Intelligent push processing method based on GIS city data mining

Abstract

The invention relates to the technical field of intelligent push processing based on GIS city data mining, and discloses an intelligent push processing method based on GIS city data mining. And a complete process from data acquisition, structure construction to hotspot identification and pushing is covered. A leaf node set is finally formed by setting a research area and grid division parameters, calculating a mean value and a standard deviation of observation values, setting a quadtree decomposition threshold value and a maximum depth, constructing a root node and performing recursive decomposition. Hot spot candidate nodes meeting hierarchical requirements are screened, spatial adjacent hot spots are merged based on an adjacent matrix, and a push list is generated according to priority ranking. According to the method, adaptive space division is realized through a quadtree, the recognition precision is improved by combining a statistical method, the hotspot aggregation capability is enhanced by utilizing connectivity analysis, and a closed-loop system from original data to intelligent pushing is constructed.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent push processing based on GIS city data mining, and in particular to an intelligent push processing method based on GIS city data mining. Background Art

[0002] As urban informatization accelerates, urban data mining and intelligent data push technologies based on geographic information systems (GIS) are becoming increasingly important components of smart city development. By analyzing large-scale spatiotemporal data such as urban traffic flows and crowd activity, dynamic monitoring of hotspots and precise services can be achieved, improving the efficiency of urban management and public services.

[0003] In existing technologies, intelligent push notification systems often rely on rule-based triggering or static threshold methods. Traditional approaches typically divide urban areas into grids of equal area, globally aggregate traffic or event statistics for each grid, and then uniformly process all grids based on a fixed segmentation threshold or preset empirical parameters. This approach ignores regional differences in data distribution, making it difficult to achieve refined real-time detection in areas of high volatility. This often results in "over-push" or "missed detection," impacting the accuracy and timeliness of hotspot identification. To improve spatial focusing capabilities, some research has attempted to apply spatial hierarchical index structures such as quadtrees to urban data analysis. Quadtrees continuously re-segment the space, enabling finer subdivision of high-density areas. However, existing solutions often employ manually set segmentation thresholds and fixed maximum depths, lacking adaptive adjustment of segmentation criteria. Furthermore, leaf node merging often relies on globally predefined parameters and cannot be dynamically optimized based on the actual data correlation between nodes. This results in hotspots being easily fragmented into isolated fragments or over-merged, reducing detection effectiveness. In terms of regional merging and connectivity processing, currently commonly used methods include clustering algorithms based on Euclidean distance or DBSCAN, which first group rasters or small fragments according to density or distance thresholds, and then complete the merging through connected component search in graph theory. This type of method is highly sensitive to parameters such as distance thresholds and minimum number of neighbors, requiring multiple rounds of tuning in different scenarios. The merging results are easily affected by noisy data, making it difficult to ensure the authenticity and stability of spatial clustering. In addition, existing technologies usually introduce empirical parameters in multiple links such as spatial division, threshold setting, decomposition depth and merging strategy, and lack a unified adaptive strategy, making the entire hotspot detection and intelligent push process too dependent on manual experience, and unable to achieve end-to-end closed-loop automation in complex urban environments with multiple sources, multiple scales and multiple time periods. Especially in typical scenarios such as traffic peaks, sudden events or holidays, it is difficult to strike a balance between real-time performance, accuracy and computational overhead.

[0004] To this end, this project aims to propose an intelligent push processing method based on GIS urban data mining. By setting reasonable spatial partitions and time windows, and recursively refining spatial regions using a quadtree decomposition algorithm, dynamic urban data (such as taxi GPS locations) is hierarchically expressed according to volatility, thereby identifying high-intensity hotspots. Subsequently, a connected component search method is used to merge adjacent hotspots, and statistical indicators are used to prioritize them. The final output is a list of hotspots for intelligent push notifications. Summary of the Invention

[0005] The present invention provides an intelligent push processing method based on GIS city data mining, which promotes the solution of the problems mentioned in the above background technology.

