Building material price dynamic prediction system based on multi-dimensional data fusion

Abstract

The invention discloses a building material price dynamic prediction system based on multi-dimensional data fusion, and relates to the technical field of data processing and price prediction, and the system comprises a data integration module which is in butt joint with a multi-industry data interface, collects multi-dimensional data, verifies and summarizes the multi-dimensional data, a data preprocessing module reads original data, outputs a standardized sequence through abnormal value processing and the like, and stores the standardized sequence. The causal mining module adopts a time sequence algorithm to identify a variable relationship and construct a causal graph, the coefficient construction module calculates and optimizes an initial coefficient to form a conduction matrix, and the prediction module collects data in real time, calculates a predicted price and outputs the predicted price; according to the method, building and associated industry data are integrated, after the quality is improved through preprocessing, a variable relation is identified through a time sequence causal algorithm, and a causal graph is constructed; a causal conduction coefficient matrix is constructed through an optimization technology, a dynamic prediction algorithm and adaptive and correction factors are combined, accurate dynamic prediction of building material prices is realized, related subjects are assisted to pre-judge trends, configuration is optimized, and risks are reduced.

Description

Technical Field

[0001] This invention relates to the field of data processing and price prediction technology, specifically a dynamic prediction system for building material prices based on multi-dimensional data fusion. Background Technology

[0002] Building material prices are influenced by a complex interplay of factors, and their fluctuations directly impact construction cost control, project investment planning, and market resource allocation efficiency. With the increasing scale and specialization of the construction industry chain, building material prices are no longer solely determined by internal supply and demand but are also deeply intertwined with the operational status of related industries such as mining, chemicals, energy, and logistics. Data from various industries is stored across different platforms, encompassing indicators such as production, sales, energy consumption, and transportation. Dynamic changes in this data can have a cascading effect on building material prices through the supply chain. Simultaneously, the complexity and uncertainty of the market environment further exacerbate price volatility, making the industry's need for accurate and real-time building material price forecasts increasingly urgent. There is a pressing need to build a technological system capable of integrating multi-dimensional data and uncovering deep correlations to provide decision support for relevant stakeholders.

[0003] Traditional building material price forecasting techniques often rely on single-industry data or limited-dimensional indicators. This limitation in data sources makes it difficult for forecasting models to comprehensively capture all factors influencing prices. Some methods rely solely on historical price data for trend inference, ignoring causal relationships and transmission lags between variables. This can easily lead to the inclusion of spurious correlations in the analysis, resulting in significant biases in the forecasts. Furthermore, traditional techniques lack systematic data preprocessing, and issues such as outliers and missing values ​​reduce data quality and affect forecast accuracy. Simultaneously, most forecasting models use fixed algorithms, failing to dynamically adapt to market changes and supply chain transmission logic. They struggle to respond to real-time industry data fluctuations, resulting in insufficient timeliness in forecasts. This fails to meet the refined price prediction needs of enterprises and regulatory authorities, and thus cannot effectively support cost control and market regulation decisions. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of existing technologies and provide a multi-dimensional data fusion-based dynamic prediction system for building material prices. This system collects multi-dimensional data from the construction and related industries through a data integration module. After cleaning and transformation by a data preprocessing module, a causal mining module identifies causal relationships and time-lag parameters between variables, constructing a causal directed acyclic graph. A coefficient construction module optimizes the initial correlation coefficients, forming a causal transmission coefficient matrix. Finally, the prediction output module, based on the latest data changes and the causal transmission coefficient matrix, uses a dynamic price prediction fusion algorithm to output real-time predicted building material prices, providing accurate price prediction services to the industry.

[0005] To solve the above-mentioned technical problems, this invention provides the following technical solution: a multi-dimensional data fusion-based dynamic prediction system for building material prices, comprising: Data integration module: Connects to data interfaces of the construction industry and related industries, collects multi-dimensional raw data through preset methods, and aggregates it into a unified data pool after format standardization and compliance verification; Data preprocessing module: Reads raw data from the data pool, performs outlier removal, missing value filling and data standardization transformation, and outputs standardized data sequences after filtering redundant indicators; Causal mining module: Receives standardized data sequences, uses time-series causal strength algorithm to identify causal direction, time delay parameters and causal strength between variables, marks pseudo-associations and constructs causal directed acyclic graph; Coefficient construction module: Calculates the initial correlation coefficient between industry indicators and building material prices using the time-series weighted cosine similarity algorithm, outputs the results using a causal directed acyclic graph, optimizes the initial correlation coefficient using causal-industry coupling coefficient optimization technology, and constructs a causal transmission coefficient matrix; Prediction output module: Collects the latest industry data in real time and calculates the changes. Based on the causal transmission coefficient matrix, it calculates the predicted price of building materials through a dynamic price prediction fusion algorithm and outputs the prediction results.

[0006] Furthermore, the data integration module connects to the construction industry and related industries, including the mining industry, chemical industry, energy industry, and logistics equipment manufacturing industry. The specific data interfaces include official data platform interfaces, chemical industry association statistical platform interfaces, automobile industry association sales data interfaces, local building materials trading market database interfaces, and large building materials production enterprise ERP system interfaces. The preset data collection methods include OAuth2.0 authorized API interface retrieval, JDBC / ODBC protocol direct database connection extraction, and compliant Scrapy crawler crawling. The collected multi-dimensional raw data specifically includes: mining equipment maintenance plans, raw material extraction volume and inventory in the mining industry; admixture production capacity, capacity utilization rate, and ex-factory price in the chemical industry; peak-valley electricity pricing standards, coal market prices, and energy consumption data of building materials production enterprises in the energy industry; heavy truck production and sales data, transportation capacity supply and demand gaps, and transportation route freight rates in the logistics equipment manufacturing industry; and real-time building material quotations, historical transaction prices, and capacity utilization rate data in the construction industry.

