A method for estimating population spatiotemporal distribution at hourly scale

Abstract

In response to the limitations of existing technologies, the present invention proposes a method for estimating the spatiotemporal distribution of population on an hourly scale. By combining dynamic location service data with population spatial prediction data constructed based on multiple ring buffers, the present invention can effectively solve the problems of missing temporal features and low resolution in existing population distribution data, and achieve a detailed depiction of the spatiotemporal characteristics of population distribution.

Description

Technical Field

[0001] The present invention relates to the technical field of population data spatiotemporalization, and in particular to a method for estimating population spatiotemporal distribution on an hourly scale. Background Art

[0002] The detailed spatial and temporal distribution of populations is a crucial basis for urban development planning and ecological and environmental protection strategies. High-precision data on population temporal and spatial distribution can characterize the population's activity levels throughout the day and night, providing crucial data support for assessing population health, addressing climate change, and allocating resources. Therefore, research on this data is of profound and far-reaching significance.

[0003] As shown in the Chinese invention patent with an announcement date of January 8, 2021: A method for dynamic population prediction at a fine spatiotemporal scale based on location big data, it constructs a multi-level regional fine spatiotemporal scale dynamic population prediction model at the district, county, and grid scales based on social perception big data and multiple machine learning models. The current traditional census data is usually counted and summarized step by step based on administrative districts. However, demographic data based on administrative units cannot be overlapped with natural geographical units, does not support spatial operations and analysis, and is difficult to reflect the spatial distribution characteristics of the population with high precision; the statistical cycle is once every 10 years, which is not timely for rapidly developing cities and cannot provide good decision-making data support.

[0004] However, current population spatialization data derived from the fusion of static, multi-source geographic information data lacks temporal heterogeneity, making it difficult to deeply characterize the mobility characteristics of urban populations. Dynamic population-related data is also lacking to supplement the temporal characteristics of population activity. Furthermore, current population distribution research often constructs models based on population-related data, such as point of interest data. However, different types of point of interest data exhibit strong correlations, and frequently occurring but meaningless place names and addresses also hinder correlation analysis. Previous studies have largely overlooked this issue, and further research on population-related data requires precise screening and characterization. Summary of the Invention

[0005] In view of the limitations of the existing technology, the present invention proposes a method for estimating the spatiotemporal distribution of population on an hourly scale. The technical solution adopted by the present invention is:

[0006] A method for estimating the spatiotemporal distribution of population at an hourly scale comprises the following steps:

[0007] S1: Obtain census data, POI data, building area and height data, road network data, and Tencent location data for each administrative unit in the target area, perform coordinate conversion on the above data, and match and spatially connect them with the boundaries of each administrative unit in the target area;

[0008] S2, based on the processing result of step S1, using the population correlation of the POI data to find the population density peak point of each administrative unit;

[0009] S3, by calculating the probability density of the two-dimensional Gaussian distribution of the population density peak point of each administrative unit, the preliminary population spatial distribution based on the two-dimensional Gaussian model is realized;

[0010] S4, establishing a multi-ring buffer zone for each administrative unit in the target area with the population density peak point of each administrative unit as the center; obtaining the population number, building area data and height data, road network length data, and various POI data in each buffer zone based on the processing results of step S1;

[0011] S5, based on the preliminary population spatial distribution obtained in step S3 and using the processing result of step S4 as an input variable, a random forest model is used to obtain population prediction data at a preset spatial resolution;

[0012] S6, fusing the population prediction data obtained in step S5 with the Tencent location data to achieve urban population distribution at a preset spatial resolution and temporal resolution.

[0013] Compared with the existing technology, the present invention can effectively solve the problems of missing time features and low resolution in existing population distribution data by combining dynamic location service data with population spatial prediction data constructed based on multi-ring buffers, and achieve a detailed characterization of the spatiotemporal characteristics of population distribution.

[0014] As a preferred solution, step S2 includes the following process:

[0015] Calculate the TF-IDF value of the POI data in each administrative unit within the target area; perform Pearson correlation analysis on the calculated TF-IDF value and the census data to obtain correlation coefficients for various types of POIs in the POI data; use the POI type with the highest correlation coefficient as the POI type with the highest population correlation, and use the distribution centroid of this type of POI point as the peak population density point of the corresponding administrative unit.

