Urban space layout planning auxiliary method and system based on big data mining

Abstract

The invention relates to big data processing, and particularly discloses an urban spatial layout planning auxiliary method and system based on big data mining, which deeply integrates static function attributes represented by urban point of interest (POI) data and a dynamic use mode disclosed by rasterized time sequence population activity data. And constructing a static function-dynamic activity joint coding feature for each urban grid unit. On the basis, the baseline level of each unit feature is learned by further utilizing spatial dynamic and static aggregation analysis, and intelligent weight distribution and information aggregation are performed according to the dynamic change and spatial proximity relation, so that the complex association and dynamic influence among urban space units are effectively captured; and aggregated coding representation capable of macroscopically representing the overall spatial features of the target city region is generated. And finally, based on the highly-condensed and information-rich aggregation coding representation, realizing accurate classification of the urban functional areas.

Description

Technical Field

[0001] This application relates to big data processing, and more specifically, to a method and system for assisting urban spatial layout planning based on big data mining. Background Technology

[0002] With the acceleration of urbanization and the continuous expansion of urban scale, the complexity of urban spatial structures is increasing, placing higher demands on the scientific rigor and precision of urban spatial planning. Traditional urban planning methods often rely on human experience, small-scale sampling surveys, and static land use data, which suffer from problems such as long cycles, high costs, strong subjectivity, and difficulty in reflecting dynamic urban changes in real time. This can lead to planning results that are out of touch with actual urban operations and residents' needs, affecting urban operational efficiency and residents' quality of life. Therefore, constructing a system that can efficiently, objectively, and dynamically assist in urban spatial planning has become an urgent need for sustainable urban development.

[0003] To address these challenges, current efforts are being made to leverage big data technologies to assist urban planning. For example, some methods utilize Points of Interest (POI) data to identify urban functional zones; however, POI data primarily reflects the static functional attributes of urban space, failing to capture its actual usage and dynamic vitality at different times. Other methods utilize population activity data (such as mobile phone signaling data and transit card data) to analyze urban dynamics, but population activity data alone cannot reveal the specific functional attributes of particular areas. For instance, a high-density population area might be a commercial center, a transportation hub, or a large park, lacking functional specificity. These existing methods often rely heavily on a single data source or fail to fully integrate the advantages of multi-source data, resulting in an incomplete and in-depth understanding of urban spatial functions and an inability to accurately depict the interaction between static functions and dynamic activities within complex urban systems.

[0004] Therefore, an optimized urban spatial layout planning auxiliary scheme is desired. Summary of the Invention

[0005] To address the aforementioned technical problems, this application is proposed. Embodiments of this application provide a method and system for assisting urban spatial layout planning based on big data mining. It deeply integrates the static functional attributes represented by points of interest (POI) data with the dynamic usage patterns revealed by rasterized time-series population activity data, constructing a static function-dynamic activity joint coding feature for each urban grid unit that encompasses both its inherent functional potential and reflects its actual dynamic vitality. This joint coding can more comprehensively and accurately characterize the essential features of basic urban units. Furthermore, considering that the formation of urban spatial functions depends not only on the unit's own attributes but also on the profound influence of the surrounding environment and spatial interactions, spatial dynamic-static aggregation analysis is further utilized to learn the baseline level of each unit's characteristics. Intelligent weight allocation and information aggregation are then performed based on their dynamic changes and spatial proximity relationships, effectively capturing the complex relationships and dynamic influences between urban spatial units and generating an aggregated coding representation that macroscopically characterizes the overall spatial features of the target urban area. Finally, based on this highly condensed and information-rich aggregated coding representation, accurate classification of urban functional areas is achieved, providing more scientific, dynamic, and refined decision support for urban spatial layout planning, effectively improving the rationality and foresight of the planning.

[0006] According to one aspect of this application, a method for assisting urban spatial layout planning based on big data mining is provided, comprising: Obtain a collection of POI data within the target city area; Obtain rasterized time-series population activity data of the target urban area, wherein the rasterized time-series population activity data includes the number of residents in each grid cell of the target urban area at different time slices; The rasterized time-series population activity data is subjected to time-series analysis to obtain a set of feature encoding vectors for the temporal pattern of population activity in the grid cells; The collection of POI data is divided into clusters according to grid cells to obtain a collection of grid cell POI data clusters; Static functional features are extracted from each grid cell POI data cluster in the set of grid cell POI data clusters to obtain a set of static functional feature encoding vectors for grid cell POI clusters; The set of static functional feature encoding vectors of the grid unit POI cluster and the set of temporal pattern feature encoding vectors of the grid unit population activity are combined to obtain the set of joint encoding vectors of static function and dynamic activity of the grid unit. Spatial dynamic-static aggregation analysis is performed on the set of joint coding feature vectors of static function and dynamic activity of the grid unit to obtain the spatial dynamic-static feature aggregation coding vector of the target urban area; Based on the aggregated encoding vector of the spatial dynamic and static features of the target urban area, the target urban area is classified into functional zones to obtain functional zone category labels.

[0007] According to another aspect of this application, a city spatial layout planning auxiliary system based on big data mining is provided, comprising: The POI data acquisition module is used to acquire a collection of POI data within the target city area; The grid-based time-series population data acquisition module is used to acquire rasterized time-series population activity data of the target urban area, wherein the rasterized time-series population activity data includes the number of residents in each grid cell of the target urban area at different time slices; The population activity data time series analysis module is used to perform population activity data time series analysis on the rasterized time series population activity data to obtain a set of feature encoding vectors of population activity time series patterns in grid cells; The grid cell cluster partitioning module is used to partition the set of POI data into clusters according to grid cells to obtain a set of grid cell POI data clusters; The grid cell static functional feature extraction module is used to extract static functional features from each grid cell POI data cluster in the set of grid cell POI data clusters to obtain a set of static functional feature encoding vectors for grid cell POI clusters. The grid cell static function-dynamic activity joint coding module is used to combine the static function feature coding vectors of the grid cell POI cluster and the population activity time-series pattern feature coding vectors of the grid cell to obtain the set of grid cell static function-dynamic activity joint coding feature coding vectors. The grid cell spatial dynamic and static aggregation analysis module is used to perform spatial dynamic and static aggregation analysis on the set of joint coding feature coding vectors of static function and dynamic activity of the grid cell to obtain the spatial dynamic and static feature aggregation coding vector of the target urban area; The functional area classification module is used to aggregate the encoding vector based on the spatial dynamic and static features of the target city area, and classify the target city area into functional areas to obtain functional area category labels.