[0007] Furthermore, in the data preprocessing module, when performing outlier removal, the mean μ and standard deviation σ of each indicator field are first calculated, and outliers are initially identified using the 3σ principle. Then, the box plot analysis method is used to calculate the quartiles Q1, Q2, Q3 and the interquartile range IQR. Data that are less than Q1-1.5IQR or greater than Q3+1.5IQR are marked as suspected outliers. Data that are identified as outliers by both methods are directly removed. Suspected outliers identified by only one method trigger a manual review process, and after review, it is determined whether to remove or retain them.

[0008] Furthermore, in the causal mining module, the mathematical expression for the temporal causal strength algorithm is: in, Industry benchmark Building material prices Causality strength with a time lag of τ days For linear-nonlinear fusion weights, for At any moment and At any moment Mutual information value, NLCC for At any moment and At any moment The nonlinear correlation coefficient, TDR is the time-delay decay coefficient. It is a time-delay reliability index used to measure the identified time-delay parameters. The credibility of Representing the current moment, It is a time delay parameter. It is the first Industry indicator variables, No. One variable is the price of building materials.

[0009] Furthermore, in the causal mining module, the specific steps for identifying the causal direction, time delay parameters, and causal strength between variables using the time-series causal strength algorithm are as follows: align the received standardized data sequence by timestamp and filter out non-redundant variables with a variance greater than 0.01; then calculate the industry indicator variables using the time-series causal strength algorithm. Building material price variables causal strength and right causal strength The direction with greater causal strength is taken as the true causal transmission direction between variables; the time lag values ​​from 0 to 30 days are traversed with a step size of 1 day. Calculate each corresponding ,Pick The maximum value corresponding to As the final time delay parameter, when there are two or more different time delay parameters Both increase the strength of causality If the values ​​reach the same maximum value, then the smallest value among them is selected. As the final time delay parameter, the time-series causality strength algorithm will be used to calculate the time delay parameter. It is directly used as the quantification result of the causal strength between variables, and its value range is [0, 1].

[0010] Furthermore, in the causal mining module, the specific steps for marking pseudo-associations are as follows: calculating industry indicators using a time-series causal strength algorithm. With building material prices The causal strength between them is determined by a preset causal strength threshold of 0.2. When the calculated causal strength is... When it is less than 0.2, a preliminary judgment is made. and This is a pseudo-association; further calculations are needed. and Synergy coefficient When Syn Furthermore, since there is no supply chain transmission logic to support a direct causal relationship between the two, their pseudo-association attribute is ultimately confirmed. For the confirmed pseudo-association, pseudo-association labels are marked in the causal relationship dataset, and the corresponding labels are recorded. Numerical value and synchronous variation coefficient Syn Numerical values, where the synchronous variation coefficient Syn The calculation formula is: in: for and The number of time points commonly covered by time-series data sequences For the first The importance weights of each feature dimension for In the Statistical values ​​under dimension , They are respectively In the Mean and standard deviation under dimensionality for In the Statistical values ​​under dimension , They are respectively In the Mean and standard deviation in the dimension.

[0011] Furthermore, the specific steps for constructing a causal directed acyclic graph in the causal mining module are as follows: Determine the nodes in the graph: Select the filtered, non-redundant industry indicator variables. and price variables of various types of building materials As nodes in the graph, each node is associated with attribute information, including variable name, industry, data type, and time series data sequence identifier; Drawing directed edges: Based on the true causal direction between variables identified by the temporal causal strength algorithm, directed edges are drawn from the cause variable node to the result variable node. Each directed edge is associated with attribute information, including the causal strength of the corresponding variable pair. Time delay parameters ; Hierarchical Node Layout: A hierarchical layout algorithm is used to arrange nodes, dividing them into three layers according to the upstream and downstream order of the industry chain, with upstream industry indicator variables... The node is located at the top layer, representing midstream industry indicator variables. The node is located in the middle layer, and the building material price variable... The node is located at the bottom layer; Hierarchical coloring of directed edges: based on causality strength Hierarchical coloring of directed edges. Use red at times, Orange was used at that time. Blue was used at that time; Eliminating spurious edges: This relates to causal strength. For pseudo-association variable pairs, no corresponding directed edges are drawn; instead, the pseudo-association variable name and corresponding... are recorded only in the attribute information of the node to which it belongs. Numerical value; Graph structure validation and optimization: Validate the graph for cyclic causal paths; if found, remove the cyclic path. Minimize the directed edges to eliminate cycles, ultimately generating a complete causal directed acyclic graph.

[0012] Furthermore, in the coefficient construction module, the mathematical expression for the time-weighted cosine similarity algorithm is: in, Industry benchmark With building material prices The initial correlation coefficient, for Time Industry Indicators Standardized data, for Construction material prices Standardized data, for Standardized data weighted mean, for Standardized data weighted mean, These are time-series trend weights preset based on a time decay function or business importance. To prevent the minimum value where the denominator is 0, It represents the total number of moments in the time series data.

