A method for analyzing the correlation between urban space three-dimensional structure and population flow

Abstract

The application discloses a kind of correlation analysis method of urban space three-dimensional structure and population flow, it is related to the field of urban space and population flow correlation analysis, the method includes the following steps: based on urban construction dataset constructs urban construction index system, by the structural characteristics analysis to urban construction index system, obtain the structural characteristics of urban construction three-dimensional space;Mobile phone signaling data is obtained using base station and identifies the commuting behavior of urban population, constructs the network structural characteristics of urban population flow;Regression model is constructed and trained using LightGBM algorithm, and the interpretive analysis of regression model is carried out by SHAP algorithm.The application analyzes the space characteristics of town and the spatio-temporal correlation of population flow, explores the importance of each feature of urban space structure to population flow, and then effectively uses population flow trend to provide valuable reference for urban development planning direction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of correlation analysis of urban space and population flow, in particular to a correlation analysis method of urban space three-dimensional structure and population flow. BACKGROUND

[0002] At present, the development of urbanization in China is rapid, and population flow is an important carrier of information flow, logistics, capital flow, and technology flow between regions. The large flow of population between cities will greatly impact the industrial structure, transportation links, infrastructure, employment situation, and urban scale of the inflow and outflow areas, thereby triggering the reorganization of urban internal factors and causing changes in the urban system structure and social development pattern. On the one hand, human activities drive the generation and decline, expansion and contraction, prosperity and decline of cities, and dominate production and life, technological innovation, and knowledge iteration in cities. On the other hand, human activities occur within the spatial organizational framework of cities and constrain human activity behavior. Therefore, in-depth identification of population flow direction, scale, network structure, and other rules, combined with geographical environment, social economy, history, and other factors, can explore the human activities and urban system construction status of different regions, enrich the theoretical system and methods of population urban planning and population geography, and provide reference and ideas for planning future population flow patterns and promoting the healthy and stable development of inflow and outflow areas.

[0003] Current research on urban space and population flow mainly focuses on the following aspects: first, by analyzing the direction, scale, and network structure of population flow, and combining geographical environment, social economic background, and other factors, the characteristics of human activities in different regions and their impact on the urban system are explored; second, based on the characteristics of human flow or from the perspective of human flow, the network system formed by human activities in the city is used to reflect the spatial distribution of activity centers and the correlation between centers, in order to reveal the essential characteristics of urban spatial structure; third, some researches map the urban spatial structure and function by the behavior patterns of residents' travel, such as establishing a system dynamics model to measure the impact of urban spatial density, traffic conditions, and economic conditions on residents' travel costs; in addition, some scholars use big data technology, such as mobile positioning data and trajectory data, to characterize the spatiotemporal characteristics of individual travel and explore the interaction between them and the urban grid and functional structure, further deepening the understanding of the characteristics of urban spatial structure.

[0004] Although current research has comprehensively covered multiple dimensions of urban spatial structure and emphasized the impact of population distribution, density, and human flow activities on urban structure, there are still deficiencies in in-depth analysis of the specific correlation between urban spatial structure and population flow, which cannot fully reveal the characteristics of population distribution through urban spatial layout, nor effectively use population flow trends to make precise urban development planning adjustments.

[0005] However, the prior art has not yet proposed an effective solution to the above problems. SUMMARY

[0006] In view of the problems in the prior art, the present application proposes a method for analyzing the correlation between the three-dimensional structure of urban space and population flow to overcome the above technical problems existing in the prior art.

[0007] To this end, the present application adopts the following specific technical solutions:

[0008] A method for analyzing the correlation between the three-dimensional structure of urban space and population flow, the correlation analysis method comprising the following steps:

[0009] S1, constructing an urban construction index system based on an urban construction dataset, and obtaining the three-dimensional spatial structure characteristics of urban construction by analyzing the structure characteristics of the urban construction index system;

[0010] S2, obtaining mobile phone signaling data using a base station and identifying the commuting behavior of urban population, and analyzing and constructing the network structure characteristics of urban population flow in combination with the GIS spatial analysis method;

[0011] S3, constructing a dataset using the three-dimensional spatial structure characteristics of urban construction and the network structure characteristics of urban population flow, and constructing and training a regression model using the LightGBM algorithm, and performing interpretive analysis on the regression model using the SHAP algorithm to obtain the correlation analysis result of the three-dimensional spatial structure of the city and population flow.

[0012] Further, the urban construction index system is constructed based on the urban construction dataset, and the three-dimensional spatial structure characteristics of urban construction are obtained by analyzing the structure characteristics of the urban construction index system, including the following steps:

[0013] S11, collecting urban building parameters to construct an urban construction dataset, and analyzing the distribution index of urban buildings and the suitability index of urban construction by statistical analysis method;

[0014] S12, constructing the urban building structure characteristics by index calculation based on the distribution index of urban buildings;

[0015] S13, calculating the urban construction suitability characteristics by expert scoring method and entropy weight method based on the suitability index of urban construction;

[0016] S14, combining the urban building structure characteristics and the urban construction suitability characteristics to obtain the three-dimensional spatial structure characteristics of urban construction.

[0017] Further, the urban building structure characteristics include the weighted average height of buildings, the building volume rate, the building coverage rate, the building volume, the proportion of high-rise buildings, and the standard deviation of building height.

[0018] Further, the calculation formula of the building weighted average height is:

[0019]

[0020] In the formula, BH represents the building weighted average height, A i represents the area occupied by the base of the building i, H i represents the height information of the i-th building, and n represents the sum of the number of areas belonging to the building.

[0021] The calculation formula of the building plot ratio is:

[0022]

[0023] In the formula, BPR represents the building plot ratio, A i represents the area occupied by the base of the building i, L i represents the number of floors of the building i, and n represents the sum of the number of areas belonging to the building, P a is the area of the area plot.

