Geographic information mining method and system based on big data

Abstract

The invention discloses a geographic information mining method and system based on big data, and relates to the technical field of geographic information mining. A geographic information mining system based on big data comprises a geographic area selection module, a geographic information mining module and a geographic information integration module. According to the method, the conditional geographic information, the regional geographic information and the social geographic information are integrated in breadth by combining the conditional geographic information, the regional geographic information and the social geographic information through technologies such as semantic analysis, space-time consistency analysis, gray scale simulation calculation and geographic relation reasoning, multi-source heterogeneous data are covered, and cross-domain comprehensive coverage is realized; in the aspect of depth, technologies such as semantic analysis, space-time consistency analysis, gray scale simulation calculation and geographic relation reasoning are adopted, fine processing is conducted on the data, the space-time change, relevance and influence factors of geographic information are deeply mined, and the hierarchical expression ability of the data is improved.

Description

Technical Field

[0001] The present invention relates to the field of geographic information mining technology, and in particular to a geographic information mining method and system based on big data. Background Art

[0002] In geographic information mining, data from different sources often have different descriptions of the same place name or region, which can easily lead to data integration errors; for example, the same place name may point to different spatial ranges, resulting in a misalignment between geographic information and actual geographical locations, which in turn affects the assessment of the level of regional geographical development; in addition, the meaning of place names may change dynamically over time, but existing geographic information mining methods find it difficult to effectively distinguish geographic entities with the same name but different locations, and it is also difficult to track changes in administrative boundaries and resolve naming conflicts in a multilingual environment in a timely manner. These problems reduce the credibility of geographic information. Current mining methods often focus on the breadth of coverage of geographic information but ignore its accuracy, which affects subsequent decisions based on geographic information.

[0003] At the same time, existing methods often find it difficult to perform effective regional matching when dealing with geographic information that does not have a clear spatial orientation, and lack an in-depth understanding of geographic information; in the context of dynamic changes in space and time, existing methods find it difficult to screen and optimize data in combination with the current social context and the process of space-time evolution. Summary of the invention

[0004] The present invention aims to provide a geographic information mining method and system based on big data, to improve the breadth and depth of geographic information collection, and to ensure the accuracy of geographic information mining and matching.

[0005] A geographic information mining method based on big data comprises the following steps: For the geographic area to be mined, obtain unique geospatial codes, geographic multi-dimensional attribute labels, and geo-social characteristic labels; Search based on the geographic area to be mined, and collect several pieces of multi-breadth geographic information from multiple data sources; Perform geographic condition screening on a number of multi-breadth geographic information to obtain conditional geographic information; Search based on geographic multi-dimensional attribute tags to collect multiple pieces of multi-depth geographic information from multiple data sources; Performing geographic area screening on a plurality of multi-depth geographic information to obtain regional geographic information; Search based on geo-social characteristic tags to obtain several multi-factor geographic information; Screening several pieces of multi-factor geographic information by geographic factors to obtain social geographic information; The conditional geographic information, regional geographic information and social geographic information are integrated to obtain the mining geographic information set that conforms to the geographic area to be mined; by constructing a geographic information semantic output model, the conditional geographic information, regional geographic information and social geographic information are divided into text features, fused with grayscale calculations, and the text is reorganized based on geographic semantic output rules, and finally the mining geographic information set is output after integration.

[0006] As a preferred technical solution of the present invention, the specific steps of integrating conditional geographic information, regional geographic information and social geographic information include: Construct geographic information semantic output model; In the geographic information semantic output model, the text feature segmentation is performed on the input text to obtain the text feature block K i ; i=1, 2, ..., I, I is the total number of data blocks for text feature division; For text feature block K i Perform fusion grayscale calculation to obtain the text semantic grayscale value Z i ; For text semantic gray value Z i The input text is reorganized with geographic semantic output rules to obtain mined geographic information; Among them, the input text is conditional geographic information, regional geographic information and social geographic information; The output mined geographic information is combined to obtain a mined geographic information set.

[0007] As a preferred technical solution of the present invention, the specific steps of performing geographical condition screening on a plurality of multi-breadth geographical information include: Extract text features from multi-breadth geographic information to obtain geographic information text features; Set the geographical conditions of the geographic area to be mined as N regional related dimension labels B n , n=1,2,…,N; Based on geographic information text features and region-related dimension labels B n Establish geographic topological relationship matrix; For each characteristic edge in the geographic topological relationship matrix, the edges that do not conform to the m regional related dimension labels B are removed. n characteristic edges, and obtain a new geographic topological relationship matrix; If the structural integrity of the new geographic topological relationship matrix is ​​greater than the preset value, the multi-breadth geographic information is retained to complete the screening; All multi-breadth geographic information is traversed and screened, and the retained multi-breadth geographic information is combined to obtain conditional geographic information.

[0008] As a preferred technical solution of the present invention, the specific steps of performing geographic area screening on a plurality of multi-depth geographic information include: Using genetic algorithm to set geographic region weight matrix; Establish a spatiotemporal knowledge graph for multi-depth geographic information and obtain a geographic information reasoning graph; Establishing geographic entity models for multi-depth geographic information to obtain simulated geographic entities; On the basis of simulating geographic entities, the geographic information reasoning map and the geographic area weight matrix are used for regional matching. The improved graph neural network is used to couple the geographic information reasoning map and the simulated geographic entities to obtain the geographic information matching degree. The trained geographic uncertainty quantification model is used in combination with a preset matching threshold to screen the geographic information matching degree and determine whether to retain multi-depth geographic information; All multi-depth geographic information is traversed and screened, and the retained multi-depth geographic information is combined to obtain regional geographic information.

