Coal-based intelligent decision generation method and device, equipment, storage medium

Abstract

The application provides a coal-based intelligent decision generation method and device, equipment and storage medium, and belongs to the technical field of intelligent decision, which comprises the following steps: coal price prediction is performed based on coal quality index data and standardized coal market data to obtain coal price prediction trend data; the coal quality index data is a comprehensive index representing the physical and chemical properties and industrial applicability of coal; knowledge vectorization processing is performed on standardized coal information text data to obtain a plurality of knowledge vectors; similarity calculation is performed based on a target retrieval vector and each knowledge vector to screen a knowledge vector corresponding to the maximum similarity; and a coal decision report is generated based on the knowledge vector corresponding to the maximum similarity and the coal price prediction trend data. The coal-based intelligent decision generation method and device, equipment and storage medium provided by the application can improve the accuracy and reliability of the generated coal-based intelligent decision.

Description

Technical Field

[0001] This application belongs to the field of intelligent decision-making technology, and more specifically, relates to a method, apparatus, equipment, and storage medium for generating intelligent decisions based on coal. Background Technology

[0002] As a core raw material for my country's industrial production and energy supply, coal's market price fluctuations directly impact the production costs and operating efficiency of downstream industries such as metallurgy, energy, and equipment manufacturing. Coal prices are influenced by a combination of factors, including resource endowment in producing areas, market supply and demand, industry policy regulation, transportation conditions, and international market conditions, exhibiting non-linear, strongly coupled, and significantly time-varying characteristics. Accurately assessing coal price trends and formulating scientific and reasonable procurement strategies have become crucial prerequisites for relevant enterprises to control costs, stabilize operations, and enhance market competitiveness.

[0003] Currently, coal price analysis and procurement decisions mainly rely on manual summarization of market information, statistical analysis of historical data, and subjective judgment based on experience. These methods have significant limitations. Existing technologies often use single-dimensional data for price calculations, making it difficult to effectively integrate and deeply utilize multi-source information. Insufficient quantitative analysis of price-influencing factors leads to low accuracy in price trend predictions, failing to provide reliable support for medium- and long-term business decisions. Furthermore, the utilization rate of unstructured data such as industry news and market dynamics is low, and the level of intelligence and accuracy in decision-making is insufficient to meet the needs of refined enterprise operations. Therefore, a method is needed to improve the accuracy and reliability of generating intelligent coal-based decisions. Summary of the Invention

[0004] This application provides a method, apparatus, device, and storage medium for generating intelligent decisions based on coal, in order to solve the problems of low accuracy and poor reliability in generating intelligent decisions based on coal, and to achieve the goal of improving the accuracy and reliability of generating intelligent decisions based on coal.

[0005] According to one aspect of the embodiments of this application, a coal-based intelligent decision generation method is provided, comprising: Obtain user search requests, which are inquiries about price trend analysis and procurement decision recommendations for coal in at least one dimension: region, type, and time. Based on user search needs, coal information and coal market data are obtained; the coal information is preprocessed to obtain standardized coal information text data; the coal market data is preprocessed to obtain standardized coal market data; and the user search needs are encoded into text vectors to obtain target search vectors. Coal price forecasting is based on coal quality index data and standardized coal market data to obtain coal price forecast trend data; coal quality index data is a comprehensive indicator that characterizes the physical and chemical properties and industrial applicability of coal. Standardized coal information text data is processed into knowledge vectors to obtain multiple knowledge vectors; similarity is calculated between the target retrieval vector and each knowledge vector, and the knowledge vector with the highest similarity is selected. A coal decision-making report is generated based on the knowledge vector corresponding to the maximum similarity and coal price forecast trend data.

[0006] According to one aspect of the embodiments of this application, a coal-based intelligent decision generation device is provided, comprising: The search demand acquisition module is used to acquire user search demands. User search demands are inquiries about price trend analysis and procurement decision suggestions for coal in at least one dimension of region, variety, and time. The preprocessing module is used to obtain coal information and coal market data based on user search requirements; perform text preprocessing on the coal information to obtain standardized coal information text data; perform data preprocessing on the coal market data to obtain standardized coal market data; and perform text vector encoding on the user search requirements to obtain the target search vector. The price forecasting module is used to forecast coal prices based on coal quality index data and standardized coal market data, and to obtain coal price forecast trend data. The coal quality index data is a comprehensive indicator that characterizes the physical and chemical properties and industrial applicability of coal. The matching module is used to perform knowledge vectorization on standardized coal information text data to obtain multiple knowledge vectors; based on the similarity calculation between the target retrieval vector and each knowledge vector, the knowledge vector corresponding to the maximum similarity is selected. The decision report generation module is used to generate coal decision reports based on the knowledge vector corresponding to the maximum similarity and coal price forecast trend data.

[0007] According to one aspect of the embodiments of this application, an electronic device is provided, the electronic device including a processor and a memory, the memory storing a computer program, the computer program being loaded and executed by the processor to implement the above-described intelligent decision generation method based on coal.

[0008] According to one aspect of an embodiment of this application, the computer program product includes a computer program stored in a computer-readable storage medium. A processor of an electronic device reads the computer program from the computer-readable storage medium and executes the computer program, causing the electronic device to perform the aforementioned coal-based intelligent decision generation method.

[0009] The beneficial effects of the technical solution provided in this application are as follows: The intelligent decision-making generation method, apparatus, equipment, and storage medium based on coal provided in this application, compared with related technologies, simultaneously acquires coal information and coal market data based on user retrieval needs. Through standardized preprocessing, it effectively integrates multi-source data, avoiding the limitations of single data sources. Simultaneously, it combines coal quality indicators with standardized coal market data for price prediction, comprehensively considering various key factors affecting coal prices, improving the accuracy of price trend prediction, and providing a reliable basis for enterprises' medium- and long-term procurement decisions. This application also performs knowledge vectorization processing on standardized coal information text data, and combines it with vector matching based on user retrieval needs to quickly filter out the most relevant knowledge vectors, solving the problem of low utilization of unstructured data in existing technologies. Finally, it combines the matched knowledge vectors with coal price prediction trend data to generate a decision report, improving the level of intelligent decision-making, helping enterprises formulate scientific procurement strategies, effectively control costs, stabilize operations, and adapt to the needs of refined enterprise operations. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 A flowchart illustrating the intelligent decision generation method based on coal provided in this application embodiment; Figure 2 A schematic diagram of the coal price forecasting process provided in the embodiments of this application; Figure 3 This is a schematic diagram of the process for generating coal decision reports provided in an embodiment of this application; Figure 4 A structural block diagram of a coal-based intelligent decision generation device provided in an embodiment of this application; Figure 5 A schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0012] To make the objectives, technical solutions, and advantages of this application clearer, the embodiments of this application will be described in further detail below with reference to the accompanying drawings.

[0013] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numerals in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0014] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any and all possible combinations of one or more of the associated listed items.