[0024] The calculation formula of the building coverage rate is:

[0025]

[0026] In the formula, BC represents the building coverage rate, A i represents the area occupied by the base of the building i, and n represents the sum of the number of areas belonging to the building, P a is the area of the area plot.

[0027] The calculation formula of the building volume is:

[0028]

[0029] In the formula, BV represents the building volume, A i represents the area occupied by the base of the building i, H i represents the height data of the building i, and n represents the sum of the number of areas belonging to the building.

[0030] The calculation formula of the high-rise building proportion is:

[0031]

[0032] In the formula, HBR represents the high-rise building proportion, N g represents the number of high-rise buildings, and N represents the total number of buildings in the area.

[0033] The calculation formula of the building height standard deviation is:

[0034]

[0035] In the formula, BHSD represents the building height standard deviation, H i represents the height information of the i-th building, μ represents the average value of the building height, and N represents the total number of buildings in the region.

[0036] Further, based on the urban construction suitability index, the urban construction suitability characteristics are calculated by expert scoring method and entropy weight method, including the following steps:

[0037] S131, normalizing each urban construction suitability index based on a normalization method to obtain a normalized index;

[0038] S132, quantitatively processing the obtained normalized index to obtain a quantitative index;

[0039] S133, combining expert scoring method and entropy weight method to analyze the weight of each quantitative index, according to the weight analysis result, using weighted summation method to calculate the comprehensive score of each quantitative index, and taking the comprehensive score as the urban construction suitability characteristics.

[0040] Further, the base station is used to obtain mobile signaling data and identify the commuting behavior of urban population, and the network structure characteristics of urban population flow are analyzed and constructed by combining GIS spatial analysis method, including the following steps:

[0041] S21, using the base station to obtain the mobile signaling data of the urban population, and preprocessing the mobile signaling data;

[0042] S22, according to the preprocessed mobile signaling data, the commuting data of the residential place and the working place during the morning peak and the evening peak are analyzed to identify the commuting behavior of the urban population;

[0043] S23, based on the commuting behavior of the urban population, the commuting volume of the urban population in the preset time period is calculated;

[0044] S24, taking the building site in the city as the node, taking the commuting behavior of the urban population as the edge, and taking the commuting volume of the urban population in the preset time period as the weight attribute of the edge, constructing the population flow network diagram of the city;

[0045] S25, using Gephi layout algorithm to analyze the population flow network diagram of the city, to obtain the network structure characteristics of the urban population flow.

[0046] Further, the preprocessing includes invalid and redundant data elimination, ping-pong data processing and drift data processing.

[0047] Further, a data set is constructed by using the three-dimensional spatial structure characteristics of urban construction and the network structure characteristics of urban population flow, and a regression model is constructed and trained by using a LightGBM algorithm, and an explanatory analysis of the regression model is performed by using a SHAP algorithm, to obtain a correlation analysis result of the three-dimensional spatial structure of the city and the population flow, including the following steps.

[0048] S31, divide the data set into a training set and a test set, construct a regression model by using a LightGBM algorithm, and initialize model parameters;

[0049] S32, based on the initialized model parameters, train the regression model by using the training set, and adjust and optimize the model parameters by cross-validation;

[0050] S33, evaluate the effect of the regression model by using the test set, calculate performance evaluation index data of the model, and obtain an evaluation result of the regression model;

[0051] S34, analyze the feature importance and influence direction of the evaluated regression model by using a SHAP algorithm, and establish a summary graph of SHAP according to the analysis result;

[0052] S35, analyze the correlation between the three-dimensional spatial structure of the city and the population flow according to the summary graph of SHAP, and obtain a correlation analysis result.

[0053] Further, the performance evaluation index data of the model includes a decision coefficient, a mean absolute error, a mean absolute percentage error, and a root mean square error.

[0054] Further, the feature importance and influence direction analysis of the evaluated regression model by using a SHAP algorithm includes the following steps:

[0055] S341, calculate the Shapley value by using the evaluated regression model based on the principle of cooperative game theory;

[0056] S342, explain and visually analyze the influence degree and feature action direction of the feature variable based on the Shapley value, and draw a visual chart by using a SHAP model, to obtain a summary graph of SHAP.

[0057] The beneficial effects of the present application are:

[0058] 1. The present invention ensures an accurate grasp of the essential characteristics of urban spatial structure by analyzing the three-dimensional characteristics of urban spatial structure. At the same time, by analyzing the network characteristics of population flow, it grasps the distribution of population flow and the characteristics of human activities in cities and towns, and further provides a basis for in-depth analysis of the relationship between urban spatial structure and population flow. Then, by analyzing the spatiotemporal correlation between urban spatial characteristics and population flow, it explores the importance of various characteristics of urban spatial structure to population flow, and then effectively utilizes population flow trends to provide valuable reference for the direction of urban development planning.

[0059] 2. Based on the town building height dataset and various town element indicators, the present invention obtains the suitability of town construction and six indicators to analyze the three-dimensional characteristics of the urban spatial structure, so as to describe the architectural form, spatial distribution and town development level of the town, and ensure the accurate grasp of the essential characteristics of the urban spatial structure.

[0060] 3. This invention identifies users' commuting behavior based on mobile phone signaling data, and then combines GIS spatial analysis methods to analyze and compare the network characteristics of population mobility in various towns, thereby grasping the distribution of population mobility in towns and the characteristics of human activities, thereby understanding the economic development trends, employment opportunities, urban planning, transportation facilities, economic activities between towns, and the intensity of social exchanges in these regions, and further providing a basis for in-depth analysis of the relationship between urban spatial structure and population mobility.