[0013] Furthermore, the specific steps for constructing the causal transmission coefficient matrix in the coefficient construction module are as follows: All selected industry indicator variables are used as row variables of the matrix, with each row variable corresponding to a unique industry category identifier and indicator name; the price variables of each type of building material are used as column variables of the matrix, with each column variable corresponding to a unique building material type identifier and price indicator name; the optimized causal transmission coefficients corresponding to each industry indicator variable and building material price variable, obtained through causal-industry coupling coefficient optimization technology, are filled into the corresponding positions of the matrix one by one according to the correspondence between row variables and column variables; for the filled matrix, the sum of all causal transmission coefficients in each row is calculated. ;when If the coefficient is not within the range of 0.95 to 1.05, then multiply each causal transmission coefficient in that row by an adjustment factor. The adjustment coefficient This normalizes the sum of all coefficients in the row to 1. After standardization, the matrix is ​​checked for completeness to ensure there are no missing coefficients or logical contradictions. Finally, a causal transmission coefficient matrix containing the quantitative transmission relationship between all industry indicator variables and building material price variables is generated.

[0014] Furthermore, in the coefficient construction module, the specific steps for optimizing the initial correlation coefficient using the causal-industry coupling coefficient optimization technique are as follows: Based on the initial correlation coefficient obtained by the time-series weighted cosine similarity algorithm... The causal strength output by combining the causal directed acyclic graph. Time delay parameters Industrial chain influence coefficient and building material cost sensitivity coefficient Optimize; for The pseudo-association is determined by its corresponding initial association coefficient. Set directly to zero; for The true causal relationship, and its optimized causal transmission coefficient. The calculation formula is: in, This represents the total number of industry indicator variables; all the results will be... Organize the data according to industry indicator variables and building material price variables, and construct a causal transmission coefficient matrix.

[0015] Furthermore, in the prediction output module, the mathematical expression of the dynamic price prediction fusion algorithm is: TSF ;in, For building materials In the future Predicted price at any time For building materials At the present moment The benchmark price For the industry Building materials The causal transmission coefficient, For the industry At any moment Real-time data changes, TSF For time lag-prediction cycle adaptation factor, For the industry Indicators for building materials Price time lag parameter, This is a market volatility correction factor. For building materials Historical 30-day price volatility It is a prediction cycle. The total number of industry indicator variables.

[0016] Compared with existing technologies, this multi-dimensional data fusion-based dynamic prediction system for building material prices has the following advantages: I. This invention integrates multi-dimensional data from the construction industry and related industries to build a comprehensive data support system. It combines multiple data collection methods and undergoes rigorous compliance verification and format standardization to ensure the data sources are broad and reliable. Preprocessing processes such as outlier removal and missing value imputation improve data quality, laying a solid foundation for subsequent analysis. A time-series causal strength algorithm accurately identifies causal relationships, time lag parameters, and strengths between variables, effectively distinguishing between true and false associations. The constructed causal directed acyclic graph clearly presents the transmission paths between variables. Based on a time-series weighted cosine similarity algorithm and causal-industry coupling coefficient optimization technology, the correlation coefficient is optimized and a causal transmission coefficient matrix is ​​constructed. This makes the quantification of the correlation between industry indicators and building material prices more consistent with actual industry logic, significantly improving the scientific rigor and accuracy of price prediction and providing a reliable basis for relevant decision-making.

[0017] II. This invention achieves dynamic prediction of building material prices by collecting the latest industry data in real time and calculating changes, combined with a dynamic price prediction fusion algorithm. This enables timely capture of the impact of market changes on prices. The introduction of a time lag-prediction cycle adaptation factor and a market fluctuation correction coefficient further improves the prediction model, reducing prediction bias caused by time lag effects and market fluctuations. A hierarchical layout and graded coloring of the causal directed acyclic graph make the causal relationships between variables intuitive and easy to understand, facilitating users' comprehension of the price impact mechanism. The normalized causal transmission coefficient matrix ensures a reasonable allocation of the impact weights of various industry indicators on building material prices, making the prediction results more valuable and helping enterprises, market regulators, and other relevant entities to anticipate price trends, optimize resource allocation, and reduce market risks.

[0018] Other advantages, objectives and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination or study, or may be learned from the practice of the invention. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0020] Figure 1 A flowchart for a dynamic prediction system for building material prices based on multi-dimensional data fusion; Figure 2 This diagram illustrates the data transmission between modules in a multi-dimensional data fusion-based dynamic price prediction system for building materials. Detailed Implementation

[0021] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0022] Example 1: Implementation scenarios for short-term (7-day) dynamic forecasting of concrete prices.

[0023] The system's data integration module interfaces with relevant data sources in the construction, mining, chemical, energy, and logistics equipment manufacturing industries. Specifically, this includes interfaces with official data platforms, chemical industry association statistical platforms, automotive industry association sales data, local building materials trading market databases, and large building materials manufacturing enterprise ERP systems. This multi-channel integration ensures comprehensive capture of various data sources affecting concrete prices. Three preset data collection methods are employed: OAuth2.0 authorized API interface retrieval, direct database connection extraction via JDBC / ODBC protocol, and compliant Scrapy crawler crawling. These methods adapt to different data storage and provision methods, ensuring flexibility and comprehensiveness in data collection. The collected multi-dimensional raw data specifically includes: mining equipment maintenance plans, raw material extraction volumes and inventory levels in the mining industry; admixture production capacity, capacity utilization rate, and ex-factory prices in the chemical industry; peak-valley electricity pricing standards, coal market prices, and energy consumption data of building material manufacturers in the energy industry; heavy truck production and sales data, transportation capacity supply and demand gaps, and transportation route freight rates in the logistics equipment manufacturing industry; and real-time concrete quotations, historical transaction prices, and capacity utilization rate data in the construction industry. This data covers all relevant factors in the concrete production, transportation, and sales chain, providing rich data for subsequent forecasting. After collection, the data undergoes format standardization and compliance verification to eliminate processing obstacles caused by data format differences, ensuring that the data conforms to usage specifications. It is then aggregated into a unified data pool to achieve centralized data management and retrieval. Figure 1 As shown.