[0016] Furthermore, the TF-IDF value is calculated according to the following formula:

[0017]

[0018]

[0019]

[0020] in, is the TF-IDF value of the i-th type of POI in the j-th administrative unit, n ij is the number of POIs of type i in the jth administrative unit, n k,j is the number of POIs of type k in the jth administrative unit, |D| is the total number of census units, |{j:t i ∈d j}| is the number of census units containing the i-th type of POI.

[0021] Furthermore, the Pearson correlation analysis was performed according to the following formula:

[0022]

[0023] Among them, TI ij is the TF-IDF value of the i-th type of POI in the j-th administrative unit, PD j is the population of the j-th administrative unit.

[0024] Furthermore, in step S2, the POI type with the highest correlation coefficient is used as the POI type with the highest population correlation, and the distribution centroid of the POI type is used as the population density peak point of the corresponding administrative unit. The distribution centroid of the POI type is obtained by the following formula:

[0025]

[0026]

[0027] Among them, x c ,y c are the longitude and latitude of the centroid of the administrative unit, N is the number of POIs of the type with the highest population correlation in the administrative unit, and x i ,y i are the longitude and latitude of the i-th POI point with the highest population correlation in the administrative unit.

[0028] As a preferred solution, in step S3, the probability density of the two-dimensional Gaussian distribution is calculated for the population density peak point of each administrative unit according to the following formula:

[0029]

[0030]

[0031]

[0032] Among them, f i (x,y) is the population distribution probability at position (x,y) in the ith census block, and x and y follow normal distributions respectively. and ρ is the correlation coefficient between and, μ x ,μ y is the longitude and latitude of the peak population density point of the census block, Dist(loc(x,y),loc(μ x ,μ y )) is the distance between the census block location (x, y) and the peak population density point.

[0033] As a preferred solution, in step S5, the population prediction data is corrected according to the following formula:

[0034]

[0035] Among them, Pop adjust Population forecast data after correction cal is the population prediction data output by the random forest model, N census is the census data in the corresponding administrative unit, N cal is the sum of the population prediction data output by the random forest model in the corresponding administrative unit.

[0036] As a preferred solution, in step S6, the population prediction data obtained in step S5 is fused with the Tencent location data in the following manner:

[0037] Using the weight coefficient method, find the Tencent location data closest to any grid in the target area, compare the positioning data of Tencent location data in any time period with the average positioning data of the preset time length, obtain the positioning data variation coefficient within the preset time length, multiply the population prediction data at the preset spatial resolution by the positioning data variation coefficient, and obtain the urban population distribution at the preset spatial resolution and time resolution.

[0038] Furthermore, the specific gravity coefficient method is implemented according to the following formula:

[0039]

[0040] Pop ik =W ij ·P k ;

[0041] Among them, TP ij is the number of positioning times of Tencent location data in the jth grid in the i-th period, W ij is the proportion of the positioning times in the jth grid in the i-th period to the total area, n is the number of regional grids, P k is the population prediction data in the kth grid obtained in step S5, Pop ikis the final population count in the kth grid in the i-th period.

[0042] The present invention also provides the following:

[0043] A computer storage medium stores a computer program, which, when executed by a processor, implements the steps of the aforementioned method for estimating the spatiotemporal distribution of population on an hourly scale.

[0044] A computer device includes a storage medium, a processor, and a computer program stored in the storage medium and executable by the processor. When the computer program is executed by the processor, the steps of the method for estimating the spatiotemporal distribution of population on an hourly scale in Example 1 are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A schematic diagram of the steps of the method for estimating the spatiotemporal distribution of population on an hourly scale provided in Example 1 of the present invention;

[0046] Figure 2 The two-dimensional Gaussian population distribution probability density of the four districts in the main urban area in the specific example of Example 1 of the present invention;

[0047] Figure 3 This is the 50m×50m resolution spatial distribution of the population of a certain street based on the construction of a multi-ring buffer zone in the specific example of Example 1 of the present invention;

[0048] Figure 4 This is a population distribution map of the main urban area and its partial display constructed based on the random forest algorithm in the specific example of Example 1 of the present invention;

[0049] Figure 5 These are the population spatial distribution characteristics of the main urban area at different time periods on an hourly scale in a specific example of Example 1 of the present invention. DETAILED DESCRIPTION

[0050] The accompanying drawings are for illustrative purposes only and are not to be construed as limiting this patent;

[0051] It should be clear that the embodiments described are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments of the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the embodiments of the present application.