[0061] 4. Based on the LightGBM algorithm, the present invention takes the comprehensive suitability of urban construction and six indicators as input, and population mobility as output to construct a model. According to the model research results, the SHAP algorithm is used to analyze the influence of various input characteristics on population mobility, so as to explore the importance of various characteristics of urban spatial structure to population mobility, accurately grasp the driving factors of population mobility, further analyze the correlation between urban spatial characteristics and population mobility, and effectively use the correlation results to provide valuable reference for the direction of urban development planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0063] Figure 1 This is a flow chart of a method for analyzing the correlation between the three-dimensional structure of urban space and population mobility according to an embodiment of the present invention;

[0064] Figure 2is a schematic diagram of the calculation of the weighted average height (BH) of buildings in a grid in a correlation analysis method of urban space three-dimensional structure and population flow according to an embodiment of the application;

[0065] Figure 3 is a schematic diagram of the calculation of the building plot ratio (BPR) in a grid in a correlation analysis method of urban space three-dimensional structure and population flow according to an embodiment of the application;

[0066] Figure 4 is a population flow network diagram in a correlation analysis method of urban space three-dimensional structure and population flow according to an embodiment of the application;

[0067] Figure 5 is a LightGBM model SHAP scatter summary diagram in a correlation analysis method of urban space three-dimensional structure and population flow according to an embodiment of the application;

[0068] Figure 6 is a LightGBM model SHAP bar summary diagram in a correlation analysis method of urban space three-dimensional structure and population flow according to an embodiment of the application. DETAILED DESCRIPTION

[0069] To further illustrate the embodiments, the present application provides drawings, which are part of the disclosure of the present application, mainly to illustrate the embodiments, and can explain the operating principle of the embodiments in conjunction with the related description of the specification. With reference to these contents, those skilled in the art should understand other possible implementations and advantages of the present application.

[0070] According to an embodiment of the present application, a correlation analysis method of urban space three-dimensional structure and population flow is provided.

[0071] The present application will be further described in conjunction with the drawings and specific embodiments, as shown in Figure 1 According to an embodiment of the present application, the correlation analysis method of urban space three-dimensional structure and population flow includes the following steps:

[0072] S1, based on the urban construction dataset, an urban construction index system is constructed, and through structural feature analysis of the urban construction index system, urban construction three-dimensional space structure features are obtained;

[0073] Specifically, the urban construction dataset includes spatial survey data, and remote sensing big data.

[0074] Spatial survey data: land survey (referred to as "three surveys") data, according to the division of 12 first-level categories, land can be classified into six types, namely grassland, farmland, forest land, water area, construction land and unused land.

[0075] Big data of earth observation: Building height data, provided by the National Earth System Science Data Center. This dataset contains all buildings in China with a height of more than 10 meters. Each building in the data contains its geographic coordinates, height information, and building outline attributes. Building height data can be used to construct urban three-dimensional feature indexes to reflect urban spatial structure characteristics.

[0076] For example, according to the "2022 Fire Prevention Code" and "General Rules for Civil Building Design", the buildings in a certain province in southern China are divided into four categories: low-rise buildings (h <= 10m), medium-rise buildings (10m < h <= 24m), medium-high-rise buildings (24m < h <= 50m), and high-rise buildings (h > 50m). Following the above rules, based on the building height dataset, the statistical analysis method of ArcGIS is used to study the distribution of the number of buildings in different height segments. The analysis results are shown in Table 1. It can be seen that the number of medium-rise buildings is the most, accounting for 78.65%; followed by medium-high-rise buildings, accounting for 18.15%; low-rise buildings and high-rise buildings have similar proportions, accounting for 1.68% and 1.52%, respectively.

[0077] Table 1 Number and proportion of buildings in different height segments in a certain province in southern China

[0078]

[0079]

[0080] Among them, six indexes are selected, including building weighted average height (Buliding Height), building plot ratio (Building PlotRatio), building coverage (Building Coverage), building volume (Building Volume), high-rise building ratio (High-rise Building Ratio), and building height standard deviation (Building Height StandardDeviation), to describe the three-dimensional structure characteristics of urban space, to describe the building form, spatial distribution and urban development level of the city. At the same time, in order to ensure the homogeneity of the evaluation unit, 1000m grid is selected as the basic evaluation unit, that is, the study area of a certain province in southern China is divided into 148551 1000*1000m grids.

[0081] Specifically, based on the urban construction dataset, an urban construction index system is constructed, and through structural characteristic analysis of the urban construction index system, the three-dimensional spatial structure characteristics of urban construction include the following steps:

[0082] S11, collect the parameters of the urban buildings to construct an urban construction dataset, and analyze the distribution indexes of the urban buildings and the suitability indexes of the urban construction by a statistical analysis method;

[0083] In addition, the suitability of the urban construction is one of the representations of the structural features of the urban construction. The evaluation of the suitability of the urban construction is an integration of the consideration of the urban land resources, the water resources, the environment, the disasters, the location and other factors. The specific conditions of each land use and functional area in the city are analyzed in detail, and the development status of the city is understood. The positioning, nature and development goal of the city are further defined.

[0084] S12, based on the distribution indexes of the urban buildings, the structural features of the urban buildings are constructed by index calculation;

[0085] Specifically, the structural features of the urban buildings include the weighted average height of the buildings, the building volume rate, the building coverage rate, the building volume, the proportion of high-rise buildings and the standard deviation of the building height.

[0086] The weighted average height of the buildings (BH) is an index reflecting the average height level of the urban buildings, and is used to represent the height information of the buildings and the vertical development degree of the city. When the index is analyzed in the grid unit, the schematic diagram of the calculation in each grid is as shown in FIG. 1. Figure 2

[0087] The building volume (BV) is an index representing the three-dimensional space occupied by the buildings, and reflecting the building scale and the building development level of the city.

[0088] The building coverage rate (BC) is an index representing the horizontal space occupied by the buildings, and reflecting the building density and the land use efficiency of the city.

[0089] The building volume rate (BPR) is an index reflecting the ratio of the volume of the buildings to the land area of the city, and can represent the intensive degree of the horizontal distribution of the buildings and the development intensity in the vertical direction. When the index is analyzed in the grid unit, the schematic diagram of the calculation is as shown in FIG. 3. Figure 3

[0090] The proportion of high-rise buildings (HBR) is an index reflecting the proportion of the high-rise buildings in the city, and can represent the building structure of the city. The index is calculated in each county of the city.