[0024] The data preprocessing module reads the raw data from the data pool and begins outlier removal. First, it calculates the mean and standard deviation of each indicator field, using the 3σ principle to initially identify outliers. Then, it combines box plot analysis to calculate the quartiles Q1, Q2, Q3, and the interquartile range (IQR). Data values ​​less than Q1 - 1.5IQR or greater than Q3 + 1.5IQR are marked as suspected outliers. This dual approach allows for more accurate filtering of data deviating from the normal range. Data identified as outliers by both methods is directly removed. Suspected outliers identified by only one method trigger a manual review process. After review, the decision to remove or retain the data is made, ensuring data accuracy while avoiding the accidental deletion of valid data. Next, missing value imputation is performed. Appropriate imputation methods are selected based on the characteristics of different indicators to fill information gaps caused by missing data and ensure data integrity. Finally, data standardization is performed to convert indicator data of different magnitudes and units into data under a unified standard, eliminating the influence of units and making various types of data comparable. Finally, redundant indicators are screened out, and indicators that are duplicated or contribute very little to the prediction results are removed to reduce the subsequent computational pressure, improve processing efficiency, and output standardized data sequences to provide a high-quality data foundation for causal mining.

[0025] The causal mining module receives standardized data sequences and uses a time-series causal strength algorithm to identify the causal direction, time delay parameters, and causal strength between variables. The mathematical expression for the time-series causal strength algorithm is: in, Industry benchmark Building material prices Causality strength with a time lag of τ days For linear-nonlinear fusion weights, for At any moment and At any moment Mutual information value, NLCC for At any moment and At any moment The nonlinear correlation coefficient, TDR is the time-delay decay coefficient. It is a time-delay reliability index used to measure the identified time-delay parameters. The credibility of Representing the current moment, It is a time delay parameter. It is the first Industry indicator variables, No. This algorithm effectively captures the correlation patterns between building material price variables over time. First, the standardized data series is aligned by timestamps, and non-redundant variables with variance greater than 0.01 are filtered to ensure that all variables involved in the analysis have actual significance and to avoid interference from invalid variables. The algorithm then calculates the causal strength of each industry indicator variable on the concrete price variable, and vice versa. The direction with the greater causal strength is taken as the true causal transmission direction between variables, accurately defining the influencing and affected relationships. Using a 1-day step, the algorithm iterates through time lag values ​​from 0 to 30 days, calculating the causal strength corresponding to each lag. The lag corresponding to the maximum causal strength is taken as the final lag parameter. If multiple lags correspond to the same maximum value, the smallest one is selected, clarifying the time delay effect of industry indicators on concrete prices.

[0026] The results calculated by the time-series causality strength algorithm are directly used as the quantitative result of the causal strength between variables, providing a quantitative basis for subsequent correlation coefficient optimization. Subsequently, pseudo-correlation relationships are marked. A preset causal strength judgment threshold of 0.2 is used; when the calculated causal strength is less than 0.2, it is initially judged as a pseudo-correlation relationship, and then the synchronization coefficient Syn is further calculated. Synchronous variation coefficient The calculation formula is: in: for and The number of time points commonly covered by time-series data sequences For the first The importance weights of each feature dimension for In the Statistical values ​​under dimension , They are respectively In the Mean and standard deviation under dimensionality for In the Statistical values ​​under dimension , They are respectively In the If the coefficient of synchronous variation of the mean and standard deviation under a given dimension is greater than the causal strength and there is no supply chain transmission logic to support a direct causal relationship between the two, then the pseudo-correlation attribute is finally confirmed, accurately excluding variable pairs without actual causal relationship and avoiding misleading prediction results. Confirmed pseudo-correlation relationships are labeled with pseudo-correlation tags in the causal relationship dataset, and relevant values ​​are recorded for subsequent traceability and verification.

[0027] Finally, a causal directed acyclic graph (DAG) is constructed, using the selected non-redundant industry indicator variables and concrete price variables as nodes. Each node is associated with attribute information such as variable name, industry, data type, and time-series data sequence identifier, making node information clear and traceable. Directed edges are drawn based on the identified true causal directions, with each edge associated with attribute information such as causal strength and time delay parameters, intuitively presenting the details of causal relationships between variables. A hierarchical layout algorithm is used to divide the nodes into three layers according to the upstream and downstream order of the industry chain: upstream industry indicator variable nodes are at the top layer, midstream industry indicator variable nodes are in the middle layer, and concrete price variable nodes are at the bottom layer. This layout, consistent with industry logic, facilitates understanding of the causal transmission path. Directed edges are colored according to causal strength: red for causal strength ≥ 0.6, orange for 0.4 ≤ causal strength < 0.6, and blue for 0.2 ≤ causal strength < 0.4, visually distinguishing the strength of causal relationships. For pseudo-correlated variable pairs with causal strength < 0.2, no corresponding directed edges are drawn; only the relevant content is recorded in the attribute information of their respective nodes, ensuring the simplicity and effectiveness of the graph structure. Verify whether there are cyclic causal paths in the graph. If so, remove the directed edge with the weakest causal strength in the cyclic path to eliminate the cycle, ensure the rationality of the graph structure, generate a complete causal directed acyclic graph, and provide clear causal relationship support for coefficient construction.