[0015] Figure 1 This is a flowchart illustrating the intelligent decision generation method based on coal provided in an embodiment of this application. The method is executed by an electronic device and may include: S101: Obtain user search requests. User search requests are inquiries about price trend analysis and procurement decision recommendations for coal in at least one dimension, including region, type, and time.

[0016] In this embodiment, the coal-producing region in the user's search query refers to the specific geographical area where coal is produced or sold, such as region A or region B; the coal type refers to the specific classification of coal, such as coking coal, lean coal, and prime coking coal; the time period refers to the price analysis period that the user is interested in, such as monthly or quarterly; the dimensions refer to different angles from which coal prices are analyzed; price trend analysis refers to the judgment of the pattern of coal price changes, such as analyzing the price trend of coking coal; and procurement decision suggestions refer to the procurement-related guidance given based on price trends. For example, the user's search query could be a quarterly price trend analysis and procurement decision suggestions for coking coal in region A.

[0017] S102: Based on user search needs, obtain coal information and coal market data; perform text preprocessing on the coal information to obtain standardized coal information text data; perform data preprocessing on the coal market data to obtain standardized coal market data; and perform text vector encoding on the user search needs to obtain the target search vector.

[0018] In this embodiment, coal information refers to industry news and policy notices related to coal, such as coking coal production capacity policies and market transaction updates. Coal market data includes data related to coal transaction prices, inventory, and production, such as the ex-works price of coking coal in region A and port inventory data. Standardized coal information text data is coal information text with a unified format after text preprocessing. Standardized coal market data is coal market data with a unified format after data preprocessing. Text vector encoding is a process of converting text into vectors, and the target retrieval vector is the vector obtained by encoding the user's search query.

[0019] For example, this embodiment, based on user search needs, collects relevant coal information and coal market data from coal industry information platforms and market trading platforms using web crawler technology. Next, the collected coal information undergoes text preprocessing: first, text cleaning to remove invalid characters and duplicate content; then, key information extraction to extract content related to coal prices and market supply and demand; and finally, normalization to unify the text format, resulting in standardized coal information text data. Then, the coal market data undergoes data preprocessing: first, cleaning to remove outliers and missing values; then, normalization to unify data units and formats, resulting in standardized coal market data. Finally, the user search needs are encoded into text vectors through text segmentation, feature extraction, and other processing, converting the textual user search needs into target search vectors for subsequent knowledge vector similarity matching.

[0020] This embodiment enables the accurate acquisition and standardized processing of coal information and coal market data, ensuring data relevance and standardization, and resolving the problems of messy, invalid, and non-standard raw data. This provides reliable data support for subsequent price forecasting and decision-making report generation. Simultaneously, this embodiment uses text vector encoding to convert user search requests into target search vectors, providing technical support for subsequent knowledge matching and improving the accuracy of knowledge matching.

[0021] In this embodiment, coal information is preprocessed to obtain standardized coal information text data, including: Text cleaning is performed on coal information to obtain cleaned coal information text data. The cleaned coal information text data is deduplicated to obtain the deduplicated coal information text data. Key information is extracted from the deduplicated coal information text data to obtain structured coal information text data. Text normalization is performed on structured coal information text data to obtain standardized coal information text data.

[0022] In this embodiment, text cleaning involves removing invalid content from coal information to obtain cleaned coal information text data; text deduplication involves deleting duplicate content from the cleaned coal information text data to obtain deduplicated coal information text data; key information extraction involves extracting core content from the deduplicated coal information text data to obtain structured coal information text data; and text normalization involves unifying the format of the structured coal information text data to obtain standardized coal information text data.

[0023] For example, this embodiment performs text cleaning on coal information. Invalid characters, random symbols, irrelevant advertising content, and blank paragraphs are filtered out, retaining only valid coal-related information, resulting in cleaned coal information text data. Next, the cleaned coal information text data is deduplicated. By comparing the similarity of text content, identical or highly similar text is identified and deleted to avoid duplicate data interference, resulting in deduplicated coal information text data. Then, key information is extracted from the deduplicated coal information text data, extracting core content related to coal prices, supply and demand, and policies, and organizing it into fixed categories to obtain structured coal information text data. Finally, the structured coal information text data is normalized, unifying the font, format, and expression standards, standardizing the same information expressed differently into a standard form, resulting in standardized coal information text data.

[0024] This embodiment addresses the issues of disorganized, invalid, repetitive, and inconsistently formatted coal information, gradually optimizing data quality to obtain standardized coal information text data. Text cleaning and deduplication reduce invalid data interference, improving data processing efficiency; key information extraction and text normalization ensure the text data is standardized and usable, providing reliable support for subsequent knowledge vectorization processing. Overall, it improves the utilization and usability of coal information, reduces the workload of subsequent processing, ensures the accuracy of subsequent stages, and ultimately helps improve the reliability and scientific nature of intelligent coal decision-making, meeting the needs of refined enterprise operations.

[0025] S103: Coal price forecasting is conducted based on coal quality index data and standardized coal market data to obtain coal price forecast trend data; coal quality index data are comprehensive indicators characterizing the physical and chemical properties and industrial applicability of coal.

[0026] In this embodiment, coal quality index data refers to comprehensive indicators characterizing the physical and chemical properties and industrial applicability of coal, such as ash content, sulfur content, and caking index. Coal price forecasting is the operation of judging future coal price changes. Coal price forecast trend data is the predicted future price change pattern data of coal. Physical properties refer to physical attributes such as appearance and hardness of coal, chemical properties refer to chemical attributes such as composition and purity of coal, and industrial applicability refers to the degree of applicability of coal in industrial production. For example, coking coal, as a core raw material for the iron and steel metallurgical industry, has more stringent quality requirements, and its price fluctuations have a more significant impact on the production costs of steel enterprises. The quality indicators of coking coal, such as ash content, sulfur content, volatile matter, caking index, and plastic layer thickness, directly determine its industrial use value, and factors such as market supply and demand, transportation conditions, environmental policies, and import quotas are more sensitive to the price of coking coal.

[0027] Considering that single data points cannot fully reflect the patterns of price changes, coal quality indicators determine coal quality and use value, forming the basis of price formation; standardized coal market data reflects market supply and demand and transaction conditions, and is the core driver of price fluctuations. Combining these two types of data for price forecasting allows for a comprehensive consideration of key factors influencing prices, addressing the problem of large prediction biases from single-data sources. Logically, this approach, centered on multi-source data fusion, ensures that the forecast results align with market realities, providing reliable support for subsequent intelligent decision-making.

[0028] In this embodiment, coal price forecasting is performed based on coal quality index data and standardized coal market data to obtain coal price forecast trend data, including: Price impact characteristics are fused based on coal quality index data and standardized coal market data to obtain fused characteristic data. By extracting the temporal variation patterns from the fused feature data, we can obtain the temporal pattern data of coal prices. By performing trend extrapolation on the time-series data of coal prices, we can obtain predicted trend data for coal prices.

