightweight functions such as cost result query, prediction trend push, and risk warning reminders, adapting to different usage scenarios. Report generation and data visualization unit: Built-in multiple standardized report templates (such as budget preparation reports, cost analysis reports, and risk assessment reports), supporting user-defined report content and format; through visualization tools such as ECharts and Highcharts, data such as cost results, price trends, cost composition, and comparative analysis are transformed into intuitive charts such as line charts, pie charts, bar charts, and heat maps, making it easy for users to quickly obtain core information.

[0025] Reference Figure 2 The above is a schematic diagram of the method flow for an engineering cost information service platform provided by the present invention, specifically as follows: Step S1: Receive user request Users can initiate requests in two ways: manually through the visual interface of the web-based management platform or mobile app, or through external systems (such as enterprise ERP) calling the core API gateway. Request types include three categories: ① Cost query and analysis requests without a BIM model (e.g., querying steel costs in a specific region over a specific time period, or analyzing cost differences between two projects); ② Cost preparation and pricing requests with a BIM model (e.g., uploading a BIM model to generate a cost report); ③ Cost trend prediction requests (which can be initiated independently or in combination with the first two types of requests).

[0026] When initiating a request, users need to submit the following parameters: Requests without a BIM model need to fill in parameters such as project area, query time period, material type, project type, and whether forecasting is required; Requests with a BIM model need to upload the BIM model file (supports breakpoint resume, and a single file can be up to 10GB), and supplement parameters such as project area, pricing rule version, and forecasting time window (optional); For separate forecasting requests, users need to fill in material type, forecast start time, forecast duration, and influencing factor weights (optional, such as users can customize the influence weight of macroeconomic factors).

[0027] After receiving a request, the core API gateway of the user service interface first performs permission verification (verifying user account and project permissions) and request validity verification (verifying parameter integrity and file format validity). After the verification is successful, a unique request ID is generated (in the format of REQ, date, and random string), and the request information (including request ID, user parameters, and file storage path) is stored in the task queue, waiting for the core processing layer module to call it.

[0028] Step S2: Determine the request type: Does it involve a BIM model? The core API gateway determines the request type based on the BIM model file identifier in the request information. This determination logic is executed by the gateway's built-in routing rule engine. If the request information contains a valid BIM model file storage path (i.e., the user has uploaded a BIM model), it is determined to be a request involving a BIM model. The routing rule engine pushes the request ID and related parameters (project area, pricing rule version, file path) to the BIM parsing module and triggers the module to start the parsing process. If the request information does not contain a BIM model file identifier, or the file verification fails (e.g., format incompatibility, file corruption), it is determined that the request does not involve a BIM model. The routing rule engine will push the request ID and query parameters (region, time period, material type, etc.) to the cost calculation and analysis module, which will then extract the key query conditions.

[0029] The core API gateway will provide real-time feedback on the judgment results to the user interface, and users can view the request processing status (such as the request has been received and is parsing the BIM model, or the request has been received and is extracting query parameters).

[0030] Step S3: Extract the bill of quantities using the BIM parsing module. After receiving the request ID and related parameters pushed by the core API gateway, the BIM parsing module operates according to the following process: File Download and Verification: Download the corresponding BIM model file from the distributed file server, initiate file integrity verification and format adaptation. If the file is not in IFC format (such as Revit format), automatically call the format conversion tool to convert it to the IFC standard format. After the conversion is completed, verify the file integrity again. Component analysis and classification: The geometric topology analysis algorithm is used to traverse all components of the BIM model, extract the geometric parameters (length, width, height, volume, etc.) and attribute information (material, specifications, construction process) of each component, and match the component classification and code in the bill of quantities pricing specification of construction projects through natural language processing technology to generate a component information table (including fields such as component ID, classification code, name, specifications, material, geometric parameters, etc.). Quantity Calculation: Based on the pricing rule version corresponding to the project area, the built-in quantity calculation rule library is called to calculate the quantity of each component according to the rules (e.g., concrete beams are calculated by volume = length × width × height, and walls are calculated by area = length × height - area of ​​door and window openings). The quantities of each sub-item project (e.g., foundation and substructure engineering, main structure engineering, decoration and renovation engineering) are automatically summarized to generate a standardized bill of quantities. The bill of quantities includes fields such as sub-item project code, name, component details, unit of measurement, quantity, and comprehensive unit price composition (currently empty, to be filled by pricing). Results storage and feedback: The component information table, bill of quantities and request ID are associated and stored in the structured database of the big data storage and cleaning platform. A status notification of BIM parsing completion is returned to the core API gateway. At the same time, the preview data of the bill of quantities is pushed to the user interface for the user to view and confirm (the user can manually modify the quantities, and the system records the change log after modification).