[0028] The coefficient construction module calculates the initial correlation coefficient between industry indicators and concrete prices using a time-weighted cosine similarity algorithm. The mathematical expression for the time-weighted cosine similarity algorithm is: in, Industry benchmark With building material prices The initial correlation coefficient, for Time Industry Indicators Standardized data, for Construction material prices Standardized data, for Standardized data weighted mean, for Standardized data weighted mean, These are time-series trend weights preset based on a time decay function or business importance. To prevent the minimum value where the denominator is 0, This represents the total number of time points in the time-series data. The algorithm fully considers the characteristics of time-series data and the weight differences of data at different time points, making the initial correlation coefficient calculation more consistent with reality. Then, combining the results output from the causal directed acyclic graph, the initial correlation coefficient is optimized using causal-industry coupling coefficient optimization technology, ensuring that the correlation coefficient reflects both causal relationships and aligns with actual industry logic. For pseudo-correlation relationships with a causal strength < 0.2, their corresponding initial correlation coefficients are directly set to zero to eliminate the interference of pseudo-correlation on the final coefficients. For true causal relationships with a causal strength ≥ 0.2, the optimized causal transmission coefficient is calculated using the corresponding formula, combining causal strength, time lag parameters, industry chain influence coefficient, and building material cost sensitivity coefficient. The calculation formula is: ,in, This represents the total number of industry indicator variables; all the results will be... By organizing industry indicator variables and building material price variables, a causal transmission coefficient matrix is ​​constructed, so that the coefficients can accurately reflect the transmission strength of industry indicators on concrete prices.

[0029] Subsequently, a causal transmission coefficient matrix was constructed. All selected industry indicator variables were used as row variables, each corresponding to a unique industry category identifier and indicator name. Concrete price was used as a column variable, corresponding to a unique building material type identifier and price indicator name, clearly defining the meaning of the variables represented by the matrix rows and columns. The optimized causal transmission coefficients were then filled into the corresponding positions in the matrix according to the correspondence between row and column variables, forming an intuitive coefficient correspondence. For the filled matrix, the sum of all causal transmission coefficients in each row was calculated. If the sum was not within the range of 0.95 to 1.05, each causal transmission coefficient in that row was multiplied by an adjustment factor to normalize the sum of all coefficients in that row to 1, ensuring the standardization and rationality of the coefficient matrix. After standardization, the matrix was checked for completeness, confirming no missing coefficients or logical contradictions. Finally, the causal transmission coefficient matrix was generated, providing accurate quantitative transmission basis for prediction calculations.

[0030] The prediction output module collects the latest data from various industries in real time, promptly capturing industry dynamics and calculating the real-time data changes of each industry indicator at the current moment, ensuring that the data input into the prediction model reflects the latest market situation. Based on the constructed causal transmission coefficient matrix, the predicted price of concrete for the next 7 days is calculated using a dynamic price prediction fusion algorithm. The mathematical expression of the dynamic price prediction fusion algorithm is: TSF ;in, For building materials In the future Predicted price at any time For building materials At the present moment The benchmark price For the industry Building materials The causal transmission coefficient, For the industry At any moment Real-time data changes, TSF For time lag-prediction cycle adaptation factor, For the industry Indicators for building materials Price time lag parameter, This is a market volatility correction factor. For building materials Historical 30-day price volatility It is a prediction cycle. The algorithm integrates multiple influencing factors to effectively fuse multi-dimensional data, considering the total number of industry indicator variables. During the calculation process, it takes into account parameters such as the current benchmark price of concrete, the causal transmission coefficient of each industry to concrete, the real-time data changes of each industry, the time lag-prediction cycle adaptation factor, the market volatility correction coefficient, and the historical 30-day price volatility of concrete. This comprehensive consideration of various influencing factors improves prediction accuracy. The final output is the predicted price of concrete for the next 7 days, providing timely and reliable reference for construction companies' procurement decisions and cost budgeting.

[0031] In summary, this embodiment focuses on short-term concrete price forecasting. A data integration module comprehensively collects multi-industry related data, forming a complete data foundation through format standardization and compliance verification. The data preprocessing stage employs dual outlier removal, missing value imputation, and standardization transformation to ensure data quality. A time-series causal strength algorithm is used to uncover the true causal relationships between variables, eliminating spurious associations and constructing a clear causal directed acyclic graph. Based on a time-series weighted cosine similarity algorithm and causal-industry coupling coefficient optimization technology, an accurate causal transmission coefficient matrix is ​​constructed. Finally, a dynamic price forecast fusion algorithm is used to calculate the predicted concrete price for the next 7 days, comprehensively considering multiple influencing factors, providing reliable support for enterprises' short-term procurement decisions.

[0032] Example 2: Implementation scenarios for medium-term (30-day) dynamic forecasting of steel bar prices.