[0029] In this embodiment, price impact characteristics are fused based on coal quality index data and standardized coal market data to obtain fused characteristic data, including: Quality attribute features are extracted from coal quality index data to obtain coal quality attribute feature data; the first price influence degree of coal quality attribute feature data is determined, and the coal quality attribute feature data is weighted based on the first price influence degree to obtain weighted coal quality attribute feature data. Market characteristics are extracted from standardized coal market data to obtain coal market characteristic data; the second price influence of the coal market characteristic data is determined, and the coal market characteristic data is weighted based on the second price influence to obtain weighted coal market characteristic data. The weighted coal quality attribute feature data and the weighted coal market condition feature data are combined to obtain fused feature data.

[0030] In this embodiment, price impact feature fusion is the operation of integrating features affecting prices from two types of data. The fused feature data is the feature data that can be used for prediction after fusion. Time-series change pattern extraction is the operation of mining the time-dimensional change patterns of the data. Coal price time-series pattern data reflects the pattern of coal price changes over time. Trend inference is the operation of predicting future price trends based on patterns. Quality attribute feature extraction is the operation of refining the core features of coal quality, and market condition feature extraction is the operation of refining the core features of market transactions. Price impact degree is the degree of influence of a feature on price, and feature splicing is the operation of integrating two types of weighted features. For example, coal quality attribute feature data may include feature information corresponding to ash content and sulfur content.

[0031] In this embodiment, coal quality index data and standardized coal market data are acquired. The coal quality index data are comprehensive indicators characterizing the physical and chemical properties and industrial applicability of coal, including indicators such as ash content and sulfur content. For example, for coking coal, the coal quality index data specifically includes core indicators that directly affect coking coal pricing and industrial applicability, such as ash content, sulfur content, volatile matter, fixed carbon, caking index G, maximum plastic layer thickness Y, Oya expansion degree, coke residue characteristics, reactivity, or strength. The standardized coal market data is coal market data in a unified format after data preprocessing, including price and inventory data. Next, quality attribute features are extracted from the coal quality index data, refining core quality features related to price to obtain coal quality attribute feature data. By analyzing the correlation between quality features and prices in historical data, the first price influence degree of the coal quality attribute feature data is determined. Based on the first price influence degree, the coal quality attribute feature data is weighted to obtain weighted coal quality attribute feature data. Then, market characteristics are extracted from standardized coal market data, refining core characteristics such as supply and demand and price fluctuations to obtain coal market characteristic data. By analyzing historical data, the impact of each characteristic in the coal market characteristic data on coal prices is quantified, determining the second price influence of the coal market characteristic data. Based on the second price influence, the coal market characteristic data is weighted to obtain weighted coal market characteristic data. Finally, the weighted coal quality attribute characteristic data and the weighted coal market characteristic data are concatenated to obtain fused characteristic data. The temporal variation patterns of the fused characteristic data are extracted to obtain coal price temporal pattern data. Trend extrapolation is performed on the coal price temporal pattern data to obtain coal price prediction trend data.

[0032] The formula for calculating the first price influence is:

[0033] in, The first price influence degree represents the comprehensive influence of coal quality attribute data on coal prices, with a value range of [0,1]. This refers to the sequence number of the coal quality attribute characteristic data. ; This represents the total number of coal quality attribute feature data. For the first The correlation coefficient between coal quality attribute data and historical coal prices reflects the strength of the linear correlation between the characteristics and prices, and the value range is [1,-1]. For the first The coefficient of variation of the coal quality attribute characteristic data (coefficient of variation = the first coal quality attribute characteristic data) Standard deviation of each coal quality attribute characteristic data / the first The mean of each coal quality attribute feature data is derived from the statistical fluctuation of the coal quality attribute feature data itself. The smaller the value, the more stable the feature is. For the first The industry benchmark weights for individual coal quality attribute characteristic data can be determined based on coal industry standards, with values ​​ranging from [0,1]. For example, in coking coal application scenarios, the industry benchmark weights are determined based on coking coal quality standards and metallurgical industry procurement requirements, with key indicators such as sulfur content, ash content, caking index, and plastic layer thickness given higher weights to align with actual coking coal pricing rules.

[0034] The above formula integrates three dimensions: correlation strength, stability, and industry importance. First, based on the correlation coefficients between various coal quality attribute data and historical coal prices, normalization is performed by dividing by the maximum value of these correlation coefficients to ensure the comparability of the influence strength of different coal quality attribute characteristics. Second, a reverse correction term for the coefficient of variation of coal quality attribute data is introduced to take the stability of this characteristic data into account; the smaller the characteristic fluctuation and the higher the stability, the higher its weight in influencing price, avoiding interference from unstable characteristics in prediction. Finally, combined with the benchmark weights determined by coal industry standards, key indicators such as calorific value and ash content are given higher fundamental importance, better aligning with the actual pricing logic of the industry. Among these, the first... The correlation coefficient between coal quality attributes and historical coal prices is calculated using the ratio of covariance to standard deviation. This correlation coefficient is used to quantify the degree of linear correlation between coal quality attributes and coal prices.

[0035]

[0036] in, For the first The correlation coefficient between individual coal quality attribute data and historical coal prices For the first Covariance of individual coal quality attribute characteristics and historical coal prices; For the first Variance of individual coal quality attribute characteristics; This represents the variance of historical coal prices.

[0037] The formula for calculating the second price influence is:

[0038] in, The second price influence degree represents the comprehensive impact of coal market characteristic data on coal prices, with a value range of [0,1]. This refers to the sequence number of the coal market characteristic data. ; This represents the total amount of data on coal market characteristics. For the first The mutual information value between coal market characteristic data and historical coal prices is obtained by calculating the log ratio of their joint probability distribution to their marginal probability distribution. It reflects the strength of the nonlinear correlation. The larger the value, the stronger the nonlinear correlation between the two. The value range is [0,1]. For the first The time-series fluctuation synchronization coefficient of coal market characteristic data is derived from the synchronization statistics of coal market characteristic data and historical coal price fluctuation trends, and the value range is [0,1]. For the first The market-driving weight of each coal market characteristic data point can be determined based on its contribution rate to historical coal price fluctuations, reflecting the actual driving force of that characteristic on prices. Specifically, the market-driving weight is determined based on the contribution rate of each coal market characteristic data point to historical coal price fluctuations, and can be calculated using the standardized regression coefficient method: a multiple linear regression model is constructed with historical coal prices as the dependent variable and each coal market characteristic data point as the independent variable; after standardizing the coal market characteristic data points and historical coal prices, the standardized regression coefficients of each coal market characteristic data point are calculated; the absolute value of each standardized regression coefficient is taken, and the absolute value of the standardized regression coefficient of each coal market characteristic data point is divided by the sum of the absolute values ​​of all the standardized regression coefficients of the characteristics to obtain the market-driving weight of that coal market characteristic data point. The value range is [0,1]. If ,but .

[0039] Suppose that the coal market characteristic data are arranged in chronological order as follows: Historical coal prices are arranged in chronological order as follows: Where T is the number of time points, .