[0091] The standard deviation of the building height (BHSD) is an index reflecting the dispersion degree of the building height in the city, and is used to represent the distribution change of the building height and the building form in the city. The index is calculated in each county of the city.

[0092] Specifically, the calculation formula of the weighted average height of the buildings is as follows:

[0093] ​​

[0094] wherein BH represents the weighted average height of buildings, A i represents the area of the base of building i, H i represents the height information of the i-th building, and n represents the sum of the number of areas belonging to the building;

[0095] The calculation formula of the building volume rate is:

[0096]

[0097] wherein BPR represents the building volume rate, A i represents the area of the base of building i, L i represents the number of floors of building i, and n represents the sum of the number of areas belonging to the building, P a is the area of the area plot;

[0098] The calculation formula of the building coverage rate is:

[0099]

[0100] wherein BC represents the building coverage rate, A i represents the area of the base of building i, n represents the sum of the number of areas belonging to the building, and P a is the area of the area plot;

[0101] The calculation formula of the building volume is:

[0102]

[0103] wherein BV represents the building volume, A i represents the area of the base of building i, H i represents the height data of building i, and n represents the sum of the number of areas belonging to the building;

[0104] The calculation formula of the high-rise building proportion is:

[0105]

[0106] wherein HBR represents the high-rise building proportion, N g represents the number of high-rise buildings (building height greater than 24m), and N represents the total number of buildings in the area;

[0107] The calculation formula of the building height standard deviation is:

[0108]

[0109] wherein BHSD represents the building height standard deviation, Hi represents the height information of the i-th building, μ represents the average value of the building height, and N represents the total number of buildings in the region.

[0110] S13, based on the urban construction suitability index, the urban construction suitability characteristics are calculated by expert scoring method and entropy weight method;

[0111] Specifically, based on the urban construction suitability index, the urban construction suitability characteristics are calculated by expert scoring method and entropy weight method, including the following steps:

[0112] S131, based on the normalization method, the normalized index is obtained by normalizing each urban construction suitability index;

[0113] S132, the obtained normalized index is quantitatively processed to obtain a quantitative index;

[0114] S133, combining expert scoring method and entropy weight method, weight analysis is performed on each quantitative index, according to the weight analysis result, the comprehensive score of each quantitative index is calculated by weighted summation, and the comprehensive score is taken as the urban construction suitability characteristic.

[0115] S14, combining the urban building structure characteristics and the urban construction suitability characteristics, the three-dimensional spatial structure characteristics of urban construction are obtained.

[0116] S2, using base stations to obtain mobile phone signaling data and identify the commuting behavior of urban population, combining with GIS spatial analysis method to analyze and construct the network structure characteristics of urban population flow;

[0117] It should be noted that the mobile phone signaling data is the data automatically captured and recorded by the system when the user triggers the relevant communication event. As long as the mobile phone is turned on, it will continuously interact with the communication base station, and when the mobile phone user moves to different locations at different time points, the communication base station they connect will also change accordingly, because the current base station position information is recorded every time interaction, then the mobile phone signaling data is generated. Each signaling data contains multiple key fields, covering base station identification code, signaling triggering location area code, signaling triggering time and other core information.

[0118] Specifically, using base stations to obtain mobile phone signaling data and identify the commuting behavior of urban population, combining with GIS spatial analysis method to analyze and construct the network structure characteristics of urban population flow includes the following steps:

[0119] S21, using base stations to obtain mobile phone signaling data of urban population, and preprocessing the mobile phone signaling data;

[0120] Specifically, the preprocessing includes invalid and redundant data elimination, ping-pong data processing and drift data processing.

[0121] Specifically, invalid and redundant data elimination, such data missing is due to mobile devices and base stations communication network conditions may be poor, delay, data transmission interruption and other problems, which will lead to the loss of base station information recorded signaling data, including base station encoding, location area code (LAC) and base station latitude and longitude, etc. Redundant data refers to the same base station, the user triggered multiple repeated mobile phone signal data, but in fact the mobile phone user did not make the corresponding position change. These invalid and redundant data will affect the accuracy and efficiency of subsequent data analysis, so in the data preprocessing stage, it is necessary to filter and clean them, and keep the effective data samples. For data missing, the corresponding data screening is used to delete it. For redundant data generated continuously with complete base station information, the method of adding is used to process it, which is to add the duration of the mobile terminal and the base station communication, and delete the data that has been merged.

[0122] Ping-pong data processing is that mobile devices may encounter weak or unstable signal conditions during communication, resulting in a decline in the quality of connection with the current base station. In order to maintain the stability of communication, the device will try to switch to other base stations. Mobile devices may cross the coverage boundary of different base stations during movement. When the device moves to a new base station coverage area, it will try to establish a connection with the new base station, which will cause frequent switching between base stations. When mobile devices repeatedly locate between different base stations, ping-pong data will be generated. For the phenomenon of ping-pong data switching, the common cell encoding sequence has two types of A-B-C-A and A-B-A. Based on the method in the existing research, first judge whether the cell encoding sequence belongs to the ping-pong data switching type, and then use the merging processing method for the sequence that meets the condition. In the processing, the start time of the first signaling of the ping-pong data and the end time of the last signaling are updated as the corresponding fields of the new signaling, and the total length of the repeated data is accumulated, which is recorded as the actual residence time of the new signaling data in the base station.