[0008] Compared with existing technologies, this application provides a method and system for assisting urban spatial layout planning based on big data mining. It deeply integrates the static functional attributes represented by points of interest (POI) data with the dynamic usage patterns revealed by rasterized time-series population activity data to construct a static function-dynamic activity joint coding feature for each urban grid unit. This joint coding can more comprehensively and accurately characterize the essential features of basic urban units. Furthermore, considering that the formation of urban spatial functions depends not only on the unit's own attributes but also on the profound influence of the surrounding environment and spatial interactions, it further utilizes spatial dynamic-static aggregation analysis to learn the baseline level of each unit's characteristics. Based on its dynamic changes and spatial proximity relationships, it performs intelligent weight allocation and information aggregation, thereby effectively capturing the complex relationships and dynamic influences between urban spatial units and generating an aggregated coding representation that can macroscopically represent the overall spatial characteristics of the target urban area. Finally, based on this highly condensed and information-rich aggregated coding representation, it achieves accurate classification of urban functional areas, providing more scientific, dynamic, and refined decision support for urban spatial layout planning, effectively improving the rationality and foresight of the planning. Attached Figure Description

[0009] The above and other objects, features, and advantages of this application will become more apparent from the more detailed description of the embodiments of this application in conjunction with the accompanying drawings. The drawings are provided to further illustrate the embodiments of this application and form part of the specification. They are used together with the embodiments of this application to explain this application and do not constitute a limitation thereof. In the drawings, the same reference numerals generally represent the same components or steps.

[0010] Figure 1 This is a flowchart of an urban spatial layout planning auxiliary method based on big data mining according to an embodiment of this application; Figure 2 This is a data flow diagram of the urban spatial layout planning auxiliary method based on big data mining according to an embodiment of this application; Figure 3 This is a flowchart illustrating the process of extracting static functional features from each grid cell POI data cluster in the set of grid cell POI data clusters according to an embodiment of this application to obtain a set of static functional feature encoding vectors for grid cell POI clusters. Figure 4 This is a flowchart illustrating the process of performing spatial dynamic and static aggregation analysis on the set of joint coding feature vectors of static function and dynamic activity of grid units to obtain the aggregated coding vector of spatial dynamic and static features of the target urban area, according to the big data mining-based urban spatial layout planning auxiliary method of this application embodiment. Figure 5This is a block diagram of an urban spatial layout planning auxiliary system based on big data mining, according to an embodiment of this application. Detailed Implementation

[0011] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0012] As indicated in this application and claims, unless the context clearly indicates otherwise, the words "a," "an," "an," and / or "the" are not specifically singular and may include plural forms. Generally speaking, the terms "comprising" and "including" only indicate the inclusion of explicitly identified steps and elements, which do not constitute an exclusive list, and the method or apparatus may also include other steps or elements.

[0013] While this application makes various references to certain modules of the systems according to embodiments of this application, any number of different modules can be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0014] Flowcharts are used in this application to illustrate the operations performed by the system according to embodiments of this application. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0015] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0016] This application proposes a big data mining-based method to assist in urban spatial layout planning, aiming to overcome the problems of strong subjectivity, outdated updates, and one-sided understanding of urban spatial functions in traditional planning methods and existing big data analysis methods. Specifically, it deeply integrates the static functional attributes represented by Points of Interest (POI) data with the dynamic usage patterns revealed by rasterized time-series population activity data to construct a static function-dynamic activity joint coding feature for each urban grid unit. This joint coding can more comprehensively and accurately characterize the essential features of basic urban units. Furthermore, considering that the formation of urban spatial functions depends not only on the unit's own attributes but also on the profound influence of the surrounding environment and spatial interactions, a spatial dynamic-static aggregation analysis mechanism is further used to learn the baseline level of each unit's characteristics and intelligently allocate weights and aggregate information based on their dynamic changes and spatial proximity relationships. This effectively captures the complex relationships and dynamic influences between urban spatial units, generating an aggregated coding vector that macroscopically represents the overall spatial characteristics of the target urban area. Ultimately, based on this highly condensed and information-rich aggregated coding vector, the precise classification of urban functional areas is achieved, providing more scientific, dynamic, and refined decision support for urban spatial layout planning, and effectively improving the rationality and foresight of the planning.

[0017] The technical solution of this application proposes an auxiliary method for urban spatial layout planning based on big data mining. Figure 1 This is a flowchart of an urban spatial layout planning auxiliary method based on big data mining, according to an embodiment of this application. Figure 2 This is a data flow diagram illustrating the urban spatial layout planning assistance method based on big data mining according to an embodiment of this application. Figure 1 and Figure 2As shown, the urban spatial layout planning assistance method based on big data mining according to an embodiment of this application includes the following steps: S100, obtaining a set of POI data within a target urban area; S200, obtaining rasterized time-series population activity data of the target urban area, wherein the rasterized time-series population activity data includes the number of residents in each grid cell of the target urban area at different time slices; S300, performing time-series analysis on the rasterized time-series population activity data to obtain a set of grid cell population activity time-series pattern feature encoding vectors; S400, dividing the set of POI data into clusters according to grid cells to obtain a set of grid cell POI data clusters; S500, extracting static functional features from each grid cell POI data cluster in the set of grid cell POI data clusters to obtain a set of grid cell POI cluster static functional feature encoding vectors; S6 S700: Combine the static functional feature encoding vectors of the POI clusters of the grid units and the temporal pattern feature encoding vectors of the population activities of the grid units in the set of static functional feature encoding vectors of the POI clusters of the grid units and the temporal pattern feature encoding vectors of the population activities of the grid units to obtain a set of static functional-dynamic activity joint encoding feature encoding vectors of the grid units; S800: Perform spatial dynamic-static aggregation analysis on the set of static functional-dynamic activity joint encoding feature encoding vectors of the grid units to obtain spatial dynamic-static feature aggregation encoding vectors of the target urban area; S800: Based on the spatial dynamic-static feature aggregation encoding vectors of the target urban area, classify the target urban area into functional zones to obtain functional zone category labels.