[0009] As a preferred technical solution of the present invention, the specific steps of performing geographical factor screening on a plurality of multi-factor geographical information include: Construct sentiment maps for multi-factor geographic information and generate positive and negative geographic information features; Analyze the emotional causes of positive and negative geographic information features to obtain positive and negative geographic keywords; Capture the periodic fluctuation characteristics of multi-factor geographic information and obtain social geographic characteristics with time series labels; Generate a multi-factor geographic matrix by combining positive geographic keywords, negative geographic keywords, and social geographic features; The dynamic weights of the multi-factor geographic matrix are adjusted based on the real-time geographic region constraint indicators to obtain an updated multi-factor geographic matrix; Calculate and update the comprehensive matching degree of the multi-factor geographic matrix, and select whether to retain the multi-factor geographic information based on the data interval of the comprehensive matching degree; All multi-factor geographic information is traversed and screened, and the retained multi-factor geographic information is combined to obtain social geographic information.

[0010] As a preferred technical solution of the present invention, keyword feature extraction is performed based on the existing geographic information set of the geographic area to be mined to obtain geographic multi-dimensional attribute labels; data extraction is performed on the real-time social attributes of the geographic area to be mined to obtain geographic social characteristic labels; A geographic information mining system based on big data, comprising: The geographic region selection module includes a region determination unit for obtaining a unique geographic space code, a geographic multi-dimensional attribute label, and a geographic social characteristic label for the geographic region to be mined; Geographic information mining module, including conditional mining unit, regional mining unit and social mining unit; The conditional mining unit is used to search based on the geographic area to be mined, collect a number of multi-breadth geographic information from multiple data sources; perform geographic condition screening on the multiple multi-breadth geographic information to obtain conditional geographic information; The regional mining unit is used to search based on geographical multi-dimensional attribute tags, collect multiple multi-depth geographical information from multiple data sources, and perform geographical region screening on multiple multi-depth geographical information to obtain regional geographical information; The social mining unit is used to search based on geographical social characteristic tags to obtain a number of multi-factor geographical information; the geographical factors of a number of multi-factor geographical information are screened to obtain social geographical information; The geographic information integration module includes a data integration unit, which is used to integrate the conditional geographic information, regional geographic information and social geographic information to obtain a mining geographic information set that conforms to the geographic area to be mined; by constructing a geographic information semantic output model, the conditional geographic information, regional geographic information and social geographic information are divided into text features, fused with grayscale calculations, and the text is reorganized based on geographic semantic output rules, and finally the mining geographic information set is output after integration.

[0011] The present invention has the following advantages: The present invention combines conditional geographic information, regional geographic information and social geographic information, and constructs a high-quality collection of mined geographic information through technologies such as semantic analysis, spatiotemporal consistency analysis, simulated grayscale calculation, and geographic relationship reasoning. It can improve the accuracy of geographic information integration, enhance data consistency, promote multi-source data fusion, and improve data utilization. In terms of breadth, it integrates conditional geographic information, regional geographic information and social geographic information, covers multi-source heterogeneous data, and achieves comprehensive cross-domain coverage. In terms of depth, it uses technologies such as semantic analysis, spatiotemporal consistency analysis, simulated grayscale calculation, and geographic relationship reasoning to refine the data, deeply explore the spatiotemporal changes, correlations and influencing factors of geographic information, and improve the hierarchical expression ability of data.

[0012] The present invention performs regional matching by using geographic information reasoning graphs and geographic area weight matrices, and uses an improved graph neural network to perform coupling analysis on geographic information reasoning graphs and simulated geographic entities to obtain geographic information matching degree; retrieves multi-depth geographic information that meets specific terrain conditions from multiple data sources through geographic multi-dimensional attribute labels, and obtains geographic data that meets regional characteristics through screening; sets the geographic area weight matrix through a genetic algorithm to make the screening process more targeted and ensure that regional information is highly matched with demand; combines spatiotemporal knowledge graph reasoning to improve the logical consistency and temporal correlation of geographic information; establishes a geographic entity model to simulate the real geographic environment and enhance the interpretability of data; uses an improved graph neural network for coupling analysis to optimize the intelligent matching of geographic data and improve the depth and efficiency of analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] Figure 1 A schematic diagram of the structure of a geographic information mining system based on big data adopted in an embodiment of the present invention.