[0052] The terms used in the embodiments of the present application are for the purpose of describing specific embodiments only and are not intended to limit the embodiments of the present application. The singular forms "a," "the," and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0053] When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims. In the description of the present application, it should be understood that the terms "first", "second", "third", etc. are only used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence, nor can they be understood as indicating or implying relative importance. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to the specific circumstances.

[0054] In addition, in the description of this application, unless otherwise specified, "plurality" refers to two or more. "And / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, and B exists alone. The character " / " generally indicates that the associated objects are in an "or" relationship. The present invention is further described below with reference to the accompanying drawings and examples.

[0055] In order to solve the limitations of the prior art, this embodiment provides a technical solution, which will be further described below in conjunction with the accompanying drawings and embodiments.

[0056] Example 1

[0057] Please refer to Figure 1 , a method for estimating the spatiotemporal distribution of population at the hourly scale, comprising the following steps:

[0058] S1: Obtain census data, POI data, building area and height data, road network data, and Tencent location data for each administrative unit in the target area, perform coordinate conversion on the above data, and match and spatially connect them with the boundaries of each administrative unit in the target area;

[0059] S2, based on the processing result of step S1, using the population correlation of the POI data to find the population density peak point of each administrative unit;

[0060] S3, by calculating the probability density of the two-dimensional Gaussian distribution of the population density peak point of each administrative unit, the preliminary population spatial distribution based on the two-dimensional Gaussian model is realized;

[0061] S4, establishing a multi-ring buffer zone for each administrative unit in the target area with the population density peak point of each administrative unit as the center; obtaining the population number, building area data and height data, road network length data, and various POI data in each buffer zone based on the processing results of step S1;

[0062] S5, based on the preliminary population spatial distribution obtained in step S3 and using the processing result of step S4 as an input variable, a random forest model is used to obtain population prediction data at a preset spatial resolution;

[0063] S6, fusing the population prediction data obtained in step S5 with the Tencent location data to achieve urban population distribution at a preset spatial resolution and temporal resolution.

[0064] Compared with the existing technology, the present invention can effectively solve the problems of missing time features and low resolution in existing population distribution data by combining dynamic location service data with population spatial prediction data constructed based on multi-ring buffers, and achieve a detailed characterization of the spatiotemporal characteristics of population distribution.

[0065] Specifically, the spatial resolution and the temporal resolution may be set to 50m spatial resolution and 1h temporal resolution.

[0066] The POI data includes the attributes of the name, type, longitude and latitude of 20 types of POI in the target area; the Tencent location data includes the attributes of time, longitude, latitude and positioning times. In step S1, the POI data will also be screened and regionally intercepted. Optionally, the above data can be matched with the boundaries of each administrative unit and spatially connected in ArcGIS.

[0067] As a preferred embodiment, step S2 includes the following process:

[0068] Calculate the TF-IDF value of the POI data in each administrative unit within the target area; perform Pearson correlation analysis on the calculated TF-IDF value and the census data to obtain correlation coefficients for various types of POIs in the POI data; use the POI type with the highest correlation coefficient as the POI type with the highest population correlation, and use the distribution centroid of this type of POI point as the peak population density point of the corresponding administrative unit.

[0069] Furthermore, the TF-IDF value is calculated according to the following formula:

[0070]

[0071]

[0072]

[0073] in, is the TF-IDF value of the i-th type of POI in the j-th administrative unit, n ij is the number of POIs of type i in the jth administrative unit, n k,j is the number of POIs of type k in the jth administrative unit, |D| is the total number of census units, |{j:t i ∈d j}| is the number of census units containing the i-th type of POI.

[0074] Furthermore, the Pearson correlation analysis was performed according to the following formula:

[0075]

[0076] Among them, TI ij is the TF-IDF value of the i-th type of POI in the j-th administrative unit, PD j is the population of the j-th administrative unit.