[0018] Specifically, in steps S100 and S200, a set of POI data within the target city area is acquired, along with rasterized time-series population activity data for the target city area. This rasterized time-series population activity data includes the number of residents in each grid cell of the target city area at different time slices. It should be understood that POI data within the target city area, such as shops, restaurants, office buildings, and parks, can reveal the static functional attributes of urban space, i.e., the intended use or service type assigned to a specific geographical location, which is the foundation for understanding urban land use and functional layout. Rasterized time-series population activity data, especially the number of residents in each grid cell at different time slices, reflects the dynamic intensity and temporal changes of urban space use, revealing the actual aggregation, distribution, and flow patterns of population in urban space, and reflecting the true vitality of the city. By combining these two heterogeneous data sources, the characteristics of urban spatial units can be comprehensively characterized from both "static potential" and "dynamic performance" dimensions, compensating for the limitations of a single data source in urban function identification, and providing a solid data foundation for subsequent more accurate urban spatial layout planning analysis. It is worth noting that here, the time slice is per hour.

[0019] Specifically, in a concrete example of this application, firstly, the target city area is defined and rasterized. The city area under study is clearly defined and divided into grid units of uniform size, such as 500m × 500m or 1km × 1km. These grid units will serve as the basic spatial units for subsequent data aggregation and analysis. Secondly, POI data is acquired and processed. POI data within the target city area is obtained through map service provider APIs (such as Gaode Maps, Baidu Maps, etc.), open data platforms (such as OpenStreetMap), or publicly available government data. This data typically includes information such as the POI's name, category (such as restaurants, shopping, companies, medical facilities, etc.), and latitude and longitude coordinates. After acquisition, the POI data needs to be cleaned, deduplicated, classified, and standardized. Each POI point is then matched to the corresponding grid unit based on its geographic coordinates, forming a POI data set within each grid unit.

[0020] Next, the acquisition and processing of rasterized time-series population activity data is performed. This type of data typically originates from mobile signaling data from mobile communication operators, LBS (Location Based Services) data from location service providers, or public transportation card swipe data, and undergoes anonymization and aggregation processing. After data acquisition, it needs to be spatially matched with preset grid cells to count the number or density of people residing in each grid cell within each preset time slice (e.g., hourly), thus forming a sequence of population occupancy in each grid cell over a continuous time period. For example, a data processing system can be used. This system receives raw, anonymized user data with timestamps and geographic locations (or base station information), maps user locations to grid cells through spatial interpolation or attribution algorithms, and counts the number of people residing in each grid cell at specified time intervals (e.g., hourly), outputting time-series data in the format of (grid ID, time slice, number of people residing).

[0021] Specifically, in step S300, time series analysis of the rasterized time-series population activity data is performed to obtain a set of feature encoding vectors for the temporal pattern of population activity in the grid units. It should be understood that from the raw, high-dimensional, and dynamically changing population residence data, the semantically meaningful dynamic activity patterns unique to each urban grid unit are extracted and condensed. Although the raw population residence time-series data reflects the dynamic distribution of the population, its direct form is too large and contains a lot of noise, making it difficult to directly use for urban function identification and comparison. Through time series analysis, especially using deep learning models, it is possible to effectively capture the complex patterns of periodicity, trends, and suddenness of population activity in different grid units at daily, weekly, or even longer time scales, such as dense daytime pedestrian traffic in commercial areas, concentrated nighttime pedestrian traffic in residential areas, and high weekend pedestrian traffic in entertainment areas. Encoding these dynamic patterns into compact feature vectors not only significantly reduces the data dimensionality, but more importantly, these vectors can serve as a quantitative representation of the "dynamic function" of the grid unit, providing key input for subsequent fusion with static functional features and the identification of urban functional areas.

[0022] More specifically, in this embodiment, performing time-series analysis on the rasterized time-series population activity data to obtain a set of feature encoding vectors for the temporal pattern of population activity in each grid cell includes: performing time-series encoding of the number of residents in each grid cell at different time slices using an LSTM model to obtain a set of feature encoding vectors for the temporal pattern of population activity in each grid cell. More specifically, firstly, the rasterized time-series population activity data for each grid cell is preprocessed. This includes data cleaning, missing value imputation (e.g., using interpolation or forward / backward imputation), and normalization to eliminate the influence of population size differences between different grid cells, ensuring that the data has a consistent scale and quality before being input into the model. Secondly, a model based on a Long Short-Term Memory (LSTM) network is constructed and trained. LSTM models are well-suited for pattern recognition of population activity time series due to their advantages in processing sequential data and capturing long-term dependencies. Specifically, for each grid cell, the sequence of residents at different time slices (e.g., hourly) is used as input to the LSTM model. This LSTM model can be designed as an encoder-decoder structure, where the encoder (one or more LSTM layers) is responsible for compressing the input time series into a fixed-length hidden state vector, while the decoder attempts to reconstruct the original time series from this hidden state vector. Through this self-supervised learning approach, the encoder is trained to learn the most representative temporal patterns in the input sequence. Finally, after model training, the hidden state vector or final cell state vector output by the trained LSTM encoder for each grid cell is extracted as the population activity temporal pattern feature encoding vector for that grid cell. These vectors are low-dimensional, abstract, and semantically rich representations of the original population activity time series; they capture the typical patterns and regularities of population activity in that grid cell over time, thus forming a set of grid cell population activity temporal pattern feature encoding vectors, laying the foundation for subsequent feature fusion and urban functional area classification.

[0023] Specifically, in step S400, the set of POI data is clustered according to grid units to obtain a set of POI data clusters for each grid unit. It should be understood that POIs in a city are densely and diversely distributed, with a single grid unit potentially containing dozens or even hundreds of POIs of different types and attributes. Directly processing massive amounts of raw POI data is not only computationally expensive but also fails to intuitively reflect the comprehensive functions of an area. Therefore, further clustering the set of POI data according to grid units "divides" or assigns the POI data to various grid units, forming a "POI data cluster" within each grid unit (i.e., the set of all POIs within that grid unit). This enables spatial aggregation of data, ensuring that each grid unit has a list of POIs describing its internal static facility configuration. This lays the foundation for subsequent quantitative evaluation of the static functions of the grid unit (such as dominant or mixed functions like commerce, office, residence, and leisure), and is a key step in understanding the heterogeneity of static functions in urban space.