[0014] Figure 2 A Loss diagram of a model training in a geographic information mining method based on big data adopted in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to enable persons skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0016] Example 1, a geographic information mining method based on big data, see Figure 2 As shown, the following steps are included: For the geographic area to be mined, obtain unique geospatial codes, geographic multi-dimensional attribute labels, and geo-social characteristic labels; Determine the geographic administrative code according to the geographical location of the geographic area to be mined; determine the geographic hash value according to the natural coordinates of the geographic area to be mined; determine the geographic terrain code according to the basic terrain of the geographic area to be mined; combine the geographic administrative code, the geographic hash value and the geographic terrain code to obtain a unique geographic space code; Extract keyword features based on the existing geographic information set of the geographic area to be mined to obtain geographic multi-dimensional attribute labels; extract data on the real-time social attributes of the geographic area to be mined to obtain geographic social characteristic labels; Unique geospatial coding needs to combine multiple key attributes of a geographic area, including administrative codes, geohash values, and terrain codes, to form a complete geographic identification. For example, standard administrative codes are obtained based on the administrative divisions of the area to be mined, such as provinces, cities, districts / counties, etc. Geohash is a coding method based on longitude and latitude, which can be used to accurately locate specific geographic areas. Based on the basic terrain of the area to be mined, such as mountains, plains, hills, rivers, lakes, etc., standard terrain codes are determined, and mountains are coded as MNT, plains are coded as PLN, and rivers are coded as RVR, etc. By combining these geomorphic feature codes and combining the spatial location of the area or other auxiliary information such as coordinates, grid numbers, etc., a unique geospatial code can be generated, thereby achieving unique identification of each area, facilitating data retrieval and spatial analysis. Geographic multi-dimensional attribute tags need to extract geographic feature information from multiple perspectives, including climate characteristics, vegetation cover, land use, water resource distribution, geological information, etc.; geographic social characteristic tags mainly reflect the attributes of a place's population, economy, culture, infrastructure and social activities, mainly including population characteristics, economic development, cultural characteristics, traffic conditions, social vitality, etc.; through a unified coding method, the management efficiency of geographic data can be improved, and seamless connection of different data sources can be achieved; combined with unique geographic coding, geographic areas can be quickly indexed through coding, which is suitable for map services, geographic analysis and other fields; combined with geographic attributes and social characteristics, applications such as urban planning, environmental monitoring, traffic optimization, and commercial site selection can be carried out; Search based on the geographic area to be mined, and collect several pieces of multi-breadth geographic information from multiple data sources; Perform geographic condition screening on a number of multi-breadth geographic information to obtain conditional geographic information; The specific steps for geographical condition screening of multiple pieces of multi-breadth geographical information include: Extract text features from multi-breadth geographic information to obtain geographic information text features; Set the geographical conditions of the geographic area to be mined as N regional related dimension labels B n , n=1,2,…,N; Based on geographic information text features and region-related dimension labels B n Establish geographic topological relationship matrix; For each characteristic edge in the geographic topological relationship matrix, the edges that do not conform to the m regional related dimension labels B are removed. n The characteristic edges of the new geographic topological relationship matrix are obtained; the specific value of m is set by professional and technical personnel according to the actual situation; If the structural integrity of the new geographic topological relationship matrix is ​​greater than the preset value, the multi-breadth geographic information is retained to complete the screening; the preset value is set by professional technicians according to actual conditions; Traverse and filter all multi-width geographic information, combine the retained multi-width geographic information, and obtain conditional geographic information; Geographic information is collected through web crawlers, open geographic databases, statistical data, remote sensing data, etc. Since geographic areas with the same name may exist, it is necessary to determine whether they meet the target location based on information such as longitude and latitude, administrative divisions, etc. The multi-breadth geographic information screening method based on the geographic area to be mined has wide application value and significant advantages; by constructing a unique geographic spatial code, extracting geographic multi-dimensional attribute labels, analyzing geographic social characteristic labels, and using a topological relationship matrix to screen geographic information, efficient and accurate geographic information matching and integration can be achieved; Through administrative divisions, longitude and latitude matching, geographic hash value comparison and other methods, the information confusion problem caused by the same place name can be effectively solved; for example, "West Lake" exists in multiple provinces and cities, but through longitude and latitude matching, the West Lake in Hangzhou can be accurately screened out; by constructing a geographic topological relationship matrix, data that does not match the characteristics of the target area can be screened out, thereby improving the reliability of the screening results; data from different sources have different formats, and this method uses unified text feature extraction + topological relationship matrix modeling to improve data comparability and facilitate fusion analysis; when geographic information increases, the matrix can be dynamically expanded, which is suitable for the storage and analysis of massive geographic data and improves scalability; combining geographic spatial characteristics, climate information, economic development, traffic conditions, population structure and other data, it provides comprehensive regional analysis capabilities to support scientific decision-making; it can improve data integrity and is suitable for integrated analysis of massive geographic data; Search based on geographic multi-dimensional attribute tags to collect multiple pieces of multi-depth geographic information from multiple data sources; Performing geographic area screening on a plurality of multi-depth geographic information to obtain regional geographic information; The specific steps for performing geographic area screening for a number of multi-depth geographic information include: The genetic algorithm is used to set the geographical area weight matrix to optimize the weight combination of various geographical area attributes (such as transportation, population, and environment) to maximize the regional matching degree; The basic steps of setting the weight matrix of geographic regions using genetic algorithms are as follows: define multiple attribute weights of geographic regions, initialize the population of weight matrices, generate multiple candidate solutions by random or heuristic methods, randomly generate X individuals, and the weight vector of each individual satisfies the condition that the sum of its elements is 1; set the fitness function to measure the effect of different weight combinations on regional matching, usually based on historical data or preset goals to optimize regional division, apply the weight matrix to the validation set, and calculate the regional matching accuracy; use the selection, crossover and mutation operations of genetic algorithms to iteratively optimize the weight matrix so that it gradually tends to the optimal configuration; in each generation of iteration, select individuals with high fitness for genetic operations to retain excellent characteristics, while at the same time Introduce diversity through mutation to prevent local optimality; after multiple iterations, select the optimal weight matrix as the core parameter for geographical area screening and matching; at the same time, the genetic algorithm can be initialized by heuristic or hierarchical methods to make the initial population more representative and improve search efficiency; the selection mechanism can combine elite retention strategy with adaptive selection to give priority to inheritance of individuals with high fitness while maintaining population diversity; crossover and mutation strategies can adopt adaptive probability adjustment to make exploration the main focus in the early stage and convergence the main focus in the later stage, thereby improving global search capabilities; local search can be combined with methods such as simulated annealing and particle swarm optimization to improve the quality of local optimal solutions, or parallel computing can be used to accelerate algorithm convergence to improve the optimization performance of genetic algorithms on complex problems; Establish a spatiotemporal knowledge graph for multi-depth geographic information and obtain a geographic information reasoning graph; Establishing geographic entity models for multi-depth geographic information to obtain simulated geographic entities; On the basis of simulating geographic entities, regional matching is performed using geographic information reasoning graphs and geographic area weight matrices. The improved graph neural network is used to couple the geographic information reasoning graphs and simulated geographic entities to obtain the geographic information matching degree. In the improved graph neural network, the attention mechanism can be used to adaptively adjust the information propagation weights according to the importance of the nodes to improve the calculation accuracy of the association between different geographic entities. Alternatively, residual connections can be used to prevent the gradient vanishing problem in deep GNNs (graph neural networks) and improve the ability to extract deep features. Adaptive neighborhood aggregation can be combined to dynamically adjust the aggregation range according to node attributes and topological structures when calculating node representations to improve matching accuracy. Hierarchical graph convolution is used to extract multi-scale features of geographic information, so that GNN can effectively capture local and global geographic patterns. The trained geographic uncertainty quantification model is used in combination with a preset matching threshold to screen the geographic information matching degree and determine whether to retain multi-depth geographic information; Training geographic uncertainty quantification models usually uses methods such as Bayesian deep learning, ensemble learning or variational inference to quantify the uncertainty in the geographic information matching process. First, a geographic uncertainty estimation model based on graph neural network, Gaussian process or Bayesian neural network is constructed, and an uncertainty modeling mechanism such as MCDropout or deep ensemble regression is introduced. Then, during