[0006] The present invention provides the following technical solution: an intelligent push processing method based on GIS city data mining, comprising:

[0007] Set the study area and initial grid division parameters, build a grid set and determine the observation indicators;

[0008] Calculate the mean and standard deviation of the observation values ​​of all grids, and set the quadtree decomposition threshold and maximum decomposition depth;

[0009] Construct the root node of the quadtree, set the region boundaries, observation mean and standard deviation;

[0010] Recursively decompose the nodes based on the set threshold and observed data, generate child nodes and update indexes and statistics until the termination condition is met;

[0011] Collect all nodes that are no longer decomposed to form a set of leaf nodes;

[0012] Filter leaf nodes that meet the preset hierarchical conditions as hotspot candidates;

[0013] Search for connected components based on the adjacency matrix and merge spatially adjacent hotspot candidate nodes;

[0014] Prioritize the merged hotspot areas and generate a push list.

[0015] Optionally, setting the study area and initial grid division parameters, constructing a grid set, and determining observation indicators specifically include:

[0016] Assume that the study area is a square area , the side length is ;

[0017] Let the minimum leaf node edge length be ,and ;

[0018] Assume the initial number of grid rows and columns is , then the actual grid side length is ;

[0019] Will Divided equally grids of the same area, forming a grid set :

[0020] , ;in, For the grids; is the total number of grids;

[0021] For each Set line number , column number , whose areas are: ;in, For grid The two-dimensional coordinate range of ; is the remainder operation;

[0022] Set the time window to the past hour;

[0023] Get the taxi GPS device in the grid within the time window The total number of reported positioning points in the area is recorded as the observation value .

[0024] Optionally, the calculation of the mean and standard deviation of the observation values ​​of all grids and setting the quadtree decomposition threshold and maximum decomposition depth specifically includes:

[0025] Calculates the arithmetic mean of all raster observations , ;

[0026] Calculate the global standard deviation , ;

[0027] Set the quadtree decomposition threshold to ;

[0028] Set the maximum decomposition depth of the quadtree to .

[0029] Optionally, constructing the root node of the quadtree and setting the region boundary, observation mean, and standard deviation specifically include:

[0030] Let the root node of the quadtree be , ;in, For nodes Depth in the tree, root node ;

[0031] Let the root node region boundary be , ;in, For nodes Left and right borders coordinate; For nodes Upper and lower boundaries Coordinates here is the root node;

[0032] Set the root node grid index set to , ;in, For nodes Grid number; To include area The collection of all raster indexes in;

[0033] Set the observation mean and observation standard deviation in the root node area to be , ;in, For nodes Mean of observations within the region; is the standard deviation of observations within the node area.

[0034] Optionally, recursively decomposing the node based on the set threshold and the observed data, generating child nodes and updating indexes and statistics until the termination condition is met, specifically includes:

[0035] For any node , initially , perform the following steps:

[0036] S101, determine whether to decompose:

[0037] like , then decompose; otherwise mark it as a leaf node and return;

[0038] S102. Calculate the node midpoint coordinates:

[0039] , ;in, For the current node The coordinates of the center point of the region;

[0040] S103. Generate four child nodes:

[0041] ;

[0042] ;

[0043] ;

[0044] ;

[0045] in, For nodes No. child nodes;

[0046] And order ;

[0047] S104, child node grid index:

[0048] For each Recalculate , ;in, is the number of elements in the combination;

[0049] like , then discard the child node; otherwise continue;

[0050] S105. Child node statistics:

[0051] , ;

[0052] For each make Repeat steps S101 to S105 for the child nodes of .

[0053] Optionally, collecting all nodes that are no longer decomposed to form a leaf node set specifically includes:

[0054] Collect all nodes that are not further decomposed as leaf nodes:

[0055] ;in, is the set of all leaf nodes that are no longer decomposed.

[0056] Optionally, the step of screening leaf nodes that meet preset hierarchical conditions as hotspot candidates specifically includes:

[0057] Select the node with the maximum level from the leaf nodes as the hotspot candidate:

[0058] ;in, It is the set of candidate hotspot nodes.

[0059] Optionally, the connected component search based on the adjacency matrix and merging spatially adjacent hotspot candidate nodes specifically includes:

[0060] Setting up the matrix ,in , for any candidate node :

[0061] ;

[0062] in, is a matrix element;

[0063] Connected component search:

[0064] S210, initialize the unvisited set , connected component list ;

[0065] S220, when :

[0066] S221, from Take a node , let the current component set ,queue , and from Removed;

[0067] S222, when , pop up node ; for each like , then join in and and from Remove ;in, is the current node and adjacent nodes in the traversal process;

[0068] S223, the amount to be completed join in ;

[0069] S230, at the end ;

[0070] For each component , calculate the The collection range of the merged hotspot areas All raster index collections contained in the area :

[0071] , .