[0024] More specifically, in a particular example of this application, firstly, it is ensured that the target urban area has been divided into standard grid cells, each with a unique identifier and a clearly defined geographical boundary.

[0025] Secondly, the acquired POI data undergoes preprocessing, including cleaning (removing invalid or duplicate data), geocoding (ensuring each POI has accurate latitude and longitude coordinates), and type standardization (mapping POI types from different sources or with varying levels of detail to a unified classification system, for example, unifying "fast food restaurants," "Chinese restaurants," and "Western restaurants" into "food service"). Thirdly, spatial matching and attribution are performed. Each POI data point is traversed, and its latitude and longitude coordinates are used to determine which grid cell's geographic range it falls into. All POI data falling into the same grid cell are grouped together to form a POI data cluster for that grid cell. This process can be efficiently accomplished using the spatial join function of a Geographic Information System (GIS). For example, a spatial index (such as an R-tree or Quadtree) can be built to accelerate the point lookup process within polygons. For each POI, its grid cell ID is queried, and it is added to the POI list corresponding to that ID. Ultimately, each grid cell corresponds to a set containing detailed information about all POIs within it (such as type, name, etc.), and this set is called the "POI data cluster" of that grid cell. After this series of operations is completed, a set of grid cell POI data clusters covering the entire target city area is obtained, providing structured input for subsequent static functional feature extraction.

[0026] Specifically, in step S500, static functional features are extracted from each grid cell POI data cluster in the set of grid cell POI data clusters to obtain a set of static functional feature encoding vectors for the grid cell POI clusters. It should be understood that although the original grid cell POI data clusters (i.e., the list of all POIs within a grid cell) contain rich static functional information, their forms are inconsistent (various POI numbers and types), making them difficult to directly use for subsequent machine learning model analysis or to effectively integrate with other features (such as dynamic activity features). Through static functional feature extraction, the POI composition of each grid cell (e.g., the density of commercial facilities, the completeness of public service facilities, the proportion of residential areas, etc.) can be abstracted into a feature vector of a unified format. This vector is used to summarize and characterize the static functional attributes of the grid cell, capturing the main or mixed static functional tendencies of the area, laying the foundation for subsequent identification of urban functional zones and understanding of urban spatial structure.

[0027] Figure 3 This is a flowchart illustrating the process of extracting static functional features from each grid cell POI data cluster in the set of grid cell POI data clusters to obtain a set of static functional feature encoding vectors for the grid cell POI clusters, according to the big data mining-based urban spatial layout planning auxiliary method of this application. Figure 3 As shown, according to the embodiment of this application, the urban spatial layout planning assistance method based on big data mining includes step S500, which includes: S510, vectorizing each POI data in the grid unit POI data cluster to obtain a set of POI encoding vectors; S520, calculating the position mean vector of the set of POI encoding vectors to obtain the static functional feature encoding vector of the grid unit POI cluster.

[0028] Specifically, in step S600, the set of static functional feature encoding vectors for the grid unit POI cluster and the set of temporal pattern feature encoding vectors for the grid unit population activity are combined to obtain a set of joint encoding vectors for static function and dynamic activity of the grid unit. It should be understood that static functional features extracted solely from POI data can only describe what facilities an area "has," i.e., its potential functional attributes; while temporal pattern features extracted solely from population activity data can only describe how an area is "used," i.e., its dynamic vitality. However, the actual function of urban space is the result of the interaction and mutual verification between static facility layout and dynamic population activity. For example, an area with a large number of commercial POIs may not fully realize its true commercial function if population activity is sparse; conversely, an area with no prominent POI type may indicate a specific function not directly reflected by the POIs if it exhibits regular population aggregation over a specific time period. Therefore, combining these two types of features can effectively compensate for the shortcomings of a single feature source, overcome the possible bias or information gaps between "static potential" and "dynamic performance", and generate a joint feature that is richer in information and more distinguishable, which includes both physical facility configuration information and actual usage patterns, thus providing a solid foundation for a more accurate understanding and division of urban functional areas.

[0029] More specifically, in this embodiment of the application, the set of static functional feature encoding vectors of the grid unit POI cluster and the set of population activity temporal pattern feature encoding vectors of the grid unit are combined to obtain a set of static function-dynamic activity joint encoding feature vectors of the grid unit. This includes: concatenating the static functional feature encoding vectors of the grid unit POI cluster and the population activity temporal pattern feature encoding vectors of the grid unit to obtain the static function-dynamic activity joint encoding feature vectors of the grid unit.

[0030] Specifically, in step S700, a spatial dynamic-static aggregation analysis is performed on the set of static function-dynamic activity joint coding feature vectors of the grid units to obtain the spatial dynamic-static feature aggregated coding vector of the target urban area. It should be understood that while the static function-dynamic activity joint coding features of a single grid unit can finely characterize the features of urban micro-units, when conducting regional-level urban spatial layout planning, it is necessary to understand and represent the overall spatial characteristics and functional pattern of the entire target urban area from a higher dimension. Simply averaging or splicing the joint coding features of all grid units makes it difficult to capture the complex interactions between different units within the region, spatial heterogeneity, and the nonlinear contribution of key units to the overall characteristics of the region. The overall characteristics of an urban area are not a simple summation of the features of its internal units, but a complex emergence determined by the features of each internal unit and their interrelationships. Therefore, in the technical solution of this application, a spatial dynamic-static aggregation analysis is further performed on the set of static function-dynamic activity joint coding feature vectors of the grid units to obtain the spatial dynamic-static feature aggregated coding vector of the target urban area. This aggregated encoding vector aims to capture macro-level feature patterns at the regional level, such as the dominant function, functional mixing, distribution of activity centers, and spatial configuration relationships between different functional units. Specifically, it first learns a feature baseline representing the "commonality" or "average level" of grid unit features within the target region. Then, it evaluates the "deviation" or "uniqueness" of each grid unit's joint encoding features relative to this baseline, and uses a Softmax mechanism to transform these deviations into dynamic, normalized importance weights. This mechanism intelligently identifies and emphasizes grid units that contribute more to the overall regional characteristics and have richer information, along with their static function-dynamic activity joint encoding features, while suppressing the influence of noise or insignificant variations. By incorporating the relative importance and uniqueness of each unit's features through a dynamic compensation mechanism, the aggregation results better reflect the region's true functional structure and dynamic activity pattern. For example, the overall business atmosphere of a region may depend not only on the average density of business POIs but also on the leading role of a few core business nodes (whose features deviate significantly from the baseline). This network can capture such non-uniform contributions, resulting in an aggregated encoding vector that more accurately reflects the region's business level. Ultimately, this high-quality regional aggregated feature vector will provide a solid foundation for more accurate urban functional zone classification, planning evaluation, and decision support, as it offers a deeper and more adaptive understanding of the complexity of urban areas.