the training process, multiple sampling strategies are used to perturb the geographic information, and the uncertainty is measured by calculating the variance of the predicted distribution. In addition, adversarial training can be combined to enhance the model's robustness to outliers and data noise. Finally, methods such as confidence threshold or quantile regression are used to screen out reliable geographic information matching results and improve the accuracy and stability of regional screening. Traverse and filter all multi-depth geographic information, combine the retained multi-depth geographic information, and obtain regional geographic information; Through geographic multi-dimensional attribute tags, multi-depth geographic information that meets specific terrain conditions is retrieved from multiple data sources, and geographic data that meets regional characteristics is obtained through screening; the geographic region weight matrix is ​​set through genetic algorithms to make the screening process more targeted and ensure that regional information is highly matched with demand; combined with spatiotemporal knowledge graph reasoning, the logical consistency and temporal correlation of geographic information are improved; by establishing a geographic entity model, simulating the real geographic environment, enhancing the interpretability of data, using improved graph neural networks for coupling analysis, optimizing the intelligent matching of geographic data, and improving the depth and efficiency of analysis; the geographic entity model is a real-world geographic entity model. The digital abstract expression of geographical objects (such as mountains, rivers, cities, roads, etc.) integrates their spatial features, attribute features and spatiotemporal dynamic features through structured data models to form computable and inferable virtual entities. Its core goal is to enhance the interpretability of geographic data and provide a semantic basis for coupled analysis of graph neural networks. Geographic multi-dimensional attribute labels adopt a dynamic update mechanism, and the public database of the Ministry of Natural Resources is updated through crawlers. Social characteristic labels generate new keywords through hot topic clustering. If an old label has not been retrieved for L consecutive days, its weight will be automatically reduced to U% of the original weight. The specific values ​​of L and U are set by professional and technical personnel according to actual conditions. Combined with the geographic uncertainty quantification model, the matching degree of geographic information is screened to effectively deal with data noise and uncertainty, and improve the stability and reliability of the screening results; the matching threshold is set to ensure that the screened information meets the actual application needs and improve the practicality of regional information; through traversal screening and combination, geographic information of different depths and sources is integrated to form more complete and richer regional geographic information; Search based on geo-social characteristic tags to obtain several multi-factor geographic information; Screening several pieces of multi-factor geographic information by geographic factors to obtain social geographic information; The specific steps for geographical factor screening of several multi-factor geographical information include: Construct sentiment maps for multi-factor geographic information and generate positive and negative geographic information features; Analyze the emotional causes of positive and negative geographic information features to obtain positive and negative geographic keywords; Capture the periodic fluctuation characteristics of multi-factor geographic information and obtain social geographic characteristics with time series labels; Generate a multi-factor geographic matrix by combining positive geographic keywords, negative geographic keywords, and social geographic features; The dynamic weights of the multi-factor geographic matrix are adjusted based on the real-time geographic region constraint indicators to obtain an updated multi-factor geographic matrix; Calculate and update the comprehensive matching degree of the multi-factor geographic matrix, and select whether to retain the multi-factor geographic information based on the data interval of the comprehensive matching degree; In the update of the multi-factor geographic matrix, rows are set to represent geographic factors, columns to represent time slices or spatial units, and quantitative values ​​are extracted from multi-source data and standardized; weights are adjusted according to external dynamic factors, and the formula for calculating the matching degree is the sum of the dynamic weight ratios of the standardized values ​​of the matrix; data intervals are set for screening, such as dividing the high confidence interval, medium confidence interval, and low confidence interval for threshold judgment; Traverse and filter all multi-factor geographic information, combine the retained multi-factor geographic information, and obtain social geographic information; By constructing a sentiment map, we can distinguish between positive and negative geographic information features, and avoid the influence of irrelevant or misleading information on social geographic analysis. By combining positive and negative geographic keywords, we can accurately capture social hot spots, public opinion trends and environmental changes, and improve the effectiveness and representativeness of the data. By analyzing the causes of sentiment, we can identify the key factors that affect the changes in social geographical features, such as policy changes, economic development, environmental governance, etc., and improve the explanatory power of social geographical phenomena. This method can be used in urban planning, public safety, tourism management and other fields to help companies better understand social needs and public opinion trends. By capturing periodic fluctuation characteristics and extracting social geographical features with time series labels, the model not only focuses on the current state, but also identifies the impact of long-term trends and seasonal changes, and improves the foresight and timeliness of information screening. For example, in scenarios such as traffic flow forecasting and commercial site selection, this method can effectively identify the periodic changes in social geographical factors and optimize resource allocation and strategy formulation. For example, positive geographic information features are spatial information features that promote the research objectives, such as high population density, convenient public transportation, concentrated medical resources, rich educational resources, etc.; negative geographic information features are spatial information features that have an inhibitory effect on the research objectives or risk prompts, such as serious environmental pollution, backward infrastructure, frequent natural disasters, etc.; positive and negative feature classification is based on the BERT sentiment analysis model to avoid subjective factors judging positive and negative indicators and affecting the final decision results; The weights of the multi-factor geographic matrix are dynamically adjusted through real-time geographic region constraint indicators, so that the screening results can be updated in real time with dynamic factors such as policies, market demand, and environmental changes, avoiding the lag of static models; for example, in real estate market analysis, this method can adjust data screening standards in real time according to factors such as housing prices, population mobility, and infrastructure development, and improve the ability to respond to market trends; adopt comprehensive matching degree calculation, combined with multi-dimensional information such as social geographic characteristics, sentiment analysis, and time series changes, to improve the objectivity and accuracy of screening, avoid bias caused by subjective judgment, filter out low-matching information, and ensure that the final combination of social geographic information has high credibility and practical value, which is suitable for urban risk assessment, social stability monitoring, disaster emergency management and other fields; real-time geographic region constraint indicators can be set to update once every 30 minutes or other values, and receive data streams through the streaming computing framework; Integrate the conditional geographic information, regional geographic information and social geographic information to obtain a mining geographic information set that meets the geographic area to be mined; construct a geographic information semantic output model, divide the conditional geographic information, regional geographic information and social geographic information into text features, perform fusion grayscale calculations, reorganize the text based on geographic semantic output rules, and finally integrate and output the mining geographic information set; when the conditional geographic information, regional geographic information and social geographic information are integrated, if there is a contradiction among the three, a three-layer conflict resolution mechanism is used to deal with it, the first is the priority rule, the second is the spatiotemporal proximity weighted rule, and the third is the expert voting mechanism; The geographic information semantic output model is a multimodal data integration framework that combines natural language processing with geographic semantic rules. It aims to transform structured and unstructured geographic information into semantic spatial expressions that can be quantified and analyzed. The geographic information semantic output model is a model trained by machine learning algorithms, which is used to integrate text data of multi-source geographic information into a structured geographic information set according to semantic rules. Its core is to eliminate data heterogeneity through text feature division and semantic grayscale calculation, and generate standardized information that meets the needs of geographic analysis. The specific steps for data integration of conditional geographic information, regional geographic information and social geographic information include: Construct geographic information semantic output model; In the geographic information semantic output model, the text feature segmentation is performed on the input text to obtain the text feature block K i ; i=1, 2, ..., I, I is the total number of data blocks for text feature division; For text feature block K i Perform fusion grayscale calculation to obtain the text semantic grayscale value Z i ; For text semantic gray value Z i The input text is reorganized with geographic semantic output rules to obtain mined geographic information; Among them, the input text is conditional geographic information, regional geographic information and social geographic information; Combining the output mining geographic information to obtain a mining geographic information set; The specific implementation steps of multi-dimensional heterogeneous information integration and screening can be divided into three stages: data preprocessing, semantic modeling, and rule-driven dynamic fusion. First, by collecting multi-source data, including conditional geographic information, regional geographic information, social geographic information, and other related information, the collected multi-dimensional data, such as multi-breadth geographic information, economic data, life service data, social geographic characteristics, and other different dimensional data, are deduplicated based on rules or clustering methods to perform the first data processing. The deduplicated data are then unified to the same spatiotemporal coordinate system for spatiotemporal alignment and the