[0077] Furthermore, in step S2, the POI type with the highest correlation coefficient is used as the POI type with the highest population correlation, and the distribution centroid of the POI type is used as the population density peak point of the corresponding administrative unit. The distribution centroid of the POI type is obtained by the following formula:

[0078]

[0079]

[0080] Among them, x c ,y c are the longitude and latitude of the centroid of the administrative unit, N is the number of POIs of the type with the highest population correlation in the administrative unit, and x i ,y i are the longitude and latitude of the i-th POI point with the highest population correlation in the administrative unit.

[0081] Assuming that the population distribution within the administrative unit conforms to the two-dimensional Gaussian probability distribution, the center of the population density distribution probability is the population density peak point obtained in step S2, and the population distribution probability at any position in the administrative unit satisfies the two-dimensional Gaussian distribution.

[0082] Then, as a preferred embodiment, in step S3, the probability density of the two-dimensional Gaussian distribution is calculated for the population density peak point of each administrative unit according to the following formula:

[0083]

[0084]

[0085]

[0086] Among them, f i (x,y) is the population distribution probability at position (x,y) in the ith census block, and x and y follow normal distributions respectively. and ρ is the correlation coefficient between and, μ x ,μ y is the longitude and latitude of the peak population density point of the census block, Dist(loc(x,y),loc(μ x ,μ y )) is the distance between the census block location (x, y) and the peak population density point.

[0087] By calculating the probability density of the two-dimensional Gaussian distribution for the population density peak point of each administrative unit, the population probability density of any location in the target area can be obtained, thereby realizing the preliminary spatial distribution of the population.

[0088] In step S4, the multi-ring buffer sampling method in ArcMap can be used, with the population density peak point of each administrative unit determined in step S2 as the center of the circle, and the radius of the circle ranging from 25m to 1000m, with a step size of 25m to adjust the radius of the buffer zone. Finally, 40 circular buffer zones are established for each administrative unit. Combined with the building area and height data and road network data corresponding to the target area, the population number, building area and height data, road network length and the number of various POIs in each buffer zone are finally obtained.

[0089] In step S5, based on the preliminary population spatial distribution obtained in step S3, the population number in the buffer zone is used as the label value of the random forest, and the number of 20 types of POIs, building area and height data, and road network length in the buffer zone are used as the feature values ​​of the random forest. The random forest algorithm is run to obtain predicted data on the population distribution of a 50m×50m grid.

[0090] As a preferred embodiment, in step S5, in order to adjust the accuracy of the prediction result, the predicted grid population data is adjusted and corrected, and the population prediction data is corrected according to the following formula:

[0091]

[0092] Among them, Pop adjust Population forecast data after correction cal is the population prediction data output by the random forest model, N census is the census data in the corresponding administrative unit, N cal is the sum of the population prediction data output by the random forest model in the corresponding administrative unit.

[0093] Therefore, by constructing a multi-ring buffer zone for sampling in the target area, the random forest algorithm was used to achieve the transition from preliminary population spatial distribution to refined grid spatial distribution with a resolution of 50m.

[0094] As a preferred embodiment, in step S6, the population prediction data obtained in step S5 is fused with the Tencent location data in the following manner:

[0095] Using the weight coefficient method, find the Tencent location data closest to any grid in the target area, compare the positioning data of Tencent location data in any time period with the average positioning data of the preset time length, obtain the positioning data variation coefficient within the preset time length, multiply the population prediction data at the preset spatial resolution by the positioning data variation coefficient, and obtain the urban population distribution at the preset spatial resolution and time resolution.

[0096] Furthermore, the specific gravity coefficient method is implemented according to the following formula:

[0097]

[0098] Pop ik =W ij ·P k ;

[0099] Among them, TP ij is the number of positioning times of Tencent location data in the jth grid in the i-th period, W ij is the proportion of the positioning times in the jth grid in the i-th period to the total area, n is the number of regional grids, P k is the population prediction data in the kth grid obtained in step S5, Pop ik is the final population count in the kth grid in the i-th period.

[0100] Specifically, we can find the Tencent location data closest to any 50×50m grid in the target area, compare its positioning data for any hour with the 24-hour average positioning data, and obtain the 24-hour positioning data variation coefficient. Multiply the predicted 50×50m grid population data by this coefficient to obtain the 24-hour dynamic population data for the 50×50m grid. This allows us to achieve urban population distribution with a 50m spatial resolution and a 1-hour temporal resolution.