[0072] Optionally, the step of prioritizing the merged hotspot areas and generating a push list specifically includes:

[0073] Calculate the The average value of hotspot areas and standard deviation :

[0074] , ;in, For collection The cardinality of

[0075] according to Arrange all merged areas from largest to smallest;

[0076] Before selection Regions , output push tuple:

[0077] , .

[0078] The present invention has the following beneficial effects:

[0079] 1. Dynamically setting the minimum leaf node edge length and the number of grid rows and columns makes spatial partitioning more scalable and adaptable. Combining time window settings with spatial partitioning enables the system to perform simultaneous spatiotemporal analysis. A clear two-dimensional coordinate conversion and indexing mechanism provides a unified spatial reference for subsequent algorithm processing. Traditional heat map methods often use static grid partitioning, which cannot automatically adjust regional granularity based on the spatial density of the data. This solution improves spatial expression capabilities through parameter setting and automatic grid numbering, providing a solid foundation for subsequent data aggregation and decomposition.

[0080] 2. The decomposition threshold is dynamically calculated using the global mean and standard deviation, implementing a data-driven decomposition judgment standard. A maximum decomposition depth control mechanism is proposed to effectively avoid the waste of computing resources caused by infinite quadtree refinement. This approach has strong generalization capabilities and is applicable to data analysis scenarios across different regions and urban scales. Compared to existing technologies that require manual experience to set static thresholds, this solution adaptively determines the necessity of regional decomposition through automatic statistical analysis, significantly improving the accuracy and efficiency of hotspot identification.

[0081] 3. Establishing a unified spatial management structure based on the entire region enables recursive traceability of region division and node management. Clear mapping of all grids to the root node index facilitates structural operations such as region merging and region searching. Initial statistics are driven by real-world data, establishing reasonable initial conditions for decomposition. Traditional methods lack dedicated structures for unified scheduling of spatial nodes, resulting in redundant subsequent operations. This method encapsulates the initial state through the root node, improving the organization and controllability of the spatial tree structure.

[0082] 4. A hierarchical recursive structure is employed to gradually refine regions, achieving hierarchical detection of spatial hotspots from coarse to fine scale. Local statistics are recalculated at each sub-node level to adaptively determine whether to continue decomposition. Invalid nodes without grids are discarded, optimizing the simplicity of the data structure and memory usage. Compared to traditional static grid analysis, this method, through the combined control of "meaningful decomposition and termination mechanism," not only avoids redundant region processing but also improves spatial resolution and enables precise hotspot location.

[0083] 5. Extract all nodes that are no longer refined as "leaf nodes," forming the base set for subsequent analysis. Filter them based on hierarchical criteria, retaining only the most fine-grained and significantly fluctuating areas as hotspot candidates. Existing technologies generally use fixed thresholds or manual determination of hotspot locations, lacking spatial hierarchical sensitivity. This solution, based on a screening mechanism that combines hierarchical decomposition with data fluctuations, is more scientific and rational.

[0084] 6. Construct an adjacency matrix and merge adjacent hotspot nodes using connected component search to fully preserve the continuity of hotspots. Efficient clustering is performed using a graph-structured breadth-first traversal algorithm to ensure the coherence and integrity of the merged regions. The merged spatial extent and raster index set are output to facilitate subsequent comprehensive evaluation. Unlike traditional methods that rely solely on numerical merging, this approach emphasizes the dual merging criteria of "spatial proximity and numerical significance," improving the completeness of the hotspot representation and the fidelity of urban spatial characteristics.

[0085] 7. For each hotspot area, calculate its statistical indicators (mean, standard deviation) and establish a unified sorting rule; output the push tuple of the hotspot area with the highest priority to achieve a closed loop from data perception to action instructions; support real-time push service deployment, such as police patrols, emergency response and other scenarios. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 It is a schematic diagram of the process of the present invention. DETAILED DESCRIPTION

[0087] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0088] Example, see Figure 1 , an intelligent push processing method based on GIS urban data mining, comprising:

[0089] Set the study area and initial grid division parameters, build a grid set and determine the observation indicators;

[0090] Calculate the mean and standard deviation of the observation values ​​of all grids, and set the quadtree decomposition threshold and maximum decomposition depth;

[0091] Construct the root node of the quadtree, set the region boundaries, observation mean and standard deviation;

[0092] Recursively decompose the nodes based on the set threshold and observed data, generate child nodes and update indexes and statistics until the termination condition is met;

[0093] Collect all nodes that are no longer decomposed to form a set of leaf nodes;

[0094] Filter leaf nodes that meet the preset hierarchical conditions as hotspot candidates;

[0095] Perform connected component search based on the adjacency matrix and merge spatially adjacent hotspot candidate nodes;

[0096] Prioritize the merged hotspot areas and generate a push list.