[0031] Step S4: Extract query parameters After receiving the request ID and original query parameters pushed by the core API gateway, the cost calculation and analysis module initiates the parameter extraction and standardization process: Parameter extraction: Extract core query conditions from the original parameters, including project area (accurate to the prefecture-level city), query time period (start date - end date, accurate to the day), material type, project type, comparison object (if it is an analysis request, parameters of two or more comparison items need to be extracted), and whether prediction is needed (Boolean value). Parameter standardization: The extracted parameters are uniformly converted according to the platform's data dictionary. For example, Beijing and Beijing Municipality are uniformly standardized to Beijing Municipality, 3 months and 90 days are uniformly standardized to 90 days, and material names are matched according to national standard codes (e.g., rebar is matched to steel-HRB400E). Parameter validation and supplementation: Validate the reasonableness of parameters (e.g., the query period cannot exceed 3 years, and the prediction duration cannot exceed 1 year). If there are unreasonable parameters, feedback will be given to the user for modification through the core API gateway. If parameters are missing (e.g., material specifications are not filled in), default values ​​will be automatically filled in (e.g., fill in according to the mainstream specifications of the material in this region), and the usage of default values ​​will be noted in subsequent reports. Results storage: The standardized query parameters are associated with the request ID and stored in the parameter database of the big data storage and cleaning platform to provide a basis for subsequent data acquisition and calculation.

[0032] Step S5: Obtain real-time market data and historical prices from the big data storage and cleaning platform. Whether it's a BIM-related request following step S3 or a non-BIM request following step S4, all requests proceed to this step, where the cost calculation and analysis module initiates a data acquisition request: Data Request Generation: Based on the request ID, the cost calculation and analysis module retrieves the corresponding bill of quantities (BIM request) or standardized query parameters (non-BIM request) from the big data storage and cleaning platform, and generates data query conditions: The query conditions for BIM requests are project area + component or material type + current time (to obtain real-time data) and project area + component or material type + the past 3 years (to obtain historical data); the query conditions for non-BIM requests are project area + material or engineering type + query time period (to obtain real-time data and historical data within that time period). Data Retrieval and Filtering: The cost calculation and analysis module initiates query requests to the big data storage and cleaning platform through a data interface. The platform filters data from the unified cost database based on the query conditions: it prioritizes retrieving real-time data (collected within 30 minutes of the current time). If there is no real-time data for a certain type of material (e.g., a niche material supplier has not updated its price), it retrieves the most recent historical data and marks the data type as historical completion. At the same time, it retrieves the corresponding macroeconomic indices (e.g., CPI and PPI data within the query period) and policy data (e.g., whether there have been adjustments to environmental protection policies) as auxiliary data for subsequent analysis and forecasting. Data Validation and Integration: After receiving the data, the cost calculation and analysis module verifies the completeness (e.g., whether it covers all component or material types) and rationality (e.g., whether the price is within a reasonable range). Missing data is supplemented using the KNN algorithm, and abnormal data (e.g., quotations exceeding the reasonable range) is marked and removed. Subsequently, real-time data and historical data are integrated according to time, material, or component dimensions to generate a structured data matrix (row dimensions are time and material, column dimensions are price, inventory, macroeconomic indicators, etc.), providing data support for subsequent calculations and forecasts.

[0033] Step S6: Request future price prediction? The cost calculation and analysis module determines whether a forecast is needed based on the "whether a forecast is needed" flag in the request parameters: If marked as yes (including standalone prediction requests, cost inquiries and compilation requests with prediction functions), the AI ​​dynamic prediction engine is triggered to start the prediction process, and the data matrix integrated by S5, prediction time window, and influencing factor weights (user-defined or default values) are pushed to the AI ​​dynamic prediction engine. If the indicator is negative, the prediction process is skipped and the process proceeds directly to step S9 (cost calculation and analysis). At this point, only the real-time and historical data obtained in step S5 are used for calculation.