[0040] The steps for calculating the synchronization coefficient of time-series fluctuations are as follows: Starting from the second time point (i.e., from t=2 to T), determine in turn whether the fluctuation direction of each pair of adjacent time points (t-1, t) is consistent; If the direction of change in coal market characteristic data ( ) and the historical trend of coal price changes ( If the changes are the same (i.e., both increase or decrease simultaneously), it is considered a "synchronization". Otherwise (if the changes are in opposite directions, or one of them does not change), it is not considered a synchronization. Count the total number of synchronous occurrences across all adjacent time points; The ratio obtained by dividing the total number of synchronizations by the total number of adjacent time points (i.e., T-1) is the timing fluctuation synchronization coefficient.

[0041] The synchronization coefficient for time-series fluctuations ranges from 0 to 1. A value closer to 1 indicates a more synchronized trend between the coal market characteristics and historical coal prices; a value closer to 0 indicates that the trends are almost asynchronous or even inversely related. For example, assuming there are 5 time points (forming 4 adjacent intervals), and the coal market characteristics and historical coal prices fluctuate in the same direction in 3 of these intervals, the synchronization coefficient is 3 / 4 = 0.75, indicating a high degree of synchronization.

[0042] The above formula addresses the nonlinear, time-varying, and strongly coupled characteristics of coal market data by integrating three dimensions: nonlinear correlation strength, temporal fluctuation synchronicity, and actual market driving force. This overcomes the limitation of a single correlation coefficient failing to accurately reflect the true impact of coal market data on prices. First, based on the mutual information value between various market data features and historical coal prices, normalization is performed by dividing by the maximum value of this mutual information value, mapping the nonlinear correlation strength to the 0-1 range to ensure comparable impact strength of market data features. Second, a temporal fluctuation synchronicity coefficient is introduced to consider the synchronicity of market data feature fluctuations with historical coal prices. The more synchronized the feature fluctuations are with the price fluctuation trends, the more stable their driving effect on prices, and the higher their corresponding weight contribution. Finally, combined with market-driven weights, market features with high contribution rates to price fluctuations are given higher fundamental importance, better aligning with the actual pricing logic in market transactions. Based on this, the three indicators are weighted, integrated, and normalized by market-driven weights, so that the final value of the second price influence is stable in the range of 0 to 1. This not only ensures that the weight distribution among different market condition characteristic data is reasonable, but also can be directly used for subsequent feature weighting calculations, providing a scientific and reliable weight basis for subsequent price influence feature integration.

[0043] For example, this embodiment can use a Bidirectional Gated Recurrent Unit (BiGRU) model, which is an improved recurrent neural network. It has the advantage of capturing the forward and backward dependencies of time series data, adapts to the nonlinear and non-stationary time series variation characteristics of coal prices, and can accurately explore the complex relationship between coal prices and multiple influencing factors.

[0044] First, the input data for the BiGRU model is prepared. The input data consists of fused feature data obtained by integrating price impact features from coal quality index data and standardized coal market data. This data integrates weighted coal quality attribute feature data and weighted coal market condition feature data, covering quality indicators such as coal ash content and sulfur content, as well as market condition features such as coal price, inventory, and production. To adapt to the input requirements of the BiGRU model, the fused feature data needs to be time-series normalized, sorted by time dimension (e.g., day, week, month) to ensure data continuity. Linear interpolation is used to fill missing values ​​and remove abnormal deviations to avoid interfering with model training accuracy. For example, in the coking coal price prediction scenario, the fused feature data further includes features such as coking coal port inventory, coking plant operating rate, steel industry procurement demand, and imported coking coal landed price, making the prediction results more consistent with the actual operating patterns of the coking coal market.

[0045] Secondly, the BiGRU model parameters are configured and trained. The BiGRU model consists of two independent GRU layers, which process the input temporal fusion feature data from the forward and backward directions respectively. By controlling the transmission and forgetting of feature information through reset and update gates, key temporal features can be effectively preserved while suppressing redundant information interference. The core model parameters are configured as follows: the input layer dimension should be consistent with the number of features in the fusion feature data, typically 8-20 dimensions; the number of hidden layer neurons should be set to 32-128, which can be flexibly adjusted according to the complexity of the fusion feature data; 64-128 neurons are used when the feature dimension is high, and 32-64 neurons are used when the feature dimension is low; mean squared error is selected as the loss function; the Adam optimization algorithm is used to iteratively optimize the model parameters, with a learning rate set to 0.001-0.01, dynamically decaying according to the number of training iterations, and the number of iterations set to 100-300 rounds; the dropout coefficient is set to 0.1-0.3 to suppress model overfitting. During training, the normalized fused feature data is divided into a training set and a validation set in a 7:3 ratio. The model is trained using the training set, and the parameters are adjusted in real time using the validation set until the model prediction error reaches a preset threshold (usually 0.01-0.05), thus completing the model training.

[0046] Next, the time-series variation patterns are extracted and trends are extrapolated. The trained BiGRU model is applied to the processed fused feature data. This model processes the data from the initial time step to the final time step through a forward GRU layer, and from the final time step to the initial time step through a backward GRU layer, capturing the forward trend and backward correlation patterns of coal prices over time, generating time-series coal price pattern data. This data clearly reflects the fluctuation characteristics and change logic of coal prices in different time periods. Based on the extracted coal price time-series pattern data, the model further extrapolates trends, combining the influence of historical price change patterns and current fused feature data to determine the fluctuation range and direction of coal prices in the future, achieving accurate prediction of coal prices.

[0047] Finally, the prediction results are output and validated. After completing the trend extrapolation, the BiGRU model outputs coal price prediction trend data, which includes the predicted price, price fluctuation range, and trend direction at different future time points. To ensure prediction accuracy, the prediction results are compared and validated with historical actual price data, and the prediction error is calculated. If the error exceeds the preset range, the model parameters are readjusted, feature data is added and integrated, and training and prediction are performed again until the prediction results meet the accuracy requirements.

[0048] Figure 2 This is a flowchart illustrating the coal price forecasting process. The process takes coal quality index data and standardized coal market data as input. First, it fuses price impact features, achieving a weighted fusion of coal quality attributes and market conditions. Then, it extracts time-series variation patterns to capture the two-way time dependence and fluctuation patterns of price data. Next, it extrapolates trends based on the extracted patterns, outputting predicted coal price trends. Simultaneously, the process includes a feedback loop to iteratively optimize the feature fusion and pattern extraction processes based on the forecast results, ensuring the accuracy of the forecasts and providing reliable price trend data for subsequent coal decision-making reports.

[0049] This embodiment, through refined feature extraction, weighted calculation, and fusion processing, fully mines key information affecting prices from coal quality and market data, highlighting the role of core features and solving the problems of large prediction deviations from single data and insufficient feature utilization. The combination of time-series pattern extraction and trend extrapolation improves the accuracy of price prediction, making the generated coal price prediction trend data more reliable. The overall price prediction process is optimized, providing high-quality support for the generation of subsequent coal decision-making reports, helping users accurately analyze price trends, rationally formulate procurement strategies, mitigate market risks, reduce operating costs, and ensure stable business operations.