[0097] A complete, intelligent push processing method for GIS urban data is proposed, covering the entire process from data collection and structural organization to hotspot identification, priority sorting, and push. By introducing a recursive quadtree structure to hierarchically organize and decompose urban spatial data, the problem of varying hotspot density in different regions is effectively addressed. Statistical methods are then used to set decomposition thresholds and priority evaluation indicators to improve the accuracy of hotspot identification. Finally, spatially adjacent hotspot areas are merged through the adjacency matrix, achieving a closed-loop process from "raw data" to "high-quality hotspot push." ​​By setting spatial partitioning parameters and combining spatiotemporal observation data, the problem of untimely and inaccurate capture of urban hotspot changes is resolved. Through recursive decomposition and connected clustering, multi-scale hotspot areas are automatically discovered, avoiding the subjectivity brought about by manual zoning. The push mechanism improves response speed and accuracy, making it widely applicable to scenarios such as urban safety warnings, traffic dispatch, and emergency management.

[0098] The setting of the study area and initial grid division parameters, the construction of the grid set and the determination of the observation indicators specifically include:

[0099] Assume that the study area is a square area , the side length is ; Clarify the overall spatial scope and provide a geometric basis for grid division;

[0100] Let the minimum leaf node edge length be ,and ;Control the minimum partition granularity and balance resolution and computational complexity;

[0101] Assume the initial number of grid rows and columns is , then the actual grid side length is ; Determine the number of rows and columns and the actual side length of the grid to ensure complete coverage of the area;

[0102] Will Divided equally grids of the same area, forming a grid set :

[0103] , ;in, For the grids; is the total number of grids; establish an index set of all grid cells to provide a unified identifier for subsequent references;

[0104] For each Set line number , column number , whose areas are: ;in, For grid The two-dimensional coordinate range of ; It is a remainder operation; each grid is mapped to the actual plane coordinates for subsequent point-region attribution judgment;

[0105] Set the time window to the past hour;

[0106] Get the taxi GPS device in the grid within the time window The total number of reported positioning points in the area is recorded as the observation value .

[0107] The fundamental steps of the entire data mining process are centered around "setting the study area, grid division, and obtaining observation indicators." By defining the urban area as a standard square and setting the minimum leaf node side length and initial grid row and column number based on the actual scenario, a unified grid system is established. Using taxi GPS data as the observation object, real-time quantification of the city's operational status is achieved. This step achieves unified digital modeling of the urban space, providing a spatial reference standard for subsequent analysis. Gridding transforms the original unstructured point data into actionable two-dimensional matrix data, enhancing the scalability and processing efficiency of data analysis. Setting a time window ensures timely data analysis and improves the system's adaptability in dynamic scenarios. These settings lay a solid spatiotemporal foundation for subsequent quadtree decomposition and hotspot identification.

[0108] The calculation of the mean and standard deviation of the observation values ​​of all grids and setting the quadtree decomposition threshold and maximum decomposition depth specifically includes:

[0109] Calculates the arithmetic mean of all raster observations , ;Provide a central tendency benchmark for region-wide data;

[0110] Calculate the global standard deviation , ; Quantify the discrete degree of data in the entire region and provide a quantitative basis for decomposition thresholds;

[0111] Set the quadtree decomposition threshold to Adaptively use global fluctuations as the criterion for continued decomposition, without the need for external parameter adjustment;

[0112] Set the maximum decomposition depth of the quadtree to ; Limit the decomposition depth to prevent excessive refinement from causing computational explosion.