[0031] Figure 4This is a flowchart illustrating the spatial dynamic and static aggregation analysis of the set of joint coding feature vectors of static functions and dynamic activities of grid units in the urban spatial layout planning auxiliary method based on big data mining, according to an embodiment of this application, to obtain the aggregated coding vector of spatial dynamic and static features of the target urban area. (See flowchart for example.) Figure 4 As shown, according to the embodiment of this application, the urban spatial layout planning assistance method based on big data mining includes step S700, which includes: S710, performing feature baseline learning on the set of grid unit static function-dynamic activity joint coding feature encoding vectors to obtain grid unit static function-dynamic activity joint coding feature baseline regression encoding vectors; S720, calculating the dynamic compensation information of the set of grid unit static function-dynamic activity joint coding feature encoding vectors relative to the grid unit static function-dynamic activity joint coding feature baseline regression encoding vectors to obtain the set of grid unit static function-dynamic activity joint coding feature dynamic compensation weight factors; S730, based on the set of grid unit static function-dynamic activity joint coding feature dynamic compensation weight factors and the set of grid unit static function-dynamic activity joint coding feature encoding vectors, performing dynamic compensation on the grid unit static function-dynamic activity joint coding feature baseline regression encoding vectors to obtain the target urban area spatial dynamic and static feature aggregate encoding vector.

[0032] Specifically, in step S710, feature baseline learning is performed on the set of joint static function-dynamic activity coding feature encoding vectors of the grid cells to obtain the grid cell static function-dynamic activity coding feature baseline regression encoding vector, expressed by the formula: , , in, This is a set of joint encoding vectors for static function and dynamic activity of grid cells. These are the 1st, 2nd, and 3rd elements in the set of joint encoding feature vectors of the static function-dynamic activity of the grid cell. The and the first The static function-dynamic activity joint encoding feature encoding vector of each grid cell and These are the learnable parameter matrix and the learnable bias vector, respectively. for Activation function The baseline regression encoding vector is the joint encoding feature of static function-dynamic activity of grid cells.

[0033] It is understandable that when analyzing the overall spatial dynamic and static characteristics of an urban area, a stable and representative "reference frame" or "average level" is needed to measure the characteristics of each grid unit within the area. While the static function-dynamic activity joint coding feature vector of each grid unit already contains rich information, differences exist between these vectors, and direct aggregation may lead to information confounding or dilution of key patterns. To better understand and quantify the contribution of each grid unit to the overall characteristics of the area, and to identify those units with significant deviations or lack of representativeness, establishing a "baseline" is crucial. This baseline can be considered a "common pattern" or "typical state" of the grid unit characteristics within the area. From the static function-dynamic activity joint coding feature vectors of all grid units within the target urban area, a "grid unit static function-dynamic activity joint coding feature baseline regression coding vector" is extracted to characterize the commonalities, trends, or statistical central trends of these feature sets. This baseline vector aims to capture the general patterns of grid unit characteristics within the area. For example, in an area primarily focused on residential functions, the joint coding characteristics of its grid units may exhibit similar patterns in terms of nighttime population activity and residential POIs; this similarity is what the baseline aims to capture. By learning this baseline, a stable reference point can be provided for subsequent dynamic compensation and aggregation, allowing subsequent analysis to focus on the "uniqueness" or "dynamic deviation" of each unit relative to this baseline, rather than simply dealing with the original absolute eigenvalues.

[0034] Accordingly, according to the embodiments of this application, step S720, calculating the dynamic compensation information of the set of grid cell static function-dynamic activity joint coding feature encoding vectors relative to the grid cell static function-dynamic activity joint coding feature baseline regression encoding vector to obtain the set of grid cell static function-dynamic activity joint coding feature dynamic compensation weight factors, includes: calculating the feature dynamic compensation factor of each grid cell static function-dynamic activity joint coding feature encoding vector in the set of grid cell static function-dynamic activity joint coding feature encoding vectors relative to the grid cell static function-dynamic activity joint coding feature baseline regression encoding vector to obtain the set of grid cell static function-dynamic activity joint coding feature dynamic compensation factors; and performing regularization processing based on the Softmax activation function on the set of grid cell static function-dynamic activity joint coding feature dynamic compensation factors to obtain the set of grid cell static function-dynamic activity joint coding feature dynamic compensation weight factors.

[0035] Specifically, the set of dynamic compensation factors for the static function-dynamic activity joint coding feature vectors of each grid cell in the set of grid cell static function-dynamic activity joint coding feature vectors is obtained by calculating the feature dynamic compensation factor of each grid cell static function-dynamic activity joint coding feature vector relative to the grid cell static function-dynamic activity joint coding feature baseline regression coding vector, and expressed by the formula: , in, Let L be the L2 norm of the vector. For positional product, for function, The set of dynamic compensation factors for the joint coding features of static function and dynamic activity of grid cells. The static function-dynamic activity joint coding feature of each grid cell is a dynamic compensation factor.