[0034] The judgment results will be synchronized to the user interface, where users can view status prompts indicating whether price prediction is in progress or no prediction is needed, and the cost will be calculated directly.

[0035] Step S7: Use the AI ​​prediction engine to generate future price trends After receiving relevant data from the cost calculation and analysis module, the AI ​​dynamic prediction engine performs predictions according to the following process: Feature extraction and preprocessing: Extract the features needed for prediction from the data matrix, including historical price series (daily prices in the past 3 years), seasonal features (generating dummy variables by month and quarter), macroeconomic features (CPI growth rate, PPI growth rate, exchange rate change rate), and policy features (such as marking environmental protection policies as 1 if implemented, otherwise as 0); standardize the extracted features (such as normalizing price data to the [0,1] interval) to eliminate the influence of units; Model selection and parameter configuration: The model is selected according to the type of material to be predicted and the time window: short-term prediction (within 1 month) mainly uses the ARIMA model (good at capturing linear trends), long-term prediction (1-6 months) mainly uses the LSTM model (good at learning nonlinear features), and ultra-long-term prediction (6-12 months) uses a fusion model of ARIMA and LSTM; the model parameters are optimized using the grid search method (such as the p, d, and q parameters of ARIMA, and the number of hidden layer neurons and learning rate of LSTM). Predictive calculation: Input the preprocessed features into the selected model and perform predictive calculation: short-term prediction uses the rolling prediction method (re-predicting every 7 days and correcting the prediction results), and long-term prediction uses the multi-step prediction method; during the calculation process, the model loss function (such as MSE mean squared error) is monitored in real time. If the loss value exceeds the preset threshold, the backup model is automatically switched (such as switching from LSTM to Transformer model). Forecast Result Calibration and Output: After the forecast is completed, the forecast result is calibrated by combining the real-time data obtained by S5 (such as correcting the forecast starting point data with the latest real-time price), generating the forecast price series (daily or weekly price), 95% confidence interval, price fluctuation risk level (calculated based on the standard deviation of the forecast price, standard deviation <5% is low risk, 5%-10% is medium risk, and >10% is high risk) and trend description (such as cement prices will rise by 8%-12% in the next 3 months due to the impact of environmental protection policies). Results Feedback: The AI ​​dynamic prediction engine associates the prediction results with the request ID, stores them in the prediction result library of the big data storage and cleaning platform, and returns a prediction completion notification to the cost calculation and analysis module. At the same time, it pushes a preview chart of the prediction trend (such as a line chart) to the user interface.

[0036] Step S8: Price Data Fusion If the user request includes a prediction function (determined as yes by S6), then proceed to this step, where the cost calculation and analysis module completes the data fusion: Data alignment: Align the real-time data (current time point) and historical data (within the query time period) obtained by S5 with the predicted data (within the future time window) generated by S7 according to the time dimension to form a complete price series of history, real-time and future. Data calibration: The weighted average method is used to calibrate the data in overlapping time periods (if there is a difference between the real-time data and the prediction starting point data, the real-time data is used as the benchmark to fine-tune the prediction data); outliers in historical data have been removed, and the confidence intervals of the prediction data are marked for easy reference in subsequent calculations; Dataset generation: Based on material or component type, the merged price data is associated with the corresponding engineering quantity (BIM request), query conditions (non-BIM request), macroeconomic parameters, and policy data to generate the final pricing dataset. This dataset contains fields such as material or component code, time, price type (historical, real-time, forecast), price value, confidence level, and influencing factors, providing complete data support for cost calculation.

[0037] If the user request does not include a prediction function (as determined by S6), then the real-time data and historical data obtained by S5 will be used directly as the pricing dataset, and this step will be skipped.

[0038] Step S9: Perform cost calculation and analysis The cost calculation and analysis module calls the pricing dataset and, in conjunction with the bill of quantities (BIM request) or query parameters (non-BIM request), performs the following core operations: Automatic pricing calculation (applies to all requests): BIM Request: Match the price data (real-time, historical, and forecast) in the pricing dataset with the components in the bill of quantities one by one, calculate the comprehensive unit price of each component (including labor costs, material costs, machinery usage fees, management fees, and profits) according to the selected pricing rules, multiply it by the quantity of work to obtain the itemized project cost, summarize all itemized project costs, add the measures fee, regulatory fee, and tax to obtain the total project cost; Non-BIM Requests: Calculate the average, maximum, and minimum cost of specified materials or project types within the corresponding time period based on query parameters, or calculate the total cost based on the project quantities entered by the user (non-BIM requests can manually enter project quantities).