[0050] S104: Perform knowledge vectorization processing on standardized coal information text data to obtain multiple knowledge vectors; calculate the similarity between the target retrieval vector and each knowledge vector, and select the knowledge vector corresponding to the maximum similarity.

[0051] In this embodiment, standardized coal information text data is processed into knowledge vectors to obtain multiple knowledge vectors, including: Standardized coal information text data is segmented into blocks to obtain segmented text data. Feature extraction is performed on the segmented text data to obtain text feature data; Vector encoding is performed on the text feature data to obtain multiple knowledge vectors.

[0052] In this embodiment, knowledge vectorization is the operation of converting standardized coal information text into vectors that can be used for similarity matching. A knowledge vector is a vector representing the core knowledge of the text after knowledge vectorization. Similarity calculation is the operation of calculating the degree of association between the target retrieval vector and each knowledge vector. Chunking is the operation of splitting long text into short segments; the segmented text data is the short text segment obtained after splitting, for example, splitting a coal policy information article into short segments such as policy content and impact analysis. Feature extraction is the operation of extracting the core information of the segmented text; the text feature data is the extracted core text features, and vector encoding is the operation of converting text features into vectors.

[0053] For example, this embodiment performs block processing on standardized coal information text data. Long texts are split into fixed-length short segments according to semantic logic, ensuring that each segment retains complete semantics and avoiding the fragmentation of core information, resulting in block-based text data. Then, feature extraction is performed on the block-based text data to extract core features related to coal prices, market supply and demand, and industry policies from each segment, while removing irrelevant and redundant information, resulting in text feature data. Next, the text feature data is vector-encoded. Through text segmentation, feature mapping, and other processing, the text feature data of each segment is transformed into vectors with uniform dimensions, resulting in multiple knowledge vectors. Finally, the target retrieval vector is obtained, and the similarity between the target retrieval vector and each knowledge vector is calculated based on a preset algorithm. By comparing all similarity values, the knowledge vector corresponding to the highest similarity is selected.

[0054] Specifically, in this embodiment, the similarity calculation is the similarity calculation result between the target retrieval vector and the i-th knowledge vector (i.e., each knowledge vector). First, the first part of the value is calculated, namely the basic correlation between the target retrieval vector and the i-th knowledge vector. This part is obtained by dividing the dot product of the target retrieval vector and the i-th knowledge vector by the product of the magnitude of the target retrieval vector and the magnitude of the i-th knowledge vector. The dot product is the result of summing the product of the corresponding dimensions of the two vectors, and the magnitude is the result of the square root of the sum of the squares of the elements of each dimension of the vector. This part is used to characterize the degree of linear correlation between the two vectors. Second, the second part of the value is calculated, namely the vector feature distribution difference correction term. The vector feature distribution difference is the degree of dispersion of the values ​​of each dimension of the vector (i.e., the fluctuation range of each dimension feature value). The standard deviation is used as its quantitative index. The larger the standard deviation, the more dispersed the values ​​within the vector are; the smaller the standard deviation, the more concentrated they are. First, calculate the feature standard deviations of the target retrieval vector and the i-th knowledge vector separately. Then, calculate the absolute difference between the two standard deviations. Divide this difference by the difference between the maximum and minimum feature standard deviations of all knowledge vectors and the target retrieval vector. Subtract 1 from the result to obtain this part of the value. The purpose of this correction term is to penalize situations where the standard deviations of two vectors differ significantly, indicating a mismatch in their numerical dispersion, which can easily affect semantic consistency or information content. This penalty is used in similarity calculation to compensate for the traditional calculation's neglect of differences in vector feature distribution. Next, determine the third part of the value, namely the coal domain adaptation correction term. This part is the value of the domain adaptation correction coefficient, which is dynamically adjusted according to the feature type of the coal information text, ranging from 0.8 to 1.2. Vectors involving core features such as coal prices and industry policies take higher values, while vectors involving secondary features take lower values, to strengthen the influence of core coal domain features on similarity calculation. Finally, the above three parts of the value are weighted and summed to obtain the final similarity calculation result. The weight coefficients of the first part of the numerical values ​​range from 0.5 to 0.7, the weight coefficients of the second part of the numerical values ​​range from 0.2 to 0.3, and the weight coefficients of the third part of the numerical values ​​range from 0.1 to 0.2. The sum of the three weight coefficients is 1. The result obtained after weighted summation is the similarity between the target retrieval vector and the i-th knowledge vector.

[0055] This embodiment addresses the problems of low utilization efficiency and inaccurate retrieval matching in standardized coal information text. Block processing and feature extraction focus on core text information, removing redundant interference and improving text processing efficiency. Vector encoding transforms text into knowledge vectors, quantifying textual knowledge and facilitating similarity calculation. Similarity filtering yields the knowledge vectors most relevant to user needs, ensuring subsequent decision reports align with user requirements, improving the relevance and reliability of decisions, reducing interference from invalid information, providing high-quality information support for intelligent coal decision-making, and meeting the needs of refined enterprise operations.

[0056] S105: Generate a coal decision report based on the knowledge vector corresponding to the maximum similarity and coal price prediction trend data.

[0057] In this embodiment, a coal decision report is generated based on the knowledge vector corresponding to the maximum similarity and coal price prediction trend data, including: Knowledge retrieval and matching are performed based on the knowledge vectors corresponding to the maximum similarity to obtain the target standardized coal information text data. By integrating standardized coal information text data with coal price forecast trend data, decision-making information is obtained, and basic data for decision analysis is derived. A comprehensive decision analysis is conducted on the basic data for decision analysis to obtain a coal decision report.

[0058] In this embodiment, the standardized coal information text data and coal price forecast trend data are fused to obtain basic data for decision analysis, including: Contextual structuring is performed on the standardized coal information text data to obtain decision context data; decision element data of the decision context data is then determined. The coal price forecast trend data is processed to obtain price trend basis data; the price trend element data of the price trend basis data is determined. Based on the feature splicing process of decision element data and price trend element data, the basic data for decision analysis is obtained.

[0059] In this embodiment, knowledge retrieval matching is the operation of matching corresponding text through knowledge vectors. The target standardized coal information text data is the standardized information text most relevant to user needs, such as information text aligned with the analysis of lean coal prices in region A. Decision information fusion is the operation of integrating two types of core data. Context structuring processing is the operation of sorting out the text context logic and standardizing the format; decision context data is the processed text context information, and decision element data is the core information in the text that affects decision-making. Trend structuring processing is the operation of standardizing the format of price trend data; price trend basis data is the processed price trend information, and price trend element data is the core price information in the trend.