[0036] It is understandable that after establishing a "baseline" of regional characteristics, it is necessary to quantify the differences or deviations between the characteristics of each specific grid unit and this baseline. The baseline represents the common or average pattern of the region, but urban space is highly heterogeneous, and each grid unit has its unique static function and dynamic activity pattern. Simply knowing the baseline is insufficient; it is more important to understand the degree and direction of each unit's "deviation" from the baseline. These deviations are key information constituting the complexity and diversity of the region. One or a set of feature dynamic compensation factors are generated for each grid unit. These factors can accurately quantify the unique information, dynamic changes, or contributions of the grid unit's static function-dynamic activity jointly encoded features relative to the overall regional baseline. These factors do not simply represent "differences," but are designed to capture those feature dimensions with information gain that are not fully covered by the baseline. For example, if the baseline represents a typical residential area pattern, then the feature vector of a commercial center grid unit will deviate significantly from the baseline. These deviations need to be quantified as compensation factors to reflect its unique commercial function. Furthermore, based on the corpus, the calculation of this deviation is ensured to be "smooth mapping induced" by introducing "displacement gradient" and "metric correlation change matrix". This means that even small feature variations can be accurately captured, and this capture is stable and will not fluctuate drastically due to small noise, thus ensuring the universality and accuracy of the compensation factor.

[0037] Specifically, the set of dynamic compensation factors for the joint encoding features of static function and dynamic activity of the grid cell is regularized based on the Softmax activation function to obtain the set of dynamic compensation weight factors for the joint encoding features of static function and dynamic activity of the grid cell, which is expressed by the formula: , in, For preset parameter values, Let e ​​be the value of an exponential function with the natural constant e as its base. The set of dynamic compensation weight factors for the joint encoding features of static function and dynamic activity of grid cells. The static function-dynamic activity joint coding feature of each grid cell is dynamically compensated by a weight factor.

[0038] It is understandable that although the unique deviation of each grid cell relative to the region baseline has been quantified (i.e., the dynamic compensation factor of the grid cell's static function-dynamic activity joint encoding feature), these original factors may have varying numerical ranges and lack the property of being directly used as "weights" for weighted aggregation (e.g., they may not be non-negative, or their sum may not be 1). To allocate the importance of each grid cell's unique information in a reasonable, comparable, and competitive manner during subsequent dynamic compensation, these compensation factors need to be standardized and normalized. The Softmax activation function plays a crucial role here, mapping any real-valued input vector to a probability distribution vector that sums to 1, ensuring that each output weight factor is non-negative. More importantly, Softmax introduces a "competition mechanism," ensuring that grid cells with larger original compensation factor values ​​receive relatively higher weights after Softmax processing, while grid cells with smaller values ​​are assigned lower weights. This enables the network to adaptively identify and "focus" on feature deviations that contribute more to the current regional aggregation task and are richer in information, while effectively suppressing or reducing the impact of noise and irrelevant variations. This gives the network the ability to dynamically allocate feature importance, thereby ensuring that the final generated regional aggregation encoding vector can more effectively integrate the unique contributions of each grid unit, thus improving the robustness and expressiveness of feature representation and enabling it to more accurately reflect the complex spatial dynamic and static characteristics of the target urban area.

[0039] Specifically, in step S730, based on the set of dynamic compensation weight factors of the joint coding features of static functions and dynamic activities of the grid cells and the set of coding vectors of the joint coding features of static functions and dynamic activities of the grid cells, dynamic compensation is performed on the baseline regression coding vector of the joint coding features of static functions and dynamic activities of the grid cells to obtain the aggregated coding vector of the spatial dynamic and static features of the target urban area, expressed by the formula: , in, Aggregate the static and dynamic features of the target city area into an encoded vector.

[0040] It is understandable that after establishing a "baseline" of regional characteristics for each grid cell and quantifying the "unique contribution" and "importance weight" of each grid cell relative to the baseline, the ultimate goal is to generate an aggregated feature vector that can comprehensively represent the entire target urban area. This aggregated vector should not be a simple copy of the baseline, nor should it be an indifferent average of all original features. Instead, it should be a comprehensive representation that reflects both regional commonalities (represented by the baseline) and dynamically incorporates the unique and important information of each cell. Therefore, in the technical solution of this application, a dynamic compensation mechanism is used to organically combine the "common baseline" of the region with the "personalized unique information" and "importance" of each grid cell, thereby generating an "aggregated encoding vector" that can highly summarize the overall spatial dynamic and static characteristics of the target urban area. This dynamic compensation is not a simple linear superposition, but allows the baseline to effectively absorb and integrate grid cell features with significant deviations and information gains that are given high weights while retaining its representation of regional commonalities. For example, if a grid cell has a very prominent commercial function and its corresponding compensation weight is high, then during the compensation process, the commercial characteristic information of that cell will be more strongly integrated into the baseline, so that the final aggregate vector can more accurately reflect the commercial vitality of the area.

[0041] Preferably, here, in evaluating the static function-dynamic activity joint coding feature encoding vector of each grid cell... Compared to the established characteristic baseline When there are differences in the degree of deviation or contribution, the expected grid cell static function-dynamic activity joint coding feature encoding vector is... To characteristic baseline It has a smooth mapping, which enables the static function-dynamic activity joint encoding feature encoding vector of each grid cell. The calculated dynamic compensation weight factor for the joint encoding features of static function and dynamic activity of grid cells has wide applicability. Therefore, the encoding vector of the joint encoding features of static function and dynamic activity of grid cells is first calculated. To characteristic baseline The displacement gradient is obtained by difference mapping, i.e.: , Then, with the displacement gradient vector Construct a matrix of changes in the correlation of measurements , wherein the displacement gradient vector This is a column vector. That is, if the above-mentioned feature encoding vector based on the joint encoding of static function and dynamic activity of grid cells is... To characteristic baseline If the differential term of the distance-related mapping is used as a parameterized mapping, then it will induce the mapping state through the displacement gradient. Therefore, by solving the displacement gradient vector... And construct the metric correlation change matrix under symmetric action. To ensure no deformation, the correlation change matrix can be measured. This determines the stability of the induced reaction.

[0042] Therefore, calculate the metric correlation change matrix. The F-norm, i.e., the low-rank stability characteristic, is used as a stability factor in the calculation of the feature dynamic compensation factor, i.e.: , This enables the joint encoding of static and dynamic functions of grid cells, resulting in a feature encoding vector. To characteristic baseline The smooth mapping induces amplification of small noise or variation disturbances, affecting the universal accuracy of the dynamic compensation weight factor of the joint encoding feature of grid cell static function-dynamic activity.