[0039] Cost accounting: Break down the total cost into its components, calculate the percentage of each cost item (materials, labor, machinery, management, fees, and taxes), and generate a cost breakdown table. Identify the top five cost items and mark them as core cost control points. Analyze the reasons for their high or low costs (such as material prices being higher than the market average, or deviations in the calculation of project quantities).

[0040] Risk Analysis: Based on the predicted price fluctuation trend, calculate the range of cost changes for the project within the predicted time window (e.g., the total cost may increase by 5%-8% in the next 3 months, corresponding to an amount of RMB 1.2 million-1.92 million). Identify high-risk materials and sub-projects (such as materials with a high predicted price fluctuation risk level) and generate risk avoidance suggestions (such as suggesting that steel be purchased in advance to lock in the current low price; or signing a price fluctuation agreement with the supplier).

[0041] Multi-dimensional comparative analysis: Cross-regional comparison: Calculate the cost difference rate of the same material or project in different regions and analyze the reasons for the difference (such as logistics costs, regional guidance price differences). Cross-time comparison: Compare the cost change trends over different time periods and explain the reasons for the changes in conjunction with macroeconomic data; Multi-project comparison: Compare the cost indicators of different projects (such as cost per unit area, unit consumption of core materials) to identify the key factors of cost control advantages and disadvantages.

[0042] Step S10: Generate a structured cost report and visualization charts After completing the calculation and analysis, the cost calculation and analysis module pushes the results to the report generation and data visualization unit of the user service interface to perform the following operations: Report generation: Select the corresponding standardized template based on the request type: A cost estimation request generates an engineering cost estimation report (including sections such as project overview, bill of quantities, cost summary table, cost composition analysis, and risk warnings); a query request generates a cost query report (including query parameters, price trends, cost results, and data sources); an analysis request generates a cost comparison analysis report (including an overview of the comparison object, difference data, analysis of the reasons for the difference, and optimization suggestions). It supports user-defined report content, allowing users to select whether to include modules such as detailed calculation processes, risk analysis, and visualization charts, and to choose the report format (PDF, Word, Excel). It also supports adding personalized content such as company logos and signature bars. The system automatically verifies the consistency of report data (such as whether the total cost and the sum of sub-item project costs are consistent), generates a report number (in the format of REP, Request ID, and Date), and stores it in a distributed file server.

[0043] Data visualization: Generate basic charts: The cost results are displayed using a pie chart to show the cost composition, the price trend is displayed using a line chart (historical, real-time, and forecast data are marked with different colors), and the comparative analysis is displayed using a bar chart to show the differences; Generate advanced charts: Risk analysis uses radar charts to show the risk level of each sub-project, and cost control uses funnel charts to show the cost breakdown process, supporting user interaction (such as clicking on the chart to view specific data, zooming in and out of the trend chart, and filtering the time range). Embed visual charts into reports and provide a separate chart viewing entry in the user interface, supporting chart export (PNG, SVG format).

[0044] Step S11: Return the result to the user via the API gateway. After the report generation and data visualization unit completes its operation, it sends a result-ready notification to the core API gateway, which then returns the result in the following manner: User interface return: The results notification is displayed on the web or mobile app. Users can view the full text of the report and visualization charts online. It supports online preview, report download and chart export. At the same time, the core results (such as total cost, predicted trend and risk level) are sent to the user's mobile device in the form of push message. API interface returns: structured data (JSON or XML format) to the external system that initiated the request, including request ID, report download address, core cost data, prediction result summary, risk warning, etc. The external system can directly call this data for subsequent business processing (such as entering it into the financial system or generating a project schedule). Results storage and traceability: The final results (report files, visualization charts, core data) are associated with the request ID, user information, and processing logs and stored in the results database of the big data storage and cleaning platform. This supports subsequent user queries and traceability (e.g., historical reports can be queried through the request ID within one year). It also provides data support for the model optimization of the AI ​​dynamic prediction engine (e.g., comparing actual cost data with predicted data for model parameter adjustment).