[0123] Drift data processing refers to the phenomenon that the position of a mobile device remains unchanged or changes slowly within a period of time, but the data deviates or drifts due to error accumulation or environmental changes. For example, during the operation of a mobile device for a long period of time, the sensor may change due to external environmental factors such as temperature fluctuations, humidity changes, or external pressure, causing the error of data output to gradually accumulate, thereby causing the drift of position data. At the same time, positioning system errors, algorithm errors, signal interference, etc. may introduce drift in the data processing process, affecting the accuracy of the position data. Based on existing research, first, determine whether the spatial distance between the base stations of the signaling data record exceeds a certain threshold value, and for identifying drift data, the primary task is to assess whether the switching speed exceeds the normal range; then integrate adjacent drift records, update the start time of the combined data to the start time of the previous signaling, and the end time to the end time of the next signaling, and add the length of the integrated drift record to the base station stay time of the new signaling data.

[0124] S22, according to the preprocessed mobile signaling data, analyze the commuting data between the residence and the workplace during the morning peak and the evening peak to identify the commuting behavior of the urban population;

[0125] S23, based on the commuting behavior of the urban population, calculate the commuting volume of the urban population in a preset time period;

[0126] For example, analyzing personnel flow first needs to determine the residence and workplace of the user object. In this paper, each natural month is taken as the statistical unit, and there are many stay points in the monthly trajectory of the user. The residence and workplace of the user need to be found among these points. The general working time period is within the range of 7 o'clock in the morning to 19 o'clock in the evening. Since the activities after work vary from person to person, some users may not directly return to their residence after work. The rest time is defined as 23 o'clock in the evening to 7 o'clock the next day. In this invention, the place with the longest stay (and the stay number exceeds 15 days) from 23 o'clock in the evening to 7 o'clock the next day in a month is defined as the residence of the signaling user; the place with the longest stay (and the stay number exceeds 10 days) from 7 o'clock in the morning to 19 o'clock in the evening in a month is defined as the workplace of the signaling user.

[0127] First, the relevant concepts are defined, and the population flow research of the present application mainly focuses on the commuting behavior between counties. The definition of commuting behavior is that the early 6-9 am and late 17-21 pm of the working day are regarded as the commuting round-trip period, and the signaling user's moving behavior between the work place and the residence place or the residence place and the work place within this time period is defined as a commuting behavior. The number of commuters is calculated according to the identity card number of the signaling user who has a commuting behavior. The monthly average number of commuters is obtained by accumulating the daily commuting number. The commuting volume is obtained by accumulating the single commuting behavior, and according to the round-trip place, it is divided into two commuting behaviors, i.e. A place to B place and B place to A place. The monthly average commuting volume is obtained by accumulating the daily commuting volume.

[0128] The specific calculation logic is that first, the signaling data is read from the base station, and the personnel whose work place and residence place do not coincide are screened out according to the job and residence point identification. The signaling data of these users are analyzed, and the commuting behavior is screened out, and the time range is the commuting round-trip period. At the same time, according to the base station longitude and latitude information in the signaling data, the belonging county is located, and the administrative district information is added. According to the record time sequence of the daily base station, the earliest time of the early and late peak appearing at the residence place and the earliest time of the early and late peak appearing at the work place are obtained. The data of the residence place and the work place during the early and late peak are integrated, and the early peak appearing at the residence place is less than the early peak appearing at the work place, and the late peak appearing at the work place is less than the late peak appearing at the residence place, which can be counted as a commuting behavior. Finally, according to this standard, the commuting volume is obtained according to the grouping of the commuting work place and the residence place, and the daily commuting volume is accumulated to obtain the monthly commuting volume; according to the user grouping, the single commuting frequency can be obtained; according to the user identity card number, the daily commuting number is obtained, and the monthly commuting number is accumulated. At the same time, according to the user's identity card information, the user's age information can be obtained, the user's age portrait is read, and the number of people in each age group is obtained by grouping according to age.

[0129] S24, taking the building site in the city as a node, taking the commuting behavior of the population in the city as an edge, and taking the commuting volume of the population in the city in a preset time period as an edge weight attribute, a population flow network diagram of the city is constructed;

[0130] S25, using Gephi layout algorithm to analyze the population flow network diagram of the city, and obtaining the network structure characteristics of the population flow of the city.

[0131] Specifically, the present application uses complex network analysis, combines with GIS spatial analysis method, studies the population flow relationship and node scale between cities, analyzes and compares the network characteristics of the population flow of each city, from the perspective of connection, and explores the ability of population flow to attract and radiate by analyzing the characteristics of population flow in different regions.

[0132] The population flow network diagram is drawn by using Gephi software (as shown inFigure 4 Gephi is an open-source network visualization tool that can be used to analyze and visualize complex network data, helping to deeply study network structure, relationships, and patterns, and intuitively display network data through graphical means. First, create a node table and edge table for input into Gephi. The node table contains the districts and counties of a certain province in southern China, and supplements the latitude and longitude information of each town. When using Gephi layout algorithms to optimize network graph layout, geographic coordinate network visualization can be achieved, and the nodes are classified by city level for subsequent visualization. The edge table inputs the contact flow information between regions, and the population commuting volume between counties in a certain province in southern China as the weight attribute of the edge to represent the directionality and strength of the network. Gephi can perform various network analysis tasks, such as calculating the degree, betweenness, clustering coefficient, and other indicators of nodes, detecting community structure, and identifying important nodes in the network. Here, the weight of the edge is used to define the style of the edge, and the greater the flow volume, the wider the edge. In order to improve the visualization effect, the size of the node and the color of the node and edge are not set too much to avoid too much clutter in the graph. Here, only the city-level administrative areas of a certain province in southern China are classified and visualized, and different colors represent different cities. From the figure, it can be seen that the northeast and central regions have frequent population commuting and strong inter-town connections, indicating that these regions have good economic development, many job opportunities, well-planned and well-constructed urban planning and transportation facilities, and intensive economic activity and social exchange between towns, attracting a large number of population flow; other regions have weak connections and do not form a core with strong population attraction and radiation, meaning that the economic development level of these regions is relatively low, and the central area of the town does not have enough influence to effectively mobilize economic exchange and attract talent.