second data processing is performed. After that, missing values ​​are filled for obviously missing data and the third data processing is performed. At the same time, data standardization methods are used to normalize and semantically standardize data of different dimensions. Unstructured addresses are geocoded and uniformly converted into longitude and latitude coordinates, and the dimensional differences are eliminated through the Z-Score standardization method. Then, a geographic information semantic output model was constructed, and the sliding window algorithm and dependency syntax analysis were used to divide the features of text information. For example, "the average annual precipitation in area A is 1200mm, and there are 5 subway stations" was divided into two independent semantic blocks, "the average annual precipitation is 1200mm" and "the number of subway stations is 5". The grayscale value of each semantic block was calculated by integrating the grayscale-like formula, in which the TF-IDF weight reflects the statistical significance of the keyword, and Word2Vec measures its semantic similarity with the predefined geographic ontology concepts; text reorganization was performed based on the geographic semantic output rule base. If a semantic block triggered a rule, Then, if the precipitation is >1000mm and the vegetation coverage field needs to be associated, the missing attributes are completed from the knowledge graph to generate a structured geographic description unit; for economic data and life service data, a hierarchical graph convolutional network is used to construct a spatiotemporal knowledge graph, and the nodes represent geographic entities such as commercial areas. The edge weights are generated by dynamic genetic algorithm optimization. When the algorithm is initialized, multiple groups of weight combinations are randomly generated and the regional matching effect is evaluated through the fitness function. During the iteration process, the weight configuration is adjusted through two-point crossover and Gaussian mutation, and the optimal weight matrix is ​​finally output to calculate the comprehensive matching degree of the multi-factor geographic matrix; In this process, the improved graph neural network introduces residual connections and adaptive attention mechanisms to dynamically aggregate neighborhood information according to node attributes. For example, for transportation hub nodes, the focus is on aggregating the characteristics of residential and commercial areas within three hops, while avoiding logical conflicts through spatial constraint loss functions. Finally, high-confidence data is screened through a threshold mechanism, and real-time social geographic features are incrementally updated in combination with a streaming processing framework to form a dynamic and scalable quantitative decision analysis. By building a geographic information semantic output model, the conditional geographic information, regional geographic information and social geographic information are deeply integrated, so that the data is no longer isolated, and the integrity and relevance of geographic information are improved; this method can effectively integrate structured and unstructured geographic information and expand the application scope of geographic information; the text feature division method is used to subdivide the geographic information text into multiple feature blocks to ensure that key information can be independently parsed and improve the parsing accuracy of the data; combined with the semantic grayscale calculation method, it can effectively process fuzzy or complex text expressions, making the processing of geographic information more accurate and intelligent; The construction process of the geographic information semantic output model is based on a large amount of annotated text data and is carried out through supervised learning. The construction steps are as follows: collect text data containing geographic information, and perform preprocessing steps such as cleaning, word segmentation, and annotation on the text. Each text should contain labels with conditional, regional, and social geographic information; perform feature division on each text sample, extract feature blocks in the text, and then use the grayscale calculation method to generate the corresponding grayscale value; use machine learning algorithms such as neural networks and support vector machines, combined with geographic semantic output rules to train the model, and optimize the objective function to enable the model to accurately predict the geographic information in the input text; during the training process, use cross-validation, accuracy and other evaluation indicators to test and adjust the model performance to ensure that the model can accurately mine semantic information related to geography; The grayscale-like calculation method is used to map geographic information from different sources into a unified semantic grayscale space to improve the integration and comparability of information. This method can be used in urban planning, regional risk assessment, traffic optimization and other fields to make the expression of geographic information more intuitive and hierarchical. The integrated information is reorganized through geographic semantic output rules to meet the needs of geographic analysis and prediction, thereby improving the availability of geographic information. It is suitable for scenarios such as smart city construction, geographic big data analysis, disaster prediction, and social event monitoring to improve the accuracy of data-driven decision-making. The spatial constraint analysis is combined with geographic semantic rules to ensure that the output data conforms to the actual geographic area division and avoids information mismatch across regions. The grayscale-like calculation method can be used to perform grayscale calculation based on fuzzy mathematics, vector space models, probability statistics and other methods, to semantically standardize information from different sources, reduce data conflicts, and improve the accuracy of data integration. The grayscale-like calculation method is used to measure the similarity between different geographic information, eliminate redundant and irrelevant data, and ensure the consistency of output information. Gray value generation is to calculate the semantic weight of each feature block, the formula is Z i =TF-IDF(K i )*α+Word2Vec(K i )*β, α+β=1, used to balance the importance and semantic relevance of keywords; TF-IDF (K i ) is a weight value, which is used to measure the importance of keywords in the text feature block in the text. It can quantify the statistical significance of keywords in Ki, such as the high weight of "mountain" when describing terrain; Word2Vec (K i ) represents the semantic relevance, which measures its similarity with the geographic ontology library, such as predefined concepts such as topography and administrative divisions; The design of geographic semantic output rules needs to ensure the logic, spatial consistency, temporal coherence and semantic accuracy of the data, so as to generate a high-quality mining geographic information set after integrating multi-source geographic information; it is necessary to follow semantic standardization, spatiotemporal consistency, semantic grayscale calculation, geographic relationship reasoning and constraint optimization to ensure the scientificity and rationality of data integration; reorganized text based on geographic semantic output rules, for example: the preset spatial rule is that if the text contains "precipitation>1000mm", it needs to be associated with "vegetation coverage ≥60%"; the preset time rule is that if the data is "working day", it needs to be associated with "morning and evening peak traffic flow"; The original input text (conditional geography, regional geography, social geography information) is finely segmented through dependency syntax analysis or semantic boundary detection algorithm, and independent semantic units are extracted as feature blocks (such as splitting "annual precipitation in West Lake District 1200mm" into "West Lake District" and "precipitation 1200mm"); then, a fusion-like grayscale calculation is performed on each feature block, and a semantic grayscale is generated through a linear weighted formula; according to the predefined geographic semantic output rule library containing three types of rules: spatial constraints, temporal associations, and social impacts, the grayscale values ​​are prioritized and matched - for example, the spatial rule requires that "feature blocks with precipitation > 1000mm must be associated with the vegetation coverage field", and if the feature block does not meet the requirement, a trigger is triggered. The rule engine automatically completes missing data, such as retrieving vegetation data of the area from the geographic database; for multi-rule conflict scenarios such as "traffic congestion period", it is necessary to simultaneously associate "workday rule" and "weather anomaly rule", adopt a dynamic weight adjustment mechanism, use real-time policy or environmental indicators, such as warning level as the basis for weight adjustment, and give priority to high-priority rules; after completing the rule matching, use conflict resolution algorithms such as weighted voting based on data authority to handle multi-source data contradictions such as differences in administrative division boundaries provided by different institutions, and use the template engine to reorganize feature blocks according to the "regional code-natural attribute-social attribute" hierarchical structure, and finally generate standardized geographic judgment standards, which are uniformly displayed in the form of scores; In this embodiment, geographic information mining for a city involves multiple factors such as population density, land use, transportation network, and ecological environment. Traditional methods are difficult to integrate multi-source data and dynamically adapt to changing trends. Through unique geographic spatial coding, geographic multi-dimensional attribute tags, geographic social characteristics tags and other technical means, combined with big data analysis, comprehensive mining and integration of urban geographic information, forming a high-quality geographic information collection, thereby optimizing urban management and decision-making. A city can be divided into multiple levels such as administrative districts, streets, communities, and business districts. Each unit is given a unique code to ensure data consistency. Through big data retrieval, information from multiple data sources such as GIS data, remote sensing images, traffic data, environmental monitoring data, social media data, and open data can be collected to provide The breadth of high-quality geographic data covers information in multiple fields such as urban infrastructure, social economy, and natural environment. The collected multi-breadth geographic information is screened to remove irrelevant or redundant data, improve the pertinence of the data, and ensure the relevance of the data to the area to be mined. The city's geographic information is deeply mined through multiple dimensions such as topography, climate, population density, land use, and ecological environment. An in-depth understanding of regional characteristics is provided to provide support for precise planning. Social media, population migration, economic data, and life service data are combined to identify the social and geographical characteristics of the city. By integrating conditional geographic information, regional geographic information, and social geographic information, a comprehensive, accurate, and dynamic urban geographic information collection is formed, forming a high-quality geographic information database to provide a precise decision-making basis for urban management.