[0113] Paying attention to "global statistics calculation and decomposition parameter setting" is the key link in whether the quadtree decomposition can be carried out scientifically. It first calculates the mean and standard deviation of the observation values ​​of all rasters to provide a benchmark for the overall data distribution in the region, and then sets the quadtree decomposition threshold and the maximum number of decomposition layers based on the standard deviation to decide when to decompose and when to terminate. By setting the decomposition threshold through data-driven, the subjectivity and inadaptability of manually setting the threshold in traditional methods are overcome, and the adaptive ability of the model is realized; the maximum decomposition depth is set to effectively control the computational overhead and ensure the stable operation of the system. This step solves the problems of arbitrary decomposition decisions and the difficulty in balancing accuracy and resource consumption. It is the core prerequisite for achieving refined hotspot identification.

[0114] The construction of the root node of the quadtree, setting the region boundary, observation mean and standard deviation, specifically includes:

[0115] Let the root node of the quadtree be , ;in, For nodes Depth in the tree, root node ; Construct the starting point of the quadtree and mark the initial level as 0;

[0116] Let the root node region boundary be , ;in, For nodes Left and right borders coordinate; For nodes Upper and lower boundaries Coordinates here is the root node; specify the complete geometric area covered by the root node;

[0117] Set the root node grid index set to , ;in, For nodes Grid number; To include area The collection of all raster indexes in;

[0118] Set the observation mean and observation standard deviation in the root node area to be , ;in, For nodes Mean of observations within the region; is the standard deviation of observations within the node area;

[0119] Establish a mapping between the root node and all grids, and calculate the initial statistics for decomposition judgment.

[0120] This paper describes the process of constructing a quadtree root node. By setting the node's initial depth, region boundaries, and set of included raster indices, and calculating their observed mean and standard deviation, it provides a starting point for establishing the entire quadtree's recursive structure. The root node uniformly carries the spatial extent, data index, and initial statistical indicators, making subsequent node decomposition directional and targeted. The unified index system and boundary definitions lay the foundation for recursive operations and hotspot location. Furthermore, the root node's initial statistical information is used to determine whether decomposition is necessary in the first round, thereby improving processing efficiency and avoiding redundant processing of the entire region.

[0121] The node is recursively decomposed based on the set threshold and observation data, child nodes are generated, and indexes and statistics are updated until the termination condition is met, specifically including:

[0122] For any node , initially , perform the following steps:

[0123] S101, determine whether to decompose:

[0124] like , then decompose; otherwise mark it as a leaf node and return it; only decompose the nodes with large fluctuations and not reaching the maximum level to ensure meaningful refinement;

[0125] S102. Calculate the node midpoint coordinates:

[0126] , ;in, For the current node The coordinates of the center point of the area; determine the center of the current node area, which is used to equally divide the sub-areas;

[0127] S103. Generate four child nodes:

[0128] ;

[0129] ;

[0130] ;

[0131] ;

[0132] in, For nodes No. child nodes;

[0133] And order ;

[0134] Split the region into parts according to the midpoint to achieve quadtree decomposition;

[0135] S104, child node grid index:

[0136] For each Recalculate , ;in, To count the elements in the combination; determine the grid set that each child node is responsible for, in preparation for local statistics;

[0137] like , then discard the child node; otherwise continue;

[0138] S105. Child node statistics:

[0139] , ; Obtain the mean and fluctuation degree of the sub-region for the next round of decomposition judgment or leaf node confirmation;

[0140] For each make Repeat steps S101 to S105 for the child nodes of ; continue iterating until all high fluctuation areas are refined to the minimum level or meet the uniformity.

[0141] It determines whether each node needs to be decomposed, layer by layer, based on the set decomposition threshold and maximum level. If the conditions are met, the region is divided into four sub-nodes, and the index and statistics of each sub-node are updated until all nodes meet the "non-decomposition" condition. By introducing a dynamic judgment mechanism, decomposition is only performed in areas with spatial data fluctuations, reducing invalid calculations. It continuously iterates and optimizes the regional granularity, ensuring that hotspot identification achieves the most reasonable spatial resolution. By driving the decomposition logic with statistical values, the accuracy and intelligence of regional processing are greatly enhanced, adapting to the needs of large-scale data processing.

[0142] The collecting of all nodes that are no longer decomposed to form a leaf node set specifically includes:

[0143] Collect all nodes that are not further decomposed as leaf nodes:

[0144] ;in, is the set of all leaf nodes that are no longer decomposed;

[0145] Summarize all nodes that terminate the decomposition to form the basic set for subsequent hotspot screening.