[0043] Specifically, in step S800, based on the aggregated encoding vector of the static and dynamic spatial features of the target urban area, the target urban area is classified into functional zones to obtain functional zone category labels. It should be understood that although the aforementioned steps have condensed the static functions and dynamic activities of the target urban area into a highly abstract and semantically rich "aggregated encoding vector of the static and dynamic spatial features of the target urban area," this vector itself is still a numerical, high-dimensional representation, lacking direct, human-understandable semantic labels. To transform this deep-level feature understanding into practical urban planning decisions and management basis, this numerical representation must be mapped to specific, clearly defined functional zone categories, such as "Central Business District," "Residential Area," "Industrial Area," "Mixed Commercial Area," or "Ecological Leisure Area," etc. Therefore, in a specific example of this application, the aggregated encoding vector of the static and dynamic spatial features of the target urban area is passed through a classifier-based target urban area functional zone analyzer to obtain functional zone classification results, which are used to represent functional zone category labels. The technical objective of this step is to achieve an intelligent mapping from abstract aggregated encoding vectors of the dynamic and static spatial features of a target urban area to specific functional zone category labels through a trained classifier. This classifier aims to learn and identify the inherent functional patterns of urban areas represented by different aggregated feature vectors and categorize them into predefined functional zone types. Its core objective is to assign a clear functional zone category label to the entire target urban area, thereby providing city managers with a macro-level, accurate, and data-driven understanding of regional functional positioning, supporting more scientific land use planning, infrastructure allocation, and policy formulation.

[0044] In summary, the urban spatial layout planning assistance method based on big data mining, as described in the embodiments of this application, is elucidated. It deeply integrates the static functional attributes represented by points of interest (POI) data with the dynamic usage patterns revealed by rasterized time-series population activity data to construct a static function-dynamic activity joint coding feature for each urban grid unit. This joint coding feature encompasses both the inherent functional potential and reflects the actual dynamic vitality of the unit. This joint coding can more comprehensively and accurately characterize the essential features of basic urban units. Furthermore, considering that the formation of urban spatial functions depends not only on the unit's own attributes but also on the profound influence of the surrounding environment and spatial interactions, spatial dynamic-static aggregation analysis is further utilized to learn the baseline level of each unit's characteristics. Intelligent weight allocation and information aggregation are then performed based on dynamic changes and spatial proximity relationships, effectively capturing the complex relationships and dynamic influences between urban spatial units and generating an aggregated coding representation that macroscopically characterizes the overall spatial features of the target urban area. Finally, based on this highly condensed and information-rich aggregated coding representation, accurate classification of urban functional areas is achieved, providing more scientific, dynamic, and refined decision support for urban spatial layout planning, effectively improving the rationality and foresight of the planning.

[0045] Furthermore, a city spatial layout planning auxiliary system based on big data mining is also provided.

[0046] Figure 5 This is a block diagram of an urban spatial layout planning auxiliary system based on big data mining, according to an embodiment of this application. Figure 5As shown, the urban spatial layout planning auxiliary system 500 based on big data mining according to an embodiment of this application includes: a POI data acquisition module 510, used to acquire a set of POI data within a target urban area; a grid-based time-series population data acquisition module 520, used to acquire rasterized time-series population activity data of the target urban area, wherein the rasterized time-series population activity data includes the number of residents in each grid unit of the target urban area at different time slices; a population activity data time-series analysis module 530, used to perform time-series analysis on the rasterized time-series population activity data to obtain a set of grid unit population activity time-series pattern feature encoding vectors; a grid unit cluster partitioning module 540, used to partition the set of POI data according to grid units to obtain a set of grid unit POI data clusters; and a grid unit static functional feature extraction module 550, used to extract static functional features of each grid unit POI in the set of grid unit POI data clusters. The data clusters undergo static functional feature extraction to obtain a set of static functional feature encoding vectors for grid unit POI clusters; the grid unit static function-dynamic activity joint encoding module 560 is used to combine the static functional feature encoding vectors of the grid unit POI clusters and the grid unit population activity time series pattern feature encoding vectors of each corresponding group in the set of static functional feature encoding vectors of the grid unit POI clusters and the set of population activity time series pattern feature encoding vectors of the grid units to obtain a set of static function-dynamic activity joint encoding feature encoding vectors for the grid units; the grid unit spatial dynamic and static aggregation analysis module 570 is used to perform spatial dynamic and static aggregation analysis on the set of static function-dynamic activity joint encoding feature encoding vectors of the grid units to obtain spatial dynamic and static feature aggregation encoding vectors of the target urban area; the functional area classification module 580 is used to classify the target urban area into functional areas based on the spatial dynamic and static feature aggregation encoding vectors of the target urban area to obtain functional area category labels.

[0047] As described above, the urban spatial layout planning assistance system 500 based on big data mining according to the embodiments of this application can be implemented in various wireless terminals, such as servers with urban spatial layout planning assistance algorithms based on big data mining. In one possible implementation, the urban spatial layout planning assistance system 500 based on big data mining according to the embodiments of this application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the urban spatial layout planning assistance system 500 based on big data mining can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the urban spatial layout planning assistance system 500 based on big data mining can also be one of many hardware modules of the wireless terminal.

[0048] The various embodiments of this disclosure have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principles, practical application, or improvement of the technology in the market, or to enable others skilled in the art to understand the embodiments disclosed herein.

Claims

1. A method for assisting urban spatial layout planning based on big data mining, characterized in that, include: Obtain a collection of POI data within the target city area; Obtain rasterized time-series population activity data of the target urban area, wherein the rasterized time-series population activity data includes the number of residents in each grid cell of the target urban area at different time slices; The rasterized time-series population activity data is subjected to time-series analysis to obtain a set of feature encoding vectors for the temporal pattern of population activity in the grid cells; The collection of POI data is divided into clusters according to grid cells to obtain a collection of grid cell POI data clusters; Static functional features are extracted from each grid cell POI data cluster in the set of grid cell POI data clusters to obtain a set of static functional feature encoding vectors for grid cell POI clusters; The set of static functional feature encoding vectors of the grid unit POI cluster and the set of temporal pattern feature encoding vectors of the grid unit population activity are combined to obtain the set of joint encoding vectors of static function and dynamic activity of the grid unit. Spatial dynamic-static aggregation analysis is performed on the set of joint coding feature vectors of static function and dynamic activity of the grid unit to obtain the spatial dynamic-static feature aggregation coding vector of the target urban area; Based on the aggregated encoding vector of the spatial dynamic and static features of the target urban area, the target urban area is classified into functional zones to obtain functional zone category labels.