[0060] This embodiment integrates two core data sets: coal information and price forecasts. Through standardized processing and fusion, it generates scientific and targeted coal decision-making reports, providing reliable support for users' procurement and other decisions. The knowledge vectors corresponding to the highest similarity are associated with the information most relevant to user needs, while coal price forecast trend data reflects price change patterns. The combination of these two provides dual assurance of information and data support. Target information is obtained through knowledge retrieval and matching to ensure its relevance; structured processing organizes the core elements of both types of data, resolving issues of data clutter and unclear logic; and feature splicing achieves data fusion, ensuring that the decision-making report is targeted, comprehensive, and scientific, meeting the actual decision-making needs of users.

[0061] In this embodiment, knowledge retrieval and matching are performed based on the knowledge vector corresponding to the maximum similarity. Through vector back mapping, the corresponding segmented text data is associated, and then the original standardized coal information text data is matched to obtain the target standardized coal information text data. Next, the target standardized coal information text data undergoes contextual structuring processing to organize the text logic and categorize the content according to policies, supply and demand, etc., to obtain decision context data. The core information influencing procurement decisions is then extracted to determine the decision element data. Simultaneously, coal price forecast trend data undergoes trend structuring processing to organize information such as price fluctuation ranges and change nodes, obtaining price trend basis data, and extracting core price information as price trend element data. Finally, the decision element data and price trend element data undergo feature concatenation processing to integrate the two types of core elements, obtaining the basic data for decision analysis. This data is then used for comprehensive decision analysis, combining professional knowledge in the coal industry to assess price trends and the impact of information, generating a coal decision report that includes price analysis and procurement recommendations.

[0062] In this embodiment, the specific process of fusing target standardized coal information text data with coal price forecast trend data to obtain basic data for decision analysis is as follows: First, the target standardized coal information text data is subjected to contextual structuring processing to obtain decision context data; the decision element data of the decision context data is then determined. Second, the coal price forecast trend data is subjected to trend structuring processing to obtain price trend basis data; the price trend element data of the price trend basis data is then determined. Finally, feature concatenation processing is performed based on the decision element data and the price trend element data to obtain basic data for decision analysis. In this process, the large model plays a core supporting role in contextual structuring, trend structuring, and feature splicing: On the one hand, the large model performs semantic analysis on the target standardized coal information text data, sorts out the contextual logic of the text, and organizes the text into structured modules such as policy guidance, market supply and demand, and industry dynamics according to the decision-making needs of the coal industry, forming decision context data. Then, it automatically extracts the core information that affects coal procurement and price judgment to determine the decision element data. On the other hand, the large model analyzes the coal price forecast trend data, organizes key information such as price fluctuation range, change nodes, and trend direction to form price trend basis data, and extracts core price information as price trend element data. Finally, through feature fusion capabilities, the large model performs standardized feature splicing of the decision element data and price trend element data to ensure that the two types of data are logically consistent and formatted uniformly, thus obtaining the basic data for decision analysis.

[0063] For example, in the knowledge retrieval and matching stage, the large model is used to achieve accurate mapping between the knowledge vector corresponding to the maximum similarity and the target standardized coal information text data. The large model first parses the knowledge vector corresponding to the maximum similarity, extracts the core text features represented by the vector, and then, through its own semantic matching capabilities, reverse-links it to the previously segmented text data, thereby matching it to the original standardized coal information text data. Finally, it selects the target standardized coal information text data that best matches the user's search needs. In this process, the large model effectively solves the problems of inaccurate mapping between knowledge vectors and text data and low matching efficiency, ensuring the relevance of the target standardized coal information text data and laying the foundation for subsequent decision-making information fusion.

[0064] In this embodiment, the large-scale model plays a core supporting role in the decision-making information fusion stage, undertaking contextual structuring, trend structuring, and feature splicing. On one hand, the large-scale model performs semantic analysis on the standardized coal information text data, sorting out the contextual logic of the text. According to the decision-making needs of the coal industry, it organizes the text into structured modules such as policy guidance, market supply and demand, and industry dynamics, forming decision context data. It then automatically extracts the core information affecting coal procurement and price judgment, determining the decision-making element data. On the other hand, the large-scale model analyzes coal price forecast trend data, organizing key information such as price fluctuation ranges, change points, and trend directions, forming price trend basis data, and extracting core price information as price trend element data. Finally, through its feature fusion capabilities, the large-scale model performs standardized feature splicing of the decision-making element data and price trend element data, ensuring logical coherence and format consistency between the two types of data, generating basic data for decision analysis.

[0065] In this embodiment, the large-scale model plays a core and leading role in the comprehensive decision analysis and coal decision report generation stages. Built with expertise in the coal industry, the model can perform in-depth logical analysis of the foundational data for decision analysis, examining the impact of standardized coal information text data (such as industry policy adjustments and market supply and demand changes) on coal price forecast trends. Combined with price fluctuation patterns, it forms scientific and rigorous analytical conclusions. Simultaneously, following the standardized format of coal decision reports, the model automatically generates a complete, professionally worded coal decision report tailored to the user's actual needs, including analytical conclusions, price trend predictions, procurement timing recommendations, and risk management measures.

[0066] In this embodiment, taking the coking coal procurement decision-making scenario as an example, knowledge retrieval and matching are performed based on the knowledge vector corresponding to the maximum similarity to obtain target standardized coking coal information text data. For example, matching industry information such as "the port inventory of prime coking coal continues to decline," "the recovery of coking plant operating rate drives the growth of procurement demand," and "changes in the customs clearance pace of imported coking coal" are obtained. The target standardized coking coal information text data is fused with coking coal price forecast trend data to obtain basic data for decision analysis. Comprehensive decision analysis is performed on the basic data for decision analysis to obtain a coking coal procurement decision report. The report may include coking coal price trend judgment, procurement timing suggestions, inventory optimization strategies, and risk warning prompts, providing support for steel companies' coking coal procurement decisions.

[0067] This embodiment achieves precise integration and in-depth utilization of standardized coal information text data and price forecast data. Knowledge retrieval and matching ensure that the information aligns with user needs, while structured processing and feature splicing address issues of data clutter and weak correlation, providing a high-quality foundation for comprehensive decision analysis. The comprehensive decision analysis, combining these two core data types, ensures that the generated coal decision report is both targeted and scientific, effectively avoiding the biases and inefficiencies of manual decision-making. This embodiment can provide users with accurate price analysis and feasible procurement decision suggestions, improving decision-making efficiency and reliability, helping enterprises to reasonably control costs, stabilize operations, and meet the needs of refined enterprise operations.

[0068] As can be concluded from the above, the intelligent decision-making generation method, apparatus, equipment, and storage medium based on coal provided in this application embodiment, compared with related technologies, simultaneously acquires coal information and coal market data based on user retrieval needs. Through standardized preprocessing, it achieves effective integration of multi-source data, avoiding the limitations of single data. Simultaneously, by combining coal quality indicators and standardized coal market data for price prediction, it can comprehensively consider various key factors affecting coal prices, improving the accuracy of price trend prediction and providing a reliable basis for enterprises' medium- and long-term procurement decisions. This application embodiment, by performing knowledge vectorization processing on standardized coal information text data and combining it with vector matching based on user retrieval needs, can quickly filter out the most relevant knowledge vectors, solving the problem of low utilization rate of unstructured data in existing technologies. Finally, by combining the matched knowledge vectors with coal price prediction trend data to generate a decision report, it can improve the level of intelligent decision-making, help enterprises formulate scientific procurement strategies, effectively control costs, stabilize operations, and adapt to the needs of refined enterprise operations.