[0017] Example 2, a geographic information mining system based on big data, see Figure 1 As shown, including: The geographic region selection module includes a region determination unit for obtaining a unique geographic space code, a geographic multi-dimensional attribute label, and a geographic social characteristic label for the geographic region to be mined; Geographic information mining module, including conditional mining unit, regional mining unit and social mining unit; The conditional mining unit is used to search based on the geographic area to be mined, collect a number of multi-breadth geographic information from multiple data sources; perform geographic condition screening on the multiple multi-breadth geographic information to obtain conditional geographic information; The regional mining unit is used to search based on geographical multi-dimensional attribute tags, collect multiple multi-depth geographical information from multiple data sources, and perform geographical region screening on multiple multi-depth geographical information to obtain regional geographical information; The social mining unit is used to search based on geographical social characteristic tags to obtain a number of multi-factor geographical information; the geographical factors of a number of multi-factor geographical information are screened to obtain social geographical information; The geographic information integration module includes a data integration unit, which is used to integrate the conditional geographic information, regional geographic information and social geographic information to obtain a mining geographic information set that conforms to the geographic area to be mined; by constructing a geographic information semantic output model, the conditional geographic information, regional geographic information and social geographic information are divided into text features, fused with grayscale calculations, and the text is reorganized based on geographic semantic output rules, and finally the mining geographic information set is output after integration.