[0133] S3, using the three-dimensional spatial structure characteristics of urban construction and the network structure characteristics of urban population flow to construct a data set, and using the LightGBM algorithm to construct and train a regression model, and using the SHAP algorithm to perform explanatory analysis on the regression model to obtain the correlation analysis results of the three-dimensional spatial structure of the town and population flow.

[0134] It should be noted that in order to capture the complex nonlinear interactions between variables, the present application uses the LightGBM algorithm and Shapley value interpretation model to further analyze the correlation between urban spatial characteristics and population flow spatio-temporal behavior, to explain and visualize the effect of urban spatial factors on population flow, and to explore the correlation mechanism.

[0135] LightGBM (GBDT, Gradient Boosting Decision Tree) refers to a gradient boosting decision tree algorithm, and its basic principle is to complete the prediction task by constructing an ensemble of multiple decision tree models. GBDT is an ensemble learning method that gradually reduces training errors by gradually constructing decision tree models. In each iteration, GBDT learns a new decision tree model that tries to capture the residual error of the samples that the previous model predicted incorrectly, thereby gradually reducing the residual error and ultimately achieving the purpose of fitting the training data. The core principle of LightGBM is to optimize the speed and efficiency of the gradient boosting algorithm through the Leaf-wise growth strategy and the histogram algorithm. The Leaf-wise growth strategy means that in the splitting process, the leaf node with the highest benefit to the current node is always split. This strategy can construct more leaf nodes at the same depth, improving the accuracy of the model. The histogram algorithm first discretizes continuous feature values into different bins, and then uses a histogram to find the best split point at each split. This algorithm can reduce computational complexity and memory usage, improving the training speed of the model.

[0136] Like machine learning models, the decision-making process of such complex and opaque model methods is becoming more and more complex, and the explainability is becoming lower and lower. According to the proposed SHAP as an explanatory index of the value of the features in the model, it is an additive feature attribution method with local and global explanatory capabilities. This concept first appeared in game theory to measure the value of each party in the game process. In a machine learning model, each feature variable of a sample can be regarded as a participant in the game process, and the Shapley value of the feature reflects the contribution of the feature to the model result. The calculation principle of Shapley value is based on the concept of cooperative game theory, which measures the contribution of each feature to the model prediction result.

[0137] Specifically, a data set is constructed using the three-dimensional spatial structure characteristics of urban construction and the network structure characteristics of urban population flow, and a regression model is constructed and trained using the LightGBM algorithm. The correlation analysis result of the three-dimensional spatial structure of the city and the population flow is obtained by using the SHAP algorithm to explain the regression model, including the following steps:

[0138] S31, divide the data set into training set and test set, construct regression model using LightGBM algorithm and initialize model parameters;

[0139] S32, based on the initialized model parameters, train the regression model using the training set, and adjust and optimize the model parameters through cross-validation;

[0140] S33, evaluate the effect of the regression model by using the test set, and calculate the performance evaluation index data of the model to obtain the evaluation result of the regression model;

[0141] It should be noted that based on the LightGBM algorithm, the present application takes each variable factor of the urban spatial structure (the three-dimensional spatial structure characteristics of urban construction) as input, takes the population flow (the network structure characteristics of urban population flow) as output, and constructs a data set. 85% of which is randomly selected as a training set, and 15% is selected as a test set to construct a model. The model parameters are set, including the number of base learners, the maximum number of leaf nodes of each base learner, the learning rate, the depth of the tree, the number of iterations and other parameters. After cross-validation and continuous adjustment and optimization of the parameters, the accuracy of the model is further improved. Finally, the parameters are set to 70, 50, 0.01, 6 and 10000 respectively. After setting the parameters, the model is trained, and the effect of the model is evaluated in combination with the test set. The present application selects the coefficient of determination (R 2 ), the mean absolute error (MAE), the mean absolute percentage error (MAPE) and the root mean square error (RMSE) as the indicators for verifying the accuracy of the model.

[0142] In the R language, the data is read, the parameters are set, the regression model of the LightGBM algorithm is constructed and trained, a fitted scatter plot is obtained, and R 2 is 0.97, RMSE is 292.04, MAE is 85.55, and MAPE is 0.13. The coefficient of determination of the model is close to 1, which indicates that the model can well explain the changes of the target variable. At the same time, the other three error values are relatively small, indicating that the prediction error of the model is small, and the fitting effect is good. The fitting result shows that the influence factor system selected by the present application is reasonable and has a strong explanatory power for the dependent variable.

[0143] S34, analyze the feature importance and influence direction of the evaluated regression model by using the SHAP algorithm, and establish a summary graph of SHAP according to the analysis result;

[0144] Specifically, the performance evaluation index data of the model includes the coefficient of determination, the mean absolute error, the mean absolute percentage error and the root mean square error.

[0145] Specifically, the feature importance and influence direction of the evaluated regression model are analyzed by using the SHAP algorithm, and the summary graph of SHAP is established according to the analysis result, including the following steps:

[0146] S341, calculate the Shapley value based on the evaluated regression model according to the principle of cooperative game theory;

[0147] S342, based on the influence degree of Shapley value on the feature variable and the feature action direction, explanation and visual analysis are carried out, and a visual chart is drawn by using SHAP model to obtain a summary chart of SHAP.

[0148] S35, according to the summary chart of SHAP, the correlation between the three-dimensional spatial structure of the town and the population flow is analyzed to obtain a correlation analysis result.

[0149] It should be noted that the summary chart of SHAP has two forms: one is a scatter plot showing all sample individual features, and the other is a bar chart based on the absolute mean of the Shapley value of each feature to reflect the importance of the feature. The importance of the feature represents the degree of reference dependence of the feature in the model prediction. Generating the SHAP summary chart can not only determine the importance ranking of all features, but also analyze the positive or negative influence of the Shapley value of each feature on the prediction result, so as to better understand the contribution of different features in the model prediction.