[0045] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An engineering cost information service platform, characterized in that, The system includes a multi-source data acquisition subsystem, a big data storage and cleaning platform, an AI dynamic prediction engine, a BIM parsing module, a cost calculation and analysis module, and a user service interface. The output of the multi-source data acquisition subsystem is connected to the input of the big data storage and cleaning platform. The output of the big data storage and cleaning platform is connected to the input of the AI ​​dynamic prediction engine, the BIM parsing module, and the cost calculation and analysis module, respectively. The outputs of the AI ​​dynamic prediction engine and the BIM parsing module are both connected to the input of the cost calculation and analysis module. The output of the cost calculation and analysis module is connected to the user service interface. The multi-source data acquisition subsystem is used to collect multi-source heterogeneous valence correlation data in real time, providing raw data support for the platform; The big data storage and cleaning platform is used to store, clean and integrate the collected data, build a unified cost database, and provide standardized data support. The AI ​​dynamic prediction engine is used to dynamically predict future cost trends based on data from a unified cost database and external influencing factors. The BIM parsing module is used to parse the BIM model, automatically extract the bill of quantities, and provide basic data for automatic pricing. The cost calculation and analysis module is used to realize automated cost calculation, cost accounting, risk analysis and multi-dimensional comparison; The user service interface is used to receive user requests, output service results, and support integration with external systems.

2. The engineering cost information service platform according to claim 1, characterized in that: The multi-source data acquisition subsystem includes a crawler unit, an API interface unit, and a data preprocessing unit. The crawler unit is built on the Python Scrapy framework to create a distributed crawler cluster, which crawls real-time price quotes and logistics cost data from building materials e-commerce platforms and logistics platforms every 15 minutes. The API interface unit connects to government e-commerce platforms, industry association data centers, enterprise databases, and authoritative statistical agencies through RESTful APIs, synchronizing government guidance prices, historical contract data, macroeconomic indices, and policy documents. The data preprocessing unit is used to perform format conversion, field completion, and data fingerprint generation on the raw data.

3. The engineering cost information service platform according to claim 1, characterized in that: The heterogeneous pricing-related data collected by the multi-source data acquisition subsystem includes real-time market price data, government-guided prices, historical contract data, logistics cost data, macroeconomic indices, policy documents, and BIM model files.

4. The engineering cost information service platform according to claim 1, characterized in that: The big data storage and cleaning platform adopts a Hadoop distributed architecture, including a distributed storage module and a data cleaning module. The distributed storage module uses HDFS to classify and store structured, semi-structured, and unstructured data. The data cleaning module processes data according to the processes of deduplication, outlier filtering, missing value filling, standardization, and association mapping to build a high-dimensional unified cost database.

5. The engineering cost information service platform according to claim 1, characterized in that: The AI ​​dynamic prediction engine adopts a fusion architecture of ARIMA model and LSTM neural network model, extracts historical price series, seasonal features, macroeconomic features and policy features as input, supports cost trend prediction for the next one month, three months, six months and custom time windows, and outputs predicted price series, confidence interval, price fluctuation risk level and trend description.

6. The engineering cost information service platform according to claim 1, characterized in that: The BIM parsing module supports uploading BIM model files in IFC, Revit, and Navisworks formats. It parses component parameters using geometric topology analysis algorithms and natural language processing technology, automatically calculates quantities and generates standardized bills of quantities according to the construction project quantity list pricing specifications and project area pricing rules, and supports matching and updating of pricing rule libraries for various regions and specialties across the country.

7. The engineering cost information service platform according to claim 1, characterized in that: The cost calculation and analysis module includes automatic pricing, cost accounting, risk analysis, and multi-dimensional comparison functions. The automatic pricing function matches real-time price data, historical price data, and predicted price data with the bill of quantities or user-input quantities, and calculates the total cost according to pricing rules. The cost accounting function breaks down the cost components and identifies core cost control points. The risk analysis function calculates the range of cost changes and high-risk items based on the prediction results. The multi-dimensional comparison function supports cost difference analysis across regions, time periods, and multiple projects, and identifies the causes.

8. The engineering cost information service platform according to claim 1, characterized in that: The user service interface includes a core API gateway, a user interface, a report generation and data visualization unit; the core API gateway implements request routing, permission verification, load balancing and external system integration. The user interface supports full operation on the web and lightweight query and alert notification on mobile devices; the report generation unit supports the generation of structured reports in PDF, Word, and Excel formats; the data visualization unit generates intuitive charts such as line charts, pie charts, and bar charts, and supports interaction and export.

Images