[0018] It should be understood that those skilled in the art can make improvements or changes based on the above description, and all such improvements and changes should fall within the scope of protection of the appended claims of the present invention. Parts not described in detail in this specification belong to the prior art known to those skilled in the art.

Claims

1. A geographic information mining method based on big data, characterized in that: The following steps are involved: For the geographic area to be mined, obtain unique geospatial codes, geographic multi-dimensional attribute labels, and geo-social characteristic labels; Search based on the geographic area to be mined, and collect several pieces of multi-breadth geographic information from multiple data sources; Perform geographic condition screening on a number of multi-breadth geographic information to obtain conditional geographic information; Search based on geographic multi-dimensional attribute tags to collect multiple pieces of multi-depth geographic information from multiple data sources; Performing geographic area screening on a plurality of multi-depth geographic information to obtain regional geographic information; Search based on geo-social characteristic tags to obtain several multi-factor geographic information; Screening several pieces of multi-factor geographic information by geographic factors to obtain social geographic information; Integrate conditional geographic information, regional geographic information and social geographic information to obtain a mining geographic information set that meets the geographic area to be mined; By constructing a geographic information semantic output model, the text features of conditional geographic information, regional geographic information and social geographic information are divided and integrated into grayscale calculations. The text is reorganized based on geographic semantic output rules, and finally the integrated output is used to mine the geographic information collection.