[0150] As shown in Figure 5 , which is a SHAP scatter plot, the vertical axis represents each influencing factor, i.e. each independent variable feature, which is arranged from top to bottom according to the importance of the feature in the model. It can be seen that the top three are building coverage (BC), high-rise building proportion (HBR) and urban construction suitability (ConsS). The horizontal axis is the Shapley value of each feature, i.e. the weight of each variable on the dependent variable, which is greater than 0, indicating a positive promoting effect on population flow, less than 0, indicating an inverse inhibitory effect, and the greater the absolute value, the greater the degree of influence. Each scatter plot corresponding to each feature in each row of the chart represents the influence of the feature, where each point is a sample, and the color of the point represents the high or low value of the feature of each sample. The closer the color is to light color, the higher the feature value, and the closer the color is to dark color, the lower the feature value.

[0151] In Figure 5As can be seen from the data, the top three features have similar effects on population mobility. Sample points with SHAP values ​​less than 0 are primarily darker in color, while those with values ​​greater than 0 are generally lighter in color. This means that higher values ​​for these features promote population mobility, and vice versa. However, it is clear that the sample distribution for building coverage (BC) is more dispersed, while the points for high-rise building ratio (HBR) and urban construction suitability (ConsS) are more densely populated and tend to be concentrated around a SHAP value of zero. This suggests that the influence of higher values ​​for these two features on population mobility is less pronounced than that for building coverage (BC). The influence of the remaining four characteristics is more complicated. Sample points with high eigenvalues ​​of the weighted average height of buildings (BH) are distributed around the SHAP zero value, and the distribution is roughly uniform. At the same time, some sample points with low eigenvalues ​​are distributed in the negative SHAP value area. This shows that a low weighted average height of buildings will have a certain inhibitory effect on population mobility, but the impact of a high weighted average height of buildings on population mobility is very complex and there is no obvious guiding pattern; most of the points with low eigenvalues ​​of building volume (BH) are distributed in the positive SHAP value area and near the zero value, while some low-value points are scattered in the negative SHAP value area, indicating that the impact mechanism of this feature on population mobility is complex; a large number of light-colored (high-value) sample points of the standard deviation of building height (BHSD) are concentrated around the SHAP zero value, indicating that this feature has no significant effect on the dependent variable. Similarly, the building volume ratio (BPR) shows that all sample points are concentrated at the SHAP zero value, indicating that the volume ratio has almost no effect on population mobility in this model.

[0152] like Figure 6 As shown in the figure, this is another summary chart of the SHAP bar chart. The vertical axis represents the various features input into the model, and the horizontal axis represents the mean of the absolute value of the SHAP of each feature, which quantifies the influence of the feature on the overall prediction of the model. Like the scatter plot, the features are ranked from top to bottom according to their importance in the model. However, unlike the scatter plot, the bar chart can reflect the relative importance of the features rather than just the ranking. Figure 6 Although the high-rise building ratio (HBR) and urban construction suitability (ConsS) rank second and third, their importance is very close, and both are only about half of the first-place building coverage ratio (BC); in addition, it is observed that the building volume ratio (BPR) not only ranks last, but its importance is almost zero.

[0153] In summary, based on the analysis results of LightGBM and SHAP, it can be known that the importance of the building coverage (BC), the high-rise building ratio (HBR) and the urban construction suitability (ConsS) is the highest among the features, and the higher the feature values are, the more the population flow is promoted. Among them, the importance of the building coverage is twice that of the second feature, and the influence on the population flow is the largest. Therefore, the spatial structure system is also a driving factor of the population flow characteristics.

[0154] In summary, by means of the above technical solutions of the present application, the three-dimensional features of the urban spatial structure are analyzed to ensure accurate understanding of the essential features of the urban spatial structure, and the network features of the population flow are analyzed to understand the population flow distribution and human activity characteristics of the city, which further provides a basis for in-depth analysis of the relationship between the urban spatial structure and the population flow. Then, the spatial correlation analysis between the urban spatial features and the population flow is performed to explore the importance of the features of the urban spatial structure to the population flow, and the population flow trend is effectively utilized to provide valuable reference for the development planning direction of the city. The urban construction suitability and six indexes are obtained based on the urban building height data set and the urban element indexes to analyze the three-dimensional features of the urban spatial structure to describe the building form, spatial distribution and urban development level of the city, so as to ensure accurate understanding of the essential features of the urban spatial structure. The user's commuting behavior is identified based on the mobile phone signaling data, and the network features of the population flow of each city are analyzed and compared by combining the GIS spatial analysis method to understand the population flow distribution and human activity characteristics of the city, so as to understand the economic development trend, employment opportunities, urban planning, transportation facilities, economic activities and social communication intensity between cities, which further provides a basis for in-depth analysis of the relationship between the urban spatial structure and the population flow. The comprehensive suitability of the urban construction and the six indexes are taken as inputs, and the population flow is taken as output based on the LightGBM algorithm to build a model, and the SHAP algorithm is used to analyze the influence of each feature on the population flow according to the model research results, so as to explore the importance of the features of the urban spatial structure to the population flow, accurately grasp the driving factors of the population flow, further analyze the correlation between the urban spatial features and the population flow, and effectively utilize the correlation results to provide valuable reference for the development planning direction of the city.

[0155] The above only describes the preferred embodiments of the present application and should not be used to limit the present application, and any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the protection scope of the present application.