[0069] Figure 3 The flowchart illustrating the generation process of the coal decision report provided in this application embodiment is as follows: First, coal information is collected, preprocessed, and stored in a knowledge base. Simultaneously, coal market data and coal quality indicator data are collected, preprocessed, and stored in a database. Then, standardized coal market data and standardized coal quality indicator data from the database are input into the coal price forecasting stage to generate coal price forecast trend data. Finally, the coal price forecast trend data and standardized coal information text data from the knowledge base are input into the comprehensive decision analysis stage. After decision information fusion and comprehensive judgment, a coal decision report is generated, providing users with price trend analysis and procurement decision support.

[0070] Corresponding to the coal-based intelligent decision generation method in the above embodiments, Figure 4 This is a structural block diagram of a coal-based intelligent decision generation device according to an embodiment of this application. For ease of explanation, only the parts relevant to the embodiment of this application are shown. References Figure 4 The coal-based intelligent decision generation device 20 may specifically include: a retrieval demand acquisition module 21, a preprocessing module 22, a price prediction module 23, a matching module 24, and a decision report generation module 25.

[0071] Among them, the search demand acquisition module 21 is used to acquire user search demands, which are inquiries about price trend analysis and procurement decision suggestions for coal in at least one dimension of region, variety and time. The preprocessing module 22 is used to obtain coal information and coal market data based on user search requirements; perform text preprocessing on the coal information to obtain standardized coal information text data; perform data preprocessing on the coal market data to obtain standardized coal market data; and perform text vector encoding on the user search requirements to obtain the target search vector. Price forecasting module 23 is used to forecast coal prices based on coal quality index data and standardized coal market data, and obtain coal price forecast trend data; the coal quality index data is a comprehensive indicator that characterizes the physical and chemical properties and industrial applicability of coal. Matching module 24 is used to perform knowledge vectorization processing on standardized coal information text data to obtain multiple knowledge vectors; based on the similarity calculation between the target retrieval vector and each knowledge vector, the knowledge vector corresponding to the maximum similarity is selected; The decision report generation module 25 is used to generate a coal decision report based on the knowledge vector corresponding to the maximum similarity and coal price prediction trend data.

[0072] In one embodiment of this application, the preprocessing module 22 is specifically used for: Text cleaning is performed on coal information to obtain cleaned coal information text data. The cleaned coal information text data is deduplicated to obtain the deduplicated coal information text data. Key information is extracted from the deduplicated coal information text data to obtain structured coal information text data. Text normalization is performed on structured coal information text data to obtain standardized coal information text data.

[0073] In one embodiment of this application, the price prediction module 23 is specifically used for: Price impact characteristics are fused based on coal quality index data and standardized coal market data to obtain fused characteristic data. By extracting the temporal variation patterns from the fused feature data, we can obtain the temporal pattern data of coal prices. By performing trend extrapolation on the time-series data of coal prices, we can obtain predicted trend data for coal prices.

[0074] In one embodiment of this application, the price prediction module 23 is specifically used for: Quality attribute features are extracted from coal quality index data to obtain coal quality attribute feature data; the first price influence degree of coal quality attribute feature data is determined, and the coal quality attribute feature data is weighted based on the first price influence degree to obtain weighted coal quality attribute feature data. Market characteristics are extracted from standardized coal market data to obtain coal market characteristic data; the second price influence of the coal market characteristic data is determined, and the coal market characteristic data is weighted based on the second price influence to obtain weighted coal market characteristic data. The weighted coal quality attribute feature data and the weighted coal market condition feature data are combined to obtain fused feature data.

[0075] In one embodiment of this application, the matching module 24 is specifically used for: Standardized coal information text data is segmented into blocks to obtain segmented text data. Feature extraction is performed on the segmented text data to obtain text feature data; Vector encoding is performed on the text feature data to obtain multiple knowledge vectors.

[0076] In one embodiment of this application, the decision report generation module 25 is specifically used for: Knowledge retrieval and matching are performed based on the knowledge vectors corresponding to the maximum similarity to obtain the target standardized coal information text data. By integrating standardized coal information text data with coal price forecast trend data, decision-making information is obtained, and basic data for decision analysis is derived. A comprehensive decision analysis is conducted on the basic data for decision analysis to obtain a coal decision report.

[0077] In one embodiment of this application, the decision report generation module 25 is specifically used for: Contextual structuring is performed on the standardized coal information text data to obtain decision context data; decision element data of the decision context data is then determined. The coal price forecast trend data is processed to obtain price trend basis data; the price trend element data of the price trend basis data is determined. Based on the feature splicing process of decision element data and price trend element data, the basic data for decision analysis is obtained.

[0078] It should be noted that the specific limitations of the coal-based intelligent decision generation device 20 embodiments provided above can be found in the limitations of the coal-based intelligent decision generation method above, and will not be repeated here. Each module of the above device can be implemented entirely or partially through software, hardware, or a combination thereof. Each module can be embedded in the processor of the electronic device in hardware form or independent of the processor, or it can be stored in the memory of the electronic device in software form, so that the processor can call and execute the corresponding operations of each module.

[0079] This application provides an electronic device, including a memory, a processor, and a computer program stored in the memory. The processor executes the computer program to implement the steps of the method provided in any optional embodiment of this application.

[0080] In one alternative embodiment, an electronic device is provided, such as Figure 5 As shown, Figure 5 The illustrated electronic device 300 includes a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, for example, via a bus 302. Optionally, the electronic device 300 may further include a transceiver 304, which can be used for data interaction between the electronic device and other electronic devices, such as sending and / or receiving data. It should be noted that in practical applications, the transceiver 304 is not limited to one type, and the structure of this electronic device 300 does not constitute a limitation on the embodiments of this application.

[0081] Processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 301 may also be a combination that implements computational functions, such as including one or more microprocessor combinations, a combination of a DSP and a microprocessor, etc.

[0082] Bus 302 may include a pathway for transmitting information between the aforementioned components. Bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 302 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0083] The memory 303 may be ROM (Read Only Memory) or other types of static storage devices capable of storing static information and instructions, RAM (Random Access Memory) or other types of dynamic storage devices capable of storing information and instructions, or EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium capable of carrying or storing computer programs and capable of being read by a computer, without limitation herein.

[0084] The memory 303 is used to store computer programs that execute the embodiments of this application, and the execution is controlled by the processor 301. The processor 301 is used to execute the computer programs stored in the memory 303 to implement the steps shown in the foregoing method embodiments.