2. The method for mining geographic information based on big data according to claim 1, characterized in that: The specific steps for data integration of conditional geographic information, regional geographic information and social geographic information include: Construct geographic information semantic output model; In the geographic information semantic output model, the text feature segmentation is performed on the input text to obtain the text feature block K i ; i=1, 2, ..., I, I is the total number of data blocks for text feature division; For text feature block K i Perform fusion grayscale calculation to obtain the text semantic grayscale value Z i ; For text semantic gray value Z i The input text is reorganized with geographic semantic output rules to obtain mined geographic information; Among them, the input text is conditional geographic information, regional geographic information and social geographic information; The output mined geographic information is combined to obtain a mined geographic information set.

3. The method for mining geographic information based on big data according to claim 2, characterized in that: The specific steps for geographical condition screening of multiple pieces of multi-breadth geographical information include: Extract text features from multi-breadth geographic information to obtain geographic information text features; Set the geographical conditions of the geographic area to be mined as N regional related dimension labels B n , n=1,2,…,N; Based on geographic information text features and region-related dimension labels B n Establish geographic topological relationship matrix; For each characteristic edge in the geographic topological relationship matrix, the edges that do not conform to the m regional related dimension labels B are removed. n characteristic edges, and obtain a new geographic topological relationship matrix; If the structural integrity of the new geographic topological relationship matrix is ​​greater than the preset value, the multi-breadth geographic information is retained to complete the screening; All multi-breadth geographic information is traversed and screened, and the retained multi-breadth geographic information is combined to obtain conditional geographic information.

4. The method for mining geographic information based on big data according to claim 3, characterized in that: The specific steps for performing geographic area screening for a number of multi-depth geographic information include: Using genetic algorithm to set geographic region weight matrix; Establish a spatiotemporal knowledge graph for multi-depth geographic information and obtain a geographic information reasoning graph; Establishing geographic entity models for multi-depth geographic information to obtain simulated geographic entities; On the basis of simulating geographic entities, the geographic information reasoning map and the geographic area weight matrix are used for regional matching. The improved graph neural network is used to couple the geographic information reasoning map and the simulated geographic entities to obtain the geographic information matching degree. The trained geographic uncertainty quantification model is used in combination with a preset matching threshold to screen the geographic information matching degree and determine whether to retain multi-depth geographic information; All multi-depth geographic information is traversed and screened, and the retained multi-depth geographic information is combined to obtain regional geographic information.

5. The method for mining geographic information based on big data according to claim 4, characterized in that: The specific steps for geographical factor screening of several multi-factor geographical information include: Construct sentiment maps for multi-factor geographic information and generate positive and negative geographic information features; Analyze the emotional causes of positive and negative geographic information features to obtain positive and negative geographic keywords; Capture the periodic fluctuation characteristics of multi-factor geographic information and obtain social geographic characteristics with time series labels; Generate a multi-factor geographic matrix by combining positive geographic keywords, negative geographic keywords, and social geographic features; The dynamic weights of the multi-factor geographic matrix are adjusted based on the real-time geographic region constraint indicators to obtain an updated multi-factor geographic matrix; Calculate and update the comprehensive matching degree of the multi-factor geographic matrix, and select whether to retain the multi-factor geographic information based on the data interval of the comprehensive matching degree; All multi-factor geographic information is traversed and screened, and the retained multi-factor geographic information is combined to obtain social geographic information.

6. The method for mining geographic information based on big data according to claim 5, characterized in that: Based on the existing geographic information set of the geographic area to be mined, keyword features are extracted to obtain geographic multi-dimensional attribute labels; data is extracted from the real-time social attributes of the geographic area to be mined to obtain geographic social characteristic labels.

7. A geographic information mining system based on big data, characterized in that: The system applies a geographic information mining method based on big data as described in any one of claims 1 to 6, including: The geographic region selection module includes a region determination unit for obtaining a unique geographic space code, a geographic multi-dimensional attribute label, and a geographic social characteristic label for the geographic region to be mined; Geographic information mining module, including conditional mining unit, regional mining unit and social mining unit; The conditional mining unit is used to search based on the geographic area to be mined, collect a number of multi-breadth geographic information from multiple data sources; perform geographic condition screening on the multiple multi-breadth geographic information to obtain conditional geographic information; The regional mining unit is used to search based on geographical multi-dimensional attribute tags, collect multiple multi-depth geographical information from multiple data sources, and perform geographical region screening on multiple multi-depth geographical information to obtain regional geographical information; The social mining unit is used to search based on geographical social characteristic tags to obtain a number of multi-factor geographical information; the geographical factors of a number of multi-factor geographical information are screened to obtain social geographical information; The geographic information integration module includes a data integration unit, which is used to integrate the conditional geographic information, regional geographic information and social geographic information to obtain a mining geographic information set that conforms to the geographic area to be mined; by constructing a geographic information semantic output model, the conditional geographic information, regional geographic information and social geographic information are divided into text features, fused with grayscale calculations, and the text is reorganized based on geographic semantic output rules, and finally the mining geographic information set is output after integration.

Images