Claims

1. A method for analyzing the correlation between urban spatial three-dimensional structure and population mobility, characterized in that: The correlation analysis method comprises the following steps: S1. Construct an urban construction index system based on the urban construction dataset. By analyzing the structural characteristics of the urban construction index system, the three-dimensional spatial structural characteristics of urban construction are obtained. S2. Use base stations to obtain mobile phone signaling data and identify the commuting behavior of urban residents. Combined with GIS spatial analysis methods, analyze and construct the network structure characteristics of urban population mobility. The S2 includes: S23. Calculate the commuting volume of the urban population in a preset time period based on the commuting behavior of the urban population; S24. Construct a population flow network diagram for a town using building locations in the town as nodes, commuting behavior of the town population as edges, and commuting volume of the town population in a preset time period as edge weight attributes; S25. Use Gephi layout algorithm to analyze the urban population flow network diagram and obtain the network structure characteristics of urban population flow; S3. Utilizing the three-dimensional spatial structure characteristics of urban construction and the network structure characteristics of urban population mobility, a dataset was constructed. A regression model was constructed and trained using the LightGBM algorithm. The regression model was then interpreted using the SHAP algorithm to obtain correlation analysis results between the three-dimensional spatial structure of urban areas and population mobility. Said S1 comprises the following steps: S11. Collect urban building parameters to construct an urban construction dataset, and analyze the distribution indicators of urban buildings and urban construction suitability indicators through statistical analysis methods; S12. Based on the distribution index of urban buildings, construct the structural characteristics of urban buildings through index calculation; S13. Based on the urban construction suitability index, calculate the urban construction suitability characteristics through expert scoring method and entropy weight method; S14. combining the town building structure characteristics and the town construction suitability characteristics to obtain the three-dimensional spatial structure characteristics of the town construction; The urban building structure characteristics include the weighted average height of buildings, building volume ratio, building coverage ratio, building volume, proportion of high-rise buildings and standard deviation of building heights.

2. The method for analyzing the correlation between the three-dimensional structure of urban space and population mobility according to claim 1, characterized in that: The calculation formula for the weighted average height of the buildings is: Where BH represents the weighted average height of buildings, A i represents the area occupied by the base of building i, H i represents the height information of the i-th building, and n represents the sum of the number of areas belonging to the building; The calculation formula for the building volume ratio is: Where BPR represents the building volume ratio, A i represents the area occupied by the base of building i, L i represents the number of floors of building i, n represents the sum of the number of areas belonging to the building, P a is the area of ​​the regional plot; The calculation formula for the building coverage rate is: Where BC represents the building coverage rate, A i represents the area occupied by the base of building i, n represents the sum of the number of areas belonging to the building, P a Indicates the area of ​​the regional plot; The calculation formula for the building volume is: In the formula, BV represents the building volume, Ai represents the base area of ​​building i, and H i represents the height data of building i, and n represents the sum of the number of areas belonging to the building; The calculation formula for the proportion of high-rise buildings is: Where HBR represents the proportion of high-rise buildings, N g represents the number of high-rise buildings, and N represents the total number of buildings in the area; The calculation formula for the standard deviation of building height is: Where BHSD represents the standard deviation of building height, H i represents the height information of the i-th building, μ represents the average building height, and N represents the total number of buildings in the area.

3. The method for analyzing the correlation between the three-dimensional structure of urban space and population mobility according to claim 1, characterized in that: The method of calculating the suitability characteristics of urban construction based on the urban construction suitability index by using the expert scoring method and the entropy weight method includes the following steps: S131. normalizing each urban construction suitability index based on a normalization method to obtain a normalized index; S132, performing quantization processing on the obtained normalized index to obtain a quantitative index; S133. Combine the expert scoring method and entropy weight method to conduct weight analysis on each quantitative indicator. Based on the weight analysis results, use the weighted summation method to calculate the comprehensive score of each quantitative indicator, and use the comprehensive score as the suitability feature of urban construction.

4. The method for analyzing the correlation between urban spatial three-dimensional structure and population mobility according to claim 1, characterized in that: The method of using base stations to obtain mobile phone signaling data and identify the commuting behavior of urban residents, and combining GIS spatial analysis methods to analyze and construct network structure characteristics of urban population flow also includes the following steps: S21. Using the base station to obtain mobile phone signaling data of urban residents, and pre-processing the mobile phone signaling data; S22. Identify the commuting behavior of the urban population by analyzing the commuting data between residence and workplace during the morning and evening peak periods based on the pre-processed mobile phone signaling data.

5. The method for analyzing the correlation between the three-dimensional structure of urban space and population mobility according to claim 4, characterized in that: The preprocessing includes invalid and redundant data elimination, ping-pong data processing and drift data processing.

6. The method for analyzing the correlation between urban spatial three-dimensional structure and population mobility according to claim 1, characterized in that: The method of constructing a data set using the three-dimensional spatial structure characteristics of urban construction and the network structure characteristics of urban population flow, constructing and training a regression model using the LightGBM algorithm, and performing an interpretative analysis of the regression model using the SHAP algorithm to obtain the correlation analysis results between the three-dimensional spatial structure of urban areas and population flow includes the following steps: S31. Divide the dataset into training set and test set, build a regression model using the LightGBM algorithm and initialize the model parameters; S32. Based on the initialized model parameters, the regression model is trained using the training set, and the model parameters are adjusted and optimized through cross-validation; S33. Evaluate the effect of the regression model using the test set, and calculate the performance evaluation index data of the model to obtain the evaluation result of the regression model; S34. Use the SHAP algorithm to analyze the feature importance and influence direction of the evaluated regression model, and create a SHAP summary diagram based on the analysis results; S35. Based on the SHAP summary diagram, the correlation between the three-dimensional spatial structure of the town and population mobility was analyzed to obtain the correlation analysis results.

7. The method for analyzing the correlation between the three-dimensional structure of urban space and population mobility according to claim 6, characterized in that: The performance evaluation index data of the model include the coefficient of determination, mean absolute error, mean absolute percentage error and root mean square error.

8. The method for analyzing the correlation between the three-dimensional structure of urban space and population mobility according to claim 7, characterized in that: The SHAP algorithm is used to analyze the feature importance and influence direction of the evaluated regression model, and the SHAP summary diagram is established according to the analysis results, including the following steps: S341. Based on the principles of cooperative game theory, the Shapley value is calculated using the evaluated regression model. S342. Based on the Shapley value, the degree of influence of the characteristic variables and the direction of the characteristic action are interpreted and analyzed visually, and the SHAP model is used to draw a visual chart to obtain a SHAP summary chart.

Images