[0085] In an exemplary embodiment, a computer-readable storage medium is also provided, which stores at least one computer program that is loaded and executed by a processor of a computer device to enable the computer to implement any of the above-described intelligent decision generation methods based on coal.

[0086] In one possible implementation, the aforementioned computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a solid-state drive (SSD), a compact disc read-only memory (CD-ROM), magnetic tape, a floppy disk, and an optical data storage device, etc. The random access memory can include resistive random access memory (ReRAM) and dynamic random access memory (DRAM).

[0087] In an exemplary embodiment, a computer program or computer program product is also provided, which includes computer instructions loaded and executed by a processor to enable a computer to implement any of the above-described intelligent decision generation methods based on coal.

[0088] It should be noted that all information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.), and signals involved in this application have been authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data must comply with the relevant laws, regulations, and standards of the relevant countries and regions. For example, the coal information and coal market data involved in this application were obtained with full authorization.

[0089] In other words, the data collection and processing in this application should strictly comply with the requirements of relevant national laws and regulations, obtain the informed consent or separate consent of the personal information subject, and carry out subsequent data use and processing within the scope of laws and regulations and the authorization of the personal information subject.

[0090] It should be further noted that the terms "first," "second," etc., used in this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. The implementation methods described in the above exemplary embodiments do not represent all implementation methods consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application.

[0091] It should be understood that "multiple" as used in this article refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following related objects have an "or" relationship.

[0092] Furthermore, the step numbers described herein are merely illustrative of one possible execution order between steps. In some other embodiments, the steps may not be executed in the order of their numbers, such as two steps with different numbers being executed simultaneously, or two steps with different numbers being executed in the reverse order of the illustration. This application does not limit this.

[0093] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. Optionally, the program is stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0094] The above description is merely an exemplary embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.

Claims

1. A method for generating intelligent decisions based on coal, characterized in that, include: Obtain user search requests, which are inquiries about price trend analysis and procurement decision suggestions for coal in at least one dimension of region, variety, and time. Based on the user's search requirements, coal information and coal market data are obtained; the coal information is preprocessed to obtain standardized coal information text data; the coal market data is preprocessed to obtain standardized coal market data; and the user's search requirements are encoded using text vectors to obtain the target search vector. Coal price forecasting is performed based on coal quality index data and the standardized coal market data to obtain coal price forecast trend data; the coal quality index data is a comprehensive indicator characterizing the physical and chemical properties and industrial applicability of coal. The standardized coal information text data is processed into knowledge vectors to obtain multiple knowledge vectors; the similarity between the target retrieval vector and each knowledge vector is calculated, and the knowledge vector with the highest similarity is selected. A coal decision report is generated based on the knowledge vector corresponding to the maximum similarity and the coal price prediction trend data.

2. The intelligent decision generation method based on coal as described in claim 1, characterized in that, The text preprocessing of the coal information to obtain standardized coal information text data includes: The coal information is cleaned to obtain cleaned coal information text data; The cleaned coal information text data is deduplicated to obtain deduplicated coal information text data. Key information is extracted from the deduplicated coal information text data to obtain structured coal information text data. The structured coal information text data is normalized to obtain standardized coal information text data.

3. The intelligent decision generation method based on coal as described in claim 1, characterized in that, The coal price forecasting based on coal quality index data and standardized coal market data yields coal price forecast trend data, including: Based on the coal quality index data and the standardized coal market data, price impact characteristics are fused to obtain fused characteristic data; The time-series variation patterns of the fused feature data are extracted to obtain time-series pattern data of coal prices; By performing trend extrapolation on the time-series data of coal prices, we can obtain predicted trend data for coal prices.

4. The intelligent decision generation method based on coal as described in claim 3, characterized in that, The process of fusing price impact features based on the coal quality index data and the standardized coal market data to obtain fused feature data includes: The coal quality index data is subjected to quality attribute feature extraction to obtain coal quality attribute feature data; the first price influence degree of the coal quality attribute feature data is determined; and the coal quality attribute feature data is weighted based on the first price influence degree to obtain weighted coal quality attribute feature data. Market characteristics are extracted from the standardized coal market data to obtain coal market characteristic data; a second price influence degree of the coal market characteristic data is determined, and the coal market characteristic data is weighted based on the second price influence degree to obtain weighted coal market characteristic data. The weighted coal quality attribute feature data and the weighted coal market condition feature data are subjected to feature splicing processing to obtain fused feature data.

5. The intelligent decision generation method based on coal as described in claim 1, characterized in that, The standardized coal information text data is subjected to knowledge vectorization processing to obtain multiple knowledge vectors, including: The standardized coal information text data is divided into blocks to obtain the block-based text data; Feature extraction is performed on the segmented text data to obtain text feature data; The text feature data is vector-encoded to obtain multiple knowledge vectors.

6. The intelligent decision generation method based on coal as described in claim 1, characterized in that, The coal decision report is generated based on the knowledge vector corresponding to the maximum similarity and the coal price prediction trend data, including: Based on the knowledge vector corresponding to the maximum similarity, knowledge retrieval and matching are performed to obtain the target standardized coal information text data. The standardized coal information text data and the coal price forecast trend data are fused together to obtain the basic data for decision analysis. A comprehensive decision analysis is performed on the aforementioned basic data to obtain a coal decision report.

7. The intelligent decision generation method based on coal as described in claim 6, characterized in that, The process of fusing the standardized coal information text data with the coal price forecast trend data to obtain basic data for decision analysis includes: The target standardized coal information text data is subjected to contextual structuring processing to obtain decision context data; the decision element data of the decision context data is then determined. The coal price forecast trend data is subjected to trend structuring processing to obtain price trend basis data; the price trend element data of the price trend basis data is determined. The decision analysis base data is obtained by performing feature concatenation processing on the decision element data and the price trend element data.

8. A coal-based intelligent decision generation device, characterized in that, include: The search requirement acquisition module is used to acquire user search requirements, which are inquiries about price trend analysis and procurement decision suggestions for coal in at least one dimension of region, variety, and time. The preprocessing module is used to obtain coal information and coal market data based on the user's search requirements. The coal information is preprocessed to obtain standardized coal information text data; the coal market data is preprocessed to obtain standardized coal market data; and the user's search request is encoded into a target search vector using text vector encoding. The price forecasting module is used to forecast coal prices based on coal quality index data and the standardized coal market data, and to obtain coal price forecast trend data; the coal quality index data is a comprehensive indicator characterizing the physical and chemical properties and industrial applicability of coal. The matching module is used to perform knowledge vectorization processing on the standardized coal information text data to obtain multiple knowledge vectors; based on the target retrieval vector and each knowledge vector, the similarity calculation is performed to select the knowledge vector corresponding to the maximum similarity. The decision report generation module is used to generate a coal decision report based on the knowledge vector corresponding to the maximum similarity and the coal price prediction trend data.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the coal-based intelligent decision generation method according to any one of claims 1 to 7 when running the computer program.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the coal-based intelligent decision generation method according to any one of claims 1 to 7.

Images