[0146] A pre-step in hotspot screening is to collect all "terminated decomposition" nodes to form a leaf node set. Leaf nodes are the endpoints of the quadtree decomposition and represent the most granular areas of information. This step aggregates the endpoint areas from the previous decomposition process to form a candidate set for the next step of hotspot identification. By focusing on these areas for analysis, the accuracy and efficiency of hotspot detection can be significantly improved. It also ensures a clear data flow and a well-structured foundation for subsequent screening and sorting operations.

[0147] The screening of leaf nodes that meet preset hierarchical conditions as hotspot candidates specifically includes:

[0148] Select the node with the maximum level from the leaf nodes as the hotspot candidate:

[0149] ;in, is the set of hotspot candidate nodes;

[0150] The areas with the finest granularity and still high fluctuation are selected as preliminary candidates for hotspots.

[0151] Based on the leaf nodes, we further screen nodes that meet the maximum level as hotspot candidates, excluding lower-level nodes or nodes with less data fluctuations. Through this hierarchical screening mechanism, we focus on the data units with the most significant fluctuations at the finest level, achieving precise positioning of abnormal areas or emergencies; avoiding misidentifying minor data fluctuations within larger areas as hotspots, improving judgment quality; and further compressing the processing data scale to improve the efficiency of subsequent clustering and sorting.

[0152] The connected component search based on the adjacency matrix and the merging of spatially adjacent hotspot candidate nodes specifically include:

[0153] Setting up the matrix ,in , for any candidate node :

[0154] ;

[0155] in, is a matrix element;

[0156] Connected component search:

[0157] S210, initialize the unvisited set , connected component list ;

[0158] S220, when :

[0159] S221, from Take a node , let the current component set ,queue , and from Removed;

[0160] S222, when , pop up node ; for each like , then join in and and from Remove ;in, is the current node and adjacent nodes in the traversal process;

[0161] S223, the amount to be completed join in ;

[0162] S230, at the end ;

[0163] Aggregate all adjacent hotspot nodes through the queue traversal algorithm;

[0164] For each component , calculate the first The collection range of the merged hotspot areas All raster index collections contained in the area :

[0165] , ;

[0166] Merge all subnode spaces within a component and summarize the grids it contains to prepare for overall hotspot statistics.

[0167] Using an adjacency matrix and connected component search algorithm, spatially contiguous candidate hotspot nodes are merged to generate a more structured and expressive set of hotspot regions. A graph structure traversal method fully considers the spatial connectivity between hotspots, avoiding fragmented judgments. It also merges multiple densely distributed small hotspots into a unified region, improving the strategic and centralized nature of information push. Compared to traditional simple threshold-based merging methods, this approach is more spatially intelligent and algorithmically robust, making it particularly suitable for complex urban spatial patterns.

[0168] Prioritizing the merged hotspot areas and generating a push list specifically includes:

[0169] Calculate the The average value of hotspot areas and standard deviation :

[0170] , ;in, For collection The base number; evaluate the overall fluctuation of each hotspot area as the basis for subsequent priority;

[0171] according to Arrange all merged areas from largest to smallest; give the areas with the largest fluctuations the highest priority;

[0172] Before selection Regions , output push tuple:

[0173] , ;

[0174] The final intelligent push instruction is formed, and the hotspot detection-push process is completed in a closed loop.

[0175] By calculating the mean and standard deviation of each hotspot, the hotspots are ranked and a number of high-priority areas are selected to form the final push list. This ranking mechanism incorporates a quantitative evaluation system, enabling hotspot push notifications to be not only based on location but also comparable in terms of data strength. The push list has a clear structure, making it easy for various management systems to directly access applications. This closed-loop management model, from data mining to task dispatch, significantly enhances the system's intelligence and practicality, and has broad application prospects in a variety of fields, including urban management, traffic emergency response, and environmental warning.

[0176] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0177] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. An intelligent push processing method based on GIS city data mining, characterized in that: include: Set the study area and initial grid division parameters, build a grid set and determine the observation indicators; Calculate the mean and standard deviation of the observation values ​​of all grids, and set the quadtree decomposition threshold and maximum decomposition depth; Construct the root node of the quadtree, set the region boundaries, observation mean and standard deviation; Recursively decompose the nodes based on the set threshold and observed data, generate child nodes and update indexes and statistics until the termination condition is met; Collect all nodes that are no longer decomposed to form a set of leaf nodes; Filter leaf nodes that meet the preset hierarchical conditions as hotspot candidates; Perform connected component search based on the adjacency matrix and merge spatially adjacent hotspot candidate nodes; Prioritize the merged hotspot areas and generate a push list.