[0033] The data integration module interfaces with data sources in the construction industry and related sectors such as mining, chemicals, energy, and logistics equipment manufacturing. It covers interfaces with official data platforms, chemical industry association statistical platforms, automotive industry association sales data, local building materials trading market databases, and ERP systems of large building materials manufacturers, broadly encompassing upstream and downstream data sources in the steel reinforcement industry chain to ensure comprehensive data collection. It utilizes three preset methods: OAuth2.0 authorized API interfaces, direct database connections via JDBC / ODBC protocols, and compliant Scrapy crawlers, flexibly adapting to different data acquisition needs and ensuring efficient and complete data collection. The collected multi-dimensional raw data specifically includes: iron ore mining equipment maintenance plans, iron ore mining volume, and inventory levels from the mining industry, which directly relate to the supply of raw materials for rebar production; steelmaking additive production capacity, capacity utilization rate, and ex-factory prices from the chemical industry, affecting auxiliary costs in rebar production; peak-valley electricity pricing standards, coal market prices, and energy consumption data from steel production enterprises from the energy industry, related to energy consumption costs in rebar production; heavy truck production and sales data, transportation capacity supply and demand gaps, and iron ore transportation route freight rates from the logistics equipment manufacturing industry, affecting transportation costs of raw materials and finished rebar; and real-time rebar quotations, historical transaction prices, and capacity utilization data from the construction industry, directly reflecting the supply and demand and price status of the rebar market. After collection, format standardization and compliance verification are performed to eliminate data format barriers, ensuring data legality, compliance, and usability. The data is then aggregated into a unified data pool for centralized storage and unified management, facilitating data retrieval for subsequent modules. Figure 2 As shown.

[0034] The data preprocessing module reads the raw data from the data pool and first performs outlier removal. It calculates the mean and standard deviation of each indicator field, uses the 3σ principle to initially identify outliers, and then calculates the quartiles Q1, Q2, Q3, and interquartile range (IQR) using box plot analysis to mark suspected outliers. This dual verification mechanism more accurately identifies outliers deviating from the normal data range, ensuring data quality. Data identified as outliers by both methods is directly removed. Suspected outliers identified by only one method undergo manual review. The decision to remove or retain outliers is based on the review results, avoiding interference from outliers and preventing the accidental deletion of valid data. Next, missing value imputation is performed. Appropriate imputation methods are selected based on the characteristics of different indicators to fill data gaps, ensuring data integrity and preventing bias in analysis results due to missing data. Finally, data standardization is implemented, converting indicator data of different units and magnitudes into a unified standard, eliminating the influence of dimensions, and making various types of data comparable horizontally, laying the foundation for subsequent causal relationship analysis and coefficient calculation. Finally, redundant indicators are screened out, and duplicate, irrelevant, or indicators with minimal impact on steel bar prices are removed. This simplifies the data dimensions, reduces the complexity of subsequent calculations, improves processing efficiency, and outputs a standardized data sequence.

[0035] The causal mining module receives standardized data sequences and employs a time-series causal strength algorithm to analyze the relationships between variables. This algorithm effectively uncovers causal associations between variables in time-series data, meeting the needs of medium-term forecasting for long-term data correlation analysis. First, the standardized data sequences are aligned by timestamps, and non-redundant variables with variance greater than 0.01 are selected to ensure that all variables involved in the analysis have actual volatility significance and to eliminate interference from worthless variables. Then, the causal strength between each industry indicator variable and the rebar price variable is calculated separately to determine the true direction of causal transmission and identify the influencing factors and the affected objects. Using a 1-day step, time lag values ​​from 0 to 30 days are traversed, and the time lag corresponding to the maximum causal strength is found as the final time lag parameter. If multiple time lags correspond to the same maximum value, the smallest time lag is selected, accurately capturing the time lag effect of industry indicators on rebar prices, providing a time dimension reference for medium-term forecasting. The calculation results of the time-series causal strength algorithm are the quantitative values ​​of the causal strength between variables, providing a quantitative basis for subsequent correlation judgment and coefficient optimization. Subsequently, pseudo-associations are marked. A pseudo-association is initially identified when the causal strength is less than 0.2. The synchronous variation coefficient is then calculated. If the synchronous variation coefficient is greater than the causal strength and there is no supply chain transmission logic to support a direct causal relationship, the pseudo-association attribute is confirmed and marked, eliminating the interference of false associations on the prediction results and ensuring the accuracy of the analysis. Finally, a causal directed acyclic graph is constructed, using non-redundant industry indicator variables and rebar price variables as nodes. Attribute information such as variable name, industry, data type, and time-series data sequence identifier is associated, ensuring complete and traceable node information. Directed edges are drawn according to the causal direction, associating causal strength and time lag parameters to intuitively present the details of the causal relationships between variables. Nodes are arranged hierarchically according to the upstream and downstream of the supply chain, with upstream industry indicator variable nodes at the top, midstream at the middle, and rebar price variable nodes at the bottom, conforming to the supply chain transmission logic and facilitating the understanding of the influence path. Directed edges are graded and colored according to causal strength, intuitively distinguishing the strength of causal relationships through color. Directed edges corresponding to pseudo-associations are removed, and relevant information is recorded to ensure the simplicity and effectiveness of the graph structure. Verify and optimize the graph structure, eliminate cyclic causal paths, ensure the rationality and logic of the graph structure, generate a complete causal directed acyclic graph, and provide clear and accurate causal relationship support for coefficient construction.