2. The urban spatial layout planning auxiliary method based on big data mining according to claim 1, characterized in that, The process of performing time series analysis on the rasterized time series population activity data to obtain a set of feature encoding vectors for the temporal pattern of population activity in grid cells includes: performing time series encoding of the number of residents in each grid cell at different time slices based on an LSTM model to obtain a set of feature encoding vectors for the temporal pattern of population activity in grid cells.

3. The urban spatial layout planning auxiliary method based on big data mining according to claim 1, characterized in that, The time slice is per hour.

4. The urban spatial layout planning auxiliary method based on big data mining according to claim 1, characterized in that, Static functional feature extraction is performed on each grid cell POI data cluster in the set of grid cell POI data clusters to obtain a set of static functional feature encoding vectors for the grid cell POI clusters, including: Vectorize each POI data in the grid cell POI data cluster to obtain a set of POI encoded vectors; Calculate the position-mean vector of the set of POI encoding vectors to obtain the static functional feature encoding vector of the POI cluster of the grid cell.

5. The urban spatial layout planning auxiliary method based on big data mining according to claim 1, characterized in that, The set of static functional feature encoding vectors of the grid unit POI cluster and the set of population activity temporal pattern feature encoding vectors of the grid unit are combined to obtain a set of static function-dynamic activity joint encoding feature vectors of the grid unit. This includes concatenating the static functional feature encoding vectors of the grid unit POI cluster and the population activity temporal pattern feature encoding vectors of the grid unit to obtain the static function-dynamic activity joint encoding feature vectors of the grid unit.

6. The urban spatial layout planning auxiliary method based on big data mining according to claim 1, characterized in that, Spatial dynamic-static aggregation analysis is performed on the set of joint coding feature vectors of static function and dynamic activity of the grid cells to obtain the spatial dynamic-static feature aggregation coding vector of the target urban area, including: Feature baseline learning is performed on the set of joint coding feature vectors of static function-dynamic activity of the grid cell to obtain the regression coding vector of the joint coding feature of static function-dynamic activity of the grid cell; Calculate the dynamic compensation information of the set of encoding vectors of the joint static function-dynamic activity of the grid cell relative to the baseline regression encoding vector of the joint static function-dynamic activity of the grid cell to obtain the set of dynamic compensation weight factors of the joint static function-dynamic activity of the grid cell; Based on the set of dynamic compensation weight factors of the joint coding features of static function and dynamic activity of the grid cell and the set of coding vectors of the joint coding features of static function and dynamic activity of the grid cell, dynamic compensation is performed on the baseline regression coding vector of the joint coding features of static function and dynamic activity of the grid cell to obtain the aggregated coding vector of the spatial dynamic and static features of the target urban area.

7. The urban spatial layout planning auxiliary method based on big data mining according to claim 6, characterized in that, Calculate the dynamic compensation information of the set of joint coding vectors of static function-dynamic activity of the grid cell relative to the baseline regression coding vector of the joint coding vector of static function-dynamic activity of the grid cell to obtain the set of dynamic compensation weight factors of the joint coding vector of static function-dynamic activity of the grid cell, including: Calculate the feature dynamic compensation factor of each grid cell static function-dynamic activity joint coding feature coding vector in the set of grid cell static function-dynamic activity joint coding feature coding vectors relative to the grid cell static function-dynamic activity joint coding feature baseline regression coding vector to obtain the set of grid cell static function-dynamic activity joint coding feature dynamic compensation factors; The set of dynamic compensation factors for the joint coding features of static function and dynamic activity of the grid cell is regularized based on the Softmax activation function to obtain the set of dynamic compensation weight factors for the joint coding features of static function and dynamic activity of the grid cell.

8. The urban spatial layout planning auxiliary method based on big data mining according to claim 7, characterized in that, Based on the aggregated encoding vector of the spatial dynamic and static features of the target urban area, the target urban area is classified into functional zones to obtain functional zone category labels, including: The aggregated encoding vector of the spatial dynamic and static features of the target urban area is passed through a classifier-based target urban area functional area analyzer to obtain the functional area classification result, which is used to represent the functional area category label.

9. A city spatial layout planning auxiliary system based on big data mining, characterized in that, include: The POI data acquisition module is used to acquire a collection of POI data within the target city area; The grid-based time-series population data acquisition module is used to acquire rasterized time-series population activity data of the target urban area, wherein the rasterized time-series population activity data includes the number of residents in each grid cell of the target urban area at different time slices; The population activity data time series analysis module is used to perform population activity data time series analysis on the rasterized time series population activity data to obtain a set of feature encoding vectors of population activity time series patterns in grid cells; The grid cell cluster partitioning module is used to partition the set of POI data into clusters according to grid cells to obtain a set of grid cell POI data clusters; The grid cell static functional feature extraction module is used to extract static functional features from each grid cell POI data cluster in the set of grid cell POI data clusters to obtain a set of static functional feature encoding vectors for grid cell POI clusters. The grid cell static function-dynamic activity joint coding module is used to combine the static function feature coding vectors of the grid cell POI cluster and the population activity time-series pattern feature coding vectors of the grid cell to obtain the set of grid cell static function-dynamic activity joint coding feature coding vectors. The grid cell spatial dynamic and static aggregation analysis module is used to perform spatial dynamic and static aggregation analysis on the set of joint coding feature coding vectors of static function and dynamic activity of the grid cell to obtain the spatial dynamic and static feature aggregation coding vector of the target urban area; The functional area classification module is used to aggregate the encoding vector based on the spatial dynamic and static features of the target city area, and classify the target city area into functional areas to obtain functional area category labels.

Images