2. The intelligent push processing method based on GIS city data mining according to claim 1 is characterized in that: The setting of the study area and initial grid division parameters, the construction of the grid set and the determination of the observation indicators specifically include: Assume that the study area is a square area , the side length is ; Let the minimum leaf node edge length be ,and ; Assume the initial number of grid rows and columns is , then the actual grid side length is ; Will Divided equally grids of the same area, forming a grid set : , ;in, For the grids; is the total number of grids; For each Set line number , column number , whose areas are: ;in, For grid The two-dimensional coordinate range of ; is the remainder operation; Set the time window to the past hour; Get the taxi GPS device in the grid within the time window The total number of reported positioning points in the area is recorded as the observation value .

3. The intelligent push processing method based on GIS city data mining according to claim 2 is characterized in that: The calculation of the mean and standard deviation of the observation values ​​of all grids and setting the quadtree decomposition threshold and maximum decomposition depth specifically includes: Calculates the arithmetic mean of all raster observations , ; Calculate the global standard deviation , ; Set the quadtree decomposition threshold to ; Set the maximum decomposition depth of the quadtree to .

4. The intelligent push processing method based on GIS city data mining according to claim 3 is characterized in that: The construction of the root node of the quadtree, setting the region boundary, observation mean and standard deviation, specifically includes: Let the root node of the quadtree be , ;in, For nodes Depth in the tree, root node ; Let the root node region boundary be , ;in, For nodes Left and right borders coordinate; For nodes Upper and lower boundaries Coordinates here is the root node; Set the root node grid index set to , ;in, For nodes Grid number; To include the area The collection of all raster indexes in; Set the observation mean and observation standard deviation in the root node area to be , ;in, For nodes Mean of observations within the region; is the standard deviation of observations within the node area.

5. The intelligent push processing method based on GIS city data mining according to claim 4 is characterized in that: The node is recursively decomposed based on the set threshold and observation data, child nodes are generated, and indexes and statistics are updated until the termination condition is met, specifically including: For any node , initially , perform the following steps: S101, determine whether to decompose: like , then decompose; otherwise mark it as a leaf node and return; S102. Calculate the node midpoint coordinates: , ;in, For the current node The coordinates of the center point of the region; S103. Generate four child nodes: ； ； ； ； in, For nodes No. child nodes; And order ; S104, child node grid index: For each Recalculate , ;in, is the number of elements in the combination; like , then discard the child node; otherwise continue; S105. Child node statistics: ， ； For each make Repeat steps S101 to S105 for the child nodes of .

6. The intelligent push processing method based on GIS city data mining according to claim 5 is characterized in that: The collecting of all nodes that are no longer decomposed to form a leaf node set specifically includes: Collect all nodes that are not further decomposed as leaf nodes: ;in, is the set of all leaf nodes that are no longer decomposed.

7. The intelligent push processing method based on GIS city data mining according to claim 6 is characterized in that: The screening of leaf nodes that meet preset hierarchical conditions as hotspot candidates specifically includes: Select the node with the maximum level from the leaf nodes as the hotspot candidate: ;in, It is the set of candidate hotspot nodes.

8. The intelligent push processing method based on GIS city data mining according to claim 7 is characterized in that: The connected component search based on the adjacency matrix and the merging of spatially adjacent hotspot candidate nodes specifically include: Setting up the matrix ,in , for any candidate node : ； in, is a matrix element; Connected component search: S210, initialize the unvisited set , connected component list ; S220, when : S221, from Take a node , let the current component set ,queue , and from Removed; S222, when , pop up node ; for each like , then join in and and from Remove ;in, is the current node and adjacent nodes in the traversal process; S223, the amount to be completed join in ; S230, at the end ; For each component , calculate the The collection range of the merged hotspot areas All raster index collections contained in the area : ， 。 9. The intelligent push processing method based on GIS city data mining according to claim 8 is characterized in that: Prioritizing the merged hotspot areas and generating a push list specifically includes: Calculate the The average value of hotspot areas and standard deviation : , ;in, For collection The cardinality of according to Arrange all merged areas from largest to smallest; Before selection Regions , output push tuple: ， 。

Images