[0036] The coefficient construction module calculates the initial correlation coefficients between various industry indicators and steel rebar prices using a time-series weighted cosine similarity algorithm. This algorithm fully considers the weight differences of data at different time points, aligning with the characteristics of time-series data, making the initial correlation coefficients more reflective of the actual correlation. Combining the output of the causal directed acyclic graph, the initial correlation coefficients are optimized using causal-industry coupling coefficient optimization technology, ensuring that the correlation coefficients both reflect causal relationships and conform to the actual transmission logic of the steel rebar industry chain, improving the rationality and effectiveness of the coefficients. Initial correlation coefficients corresponding to spurious correlations are directly set to zero to eliminate the impact of false correlations. For genuine causal relationships, optimized causal transmission coefficients are calculated by combining causal strength, time lag parameters, industry chain influence coefficients, and building material cost sensitivity coefficients, enabling the coefficients to accurately quantify the transmission strength of various industry indicators on steel rebar prices. When constructing the causal transmission coefficient matrix, industry indicator variables are used as row variables with relevant labels, and steel rebar price variables are used as column variables with relevant labels, clearly defining the meaning of the matrix rows and columns. The optimized causal transmission coefficients are then filled in, forming a quantitative correspondence between industry indicators and steel rebar prices. The coefficients are calculated row by row. If the sum is not in the range of 0.95-1.05, the coefficients are adjusted for normalization to ensure the standardization and rationality of the matrix coefficients, facilitating subsequent prediction calculations. After completion, the matrix integrity and logical consistency are verified to ensure there are no missing or incorrect coefficients, generating the final causal transmission coefficient matrix, providing accurate quantitative basis for medium-term price forecasting.

[0037] The forecast output module collects the latest data from various related industries in real time, promptly grasps industry dynamics, and calculates the real-time data changes of each industry indicator at the current moment. This ensures that the data input into the forecast model reflects the latest market conditions and meets the timeliness requirements of medium-term forecasts. Based on the causal transmission coefficient matrix, a dynamic price forecast fusion algorithm is used. This algorithm integrates multiple influencing factors and effectively merges multi-dimensional data, meeting the needs of medium-term forecasts for comprehensive consideration of complex factors. During the calculation process, parameters such as the current benchmark price of rebar, the causal transmission coefficients of each industry to rebar, the real-time data changes of each industry, the time lag-forecast cycle adaptation factor, the market volatility correction coefficient, and the historical 30-day price volatility of rebar are combined to comprehensively cover various factors affecting rebar prices, improving the accuracy of medium-term forecasts. Finally, the predicted price of rebar for the next 30 days is calculated, providing a reliable price reference for construction companies to formulate medium- and long-term procurement plans and project cost accounting, helping companies mitigate price volatility risks.

[0038] In summary, this embodiment addresses the need for medium-term steel rebar price forecasting by integrating upstream and downstream data from multiple channels across the industry chain, ensuring data coverage of all stages, including raw material supply, production, and transportation. The pre-processed standardized data is then used to accurately identify the causal direction, time lag parameters, and intensity of variables through a time-series causal strength algorithm, constructing a causal directed acyclic graph that conforms to industry logic. A standardized causal transmission coefficient matrix is ​​generated by combining a time-series weighted cosine similarity algorithm with optimization techniques. Finally, a dynamic price forecasting fusion algorithm, incorporating real-time data changes and market fluctuations, efficiently calculates the steel rebar price for the next 30 days. This progressive process provides a scientific basis for enterprises to formulate medium- and long-term procurement plans and control costs.

[0039] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A multi-dimensional data fusion-based dynamic price prediction system for building materials, characterized in that: The system includes: Data integration module: Connects to data interfaces of the construction industry and related industries, collects multi-dimensional raw data through preset methods, and aggregates it into a unified data pool after format standardization and compliance verification; Data preprocessing module: Reads raw data from the data pool, performs outlier removal, missing value filling and data standardization transformation, and outputs standardized data sequences after filtering redundant indicators; Causal mining module: Receives standardized data sequences, uses time-series causal strength algorithm to identify causal direction, time delay parameters and causal strength between variables, marks pseudo-associations and constructs causal directed acyclic graph; Coefficient construction module: Calculates the initial correlation coefficient between industry indicators and building material prices using the time-series weighted cosine similarity algorithm, outputs the results using a causal directed acyclic graph, optimizes the initial correlation coefficient using causal-industry coupling coefficient optimization technology, and constructs a causal transmission coefficient matrix; Prediction output module: Collects the latest industry data in real time and calculates the changes. Based on the causal transmission coefficient matrix, it calculates the predicted price of building materials through a dynamic price prediction fusion algorithm and outputs the prediction results.

2. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 1, characterized in that, The data integration module connects to the construction industry and related industries, including the mining industry, chemical industry, energy industry, and logistics equipment manufacturing industry. The specific data interfaces are the official data platform interface, the chemical industry association statistical platform interface, the automobile industry association sales data interface, the local building materials trading market database interface, and the ERP system interface of large building materials production enterprises.

3. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 1, characterized in that, The data integration module includes preset collection methods such as API interface retrieval authorized by OAuth2.0, direct database connection extraction via JDBC / ODBC protocol, and Scrapy crawler crawling with compliant configuration. The collected multi-dimensional raw data specifically includes mining equipment maintenance plans, raw material mining volume and inventory in the mining industry; admixture production capacity, capacity utilization rate and ex-factory price in the chemical industry; peak and valley electricity price standards, coal market prices and energy consumption data of building material production enterprises in the energy industry; heavy truck production and sales data, transportation capacity supply and demand gap and transportation route freight rates in the logistics equipment manufacturing industry; and real-time quotations, historical transaction prices and capacity utilization rate data of building materials in the construction industry.

4. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 1, characterized in that, In the data preprocessing module, when performing outlier removal, the mean μ and standard deviation σ of each indicator field are first calculated, and outliers are initially identified using the 3σ principle. Then, the box plot analysis method is used to calculate the quartiles Q1, Q2, Q3 and the interquartile range IQR. Data that are less than Q1-1.5IQR or greater than Q3+1.5IQR are marked as suspected outliers. Data that are identified as outliers by both methods are directly removed. Suspected outliers identified by only one method trigger a manual review process, and the removal or retention is determined after review.

5. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 1, characterized in that, In the causal mining module, the mathematical expression for the temporal causal strength algorithm is: in, Industry benchmark Building material prices Causality strength with a time lag of τ days For linear-nonlinear fusion weights, for At any moment and At any moment Mutual information value, NLCC for At any moment and At any moment The nonlinear correlation coefficient, TDR is the time-delay decay coefficient. It is a time-delay reliability index used to measure the identified time-delay parameters. The credibility of Representing the current moment, It is a time delay parameter. It is the first Industry indicator variables, No. One variable is the price of building materials.

6. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 5, characterized in that, In the causal mining module, the specific steps for identifying the causal direction, time delay parameters, and causal strength between variables using the time-series causal strength algorithm are as follows: The received standardized data sequences are aligned by timestamps, and non-redundant variables with variance greater than 0.01 are filtered out. The industry indicator variables are then calculated using the time-series causal strength algorithm. Building material price variables causal strength and right causal strength The direction with greater causal strength is taken as the true causal transmission direction between variables; the time lag values ​​from 0 to 30 days are traversed with a step size of 1 day. Calculate each corresponding ,Pick The maximum value corresponding to As the final time delay parameter, when there are two or more different time delay parameters Both increase the strength of causality If the values ​​reach the same maximum value, then the smallest value among them is selected. As the final time delay parameter, the time-series causality strength algorithm will be used to calculate the time delay parameter. It is directly used as the quantification result of the causal strength between variables, and its value range is [0, 1].

7. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 5, characterized in that, In the causal mining module, the specific steps for marking pseudo-associations are as follows: calculating industry indicators using a time-series causal strength algorithm. With building material prices The causal strength between them is determined by a preset causal strength threshold of 0.

2. When the calculated causal strength is... When it is less than 0.2, a preliminary judgment is made. and This is a pseudo-association; further calculations are needed. and Synergy coefficient When Syn Furthermore, since there is no supply chain transmission logic to support a direct causal relationship between the two, their pseudo-association attribute is ultimately confirmed. For the confirmed pseudo-association, pseudo-association labels are marked in the causal relationship dataset, and the corresponding labels are recorded. Numerical value and synchronous variation coefficient Syn Numerical value.

8. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 5, characterized in that, The specific steps for constructing a causal directed acyclic graph in the causal mining module are as follows: Determine the nodes in the graph: Select the filtered, non-redundant industry indicator variables. and price variables of various types of building materials As nodes in the graph, each node is associated with attribute information, including variable name, industry, data type, and time series data sequence identifier; Drawing directed edges: Based on the true causal direction between variables identified by the temporal causal strength algorithm, directed edges are drawn from the cause variable node to the result variable node. Each directed edge is associated with attribute information, including the causal strength of the corresponding variable pair. Time delay parameters ; Hierarchical Node Layout: A hierarchical layout algorithm is used to arrange nodes, dividing them into three layers according to the upstream and downstream order of the industry chain, with upstream industry indicator variables... The node is located at the top layer, representing midstream industry indicator variables. The node is located in the middle layer, and the building material price variable... The node is located at the bottom layer; Hierarchical coloring of directed edges: based on causality strength Hierarchical coloring of directed edges. Use red at times, Orange was used at that time. Blue was used at that time; Eliminating spurious edges: This relates to causal strength. For pseudo-association variable pairs, no corresponding directed edges are drawn; instead, the pseudo-association variable name and corresponding... are recorded only in the attribute information of the node to which it belongs. Numerical value; Graph structure validation and optimization: Validate the graph for cyclic causal paths; if found, remove the cyclic path. Minimize the directed edges to eliminate cycles, ultimately generating a complete causal directed acyclic graph.

9. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 1, characterized in that, In the coefficient construction module, the mathematical expression for the time-weighted cosine similarity algorithm is: in, Industry benchmark With building material prices The initial correlation coefficient, for Time Industry Indicators Standardized data, for Construction material prices Standardized data, for Standardized data weighted mean, for Standardized data weighted mean, These are time-series trend weights preset based on a time decay function or business importance. To prevent the minimum value where the denominator is 0, It represents the total number of moments in the time series data.

10. The multi-dimensional data fusion-based dynamic prediction system for building material prices according to claim 1, characterized in that, The specific steps for constructing the causal transmission coefficient matrix in the coefficient construction module are as follows: All selected industry indicator variables are used as row variables of the matrix, with each row variable corresponding to a unique industry category identifier and indicator name; the price variables of each type of building material are used as column variables of the matrix, with each column variable corresponding to a unique building material type identifier and price indicator name; the optimized causal transmission coefficients corresponding to each industry indicator variable and building material price variable, obtained through causal-industry coupling coefficient optimization technology, are filled into the corresponding positions of the matrix one by one according to the correspondence between row variables and column variables; for the completed matrix, the sum of all causal transmission coefficients in each row is calculated. ;when If the coefficient is not within the range of 0.95 to 1.05, then multiply each causal transmission coefficient in that row by an adjustment factor. The adjustment coefficient This normalizes the sum of all coefficients in the row to 1. After standardization, the matrix is ​​checked for completeness to ensure there are no missing coefficients or logical contradictions. Finally, a causal transmission coefficient matrix containing the quantitative transmission relationship between all industry indicator variables and building material price variables is generated.

Images