[0101] Next, this embodiment will be further explained with reference to specific examples and data, with the main urban areas of Guangzhou (Yuexiu District, Haizhu District, Tianhe District, and Liwan District) as the target area:

[0102] In step S1, population census data, POI data, building area and height data, road network data, and Tencent location data for Guangzhou's main urban area are obtained. The population census data comes from the seventh population census conducted by the Guangzhou Municipal Bureau of Statistics. The POI data, including the name, type, longitude, and latitude of 20 types of points of interest within the research scope, is obtained through the AutoNavi map service, totaling over 140,000 items. For each type of data, its scope within Guangzhou's main urban area is captured and matched to administrative unit boundaries. The POI data is then converted to the WGS-84 coordinate system. The data layers are spatially connected in ArcGIS to form a comprehensive layer for Guangzhou's main urban area.

[0103] In step S2, the TF-IDF values ​​of 20 POI types in Guangzhou's main urban area were calculated, and the Pearson correlation coefficients with the main urban area population were calculated and ranked. The results are shown in Table 1 (Note: ** indicates data with a significant correlation at the 0.01 level, and * indicates data with a significant correlation at the 0.05 level). At the main urban area scale, the three POI types with the highest population correlation coefficients are auto repair, sports and leisure services, and commercial and residential.

[0104] Table 1 Pearson correlation coefficient ranking between TF-IDF values ​​and population density of different types of POI data in the main urban area of ​​Guangzhou

[0105]

[0106]

[0107] In step S3, after determining the POI type with the highest correlation coefficient in each district, the centroid of the POI data distribution is used as the population density peak point of the district to achieve preliminary population distribution based on the two-dimensional Gaussian distribution model. The probability distribution of population density in each administrative district is as follows: Figure 2 shown

[0108] In step S4, a multi-ring buffer zone is established for each district, and the population, building area and height data, road network length data, and various POI data within each buffer zone are obtained. The specific process is as follows:

[0109] (1) Select an area i and calculate the coordinate center point p of the POI in the area that is most relevant to population density calculated in step S3. i , let p i as the centroid of the region;

[0110] (2) With p i With 25 m as the center, 25 m as the minimum radius, 25 m as the step length, and 1000 m as the maximum radius, a circular sampling area is constructed. That is, a circular sampling area with a radius of r = (25 m, 50 m, 75 m, ..., 1000 m) is constructed for each area. The jth circular sampling area in area i is recorded as p ij ;

[0111] (3) The jth circular sampling area in area i is denoted as p ij Sampling of input and output data, sampling p ij Population density, building area and height data, road network length data and various POI density data;

[0112] (4) The process (3) was repeated until all data samples in the four main urban districts of Guangzhou were collected. Finally, a training set containing 3200 data items was constructed and abnormal data were removed.

[0113] The random forest model was run on the obtained population data and covariates to obtain the predicted population distribution data under 50×50m grid and calibrate it. The specific steps are as follows:

[0114] The model was trained using the random forest algorithm, with 100 decision trees. 70% of the samples were used as training data, and 30% as testing data for cross-validation. After model training, the model was compared with census data. The overall accuracy evaluation results are shown in Table 2. It can be seen that the population prediction based on the multi-ring buffer system is highly accurate, with a mean absolute percentage error (MAPE) of 8.52%.

[0115] Table 2 Population prediction accuracy results at 50m spatial resolution

[0116] Pearson R <![CDATA[Standard R 2 ]]> MAPE 0.8926 0.7856 8.52％

[0117] Taking Beijing Street in Yuexiu District as the test object, the spatial population distribution model trained based on the random forest model was applied to Beijing Street in Yuexiu District to predict the population distribution map of the 50m×50m grid in the area. Due to the irregularity of the regional boundaries and incomplete grids, the area weighting method was used to redistribute the population predicted by the grids, and finally a complete grid population distribution was obtained. The final prediction results can be seen Figure 3 .Will Figure 3 The population data for the predicted grid was tallied and summed, and compared with the census data for Beijing Subdistrict in Yuexiu District to determine the deviation from the statistical results. Overall, the predicted total population for Beijing Subdistrict in Yuexiu District was 47,350, while the census result was 43,632, a decrease of 3,717 people, for a deviation of approximately 8.52%.

[0118] The population spatialization model based on the multi-ring buffer zone was applied to the main urban area of ​​Guangzhou. The population spatial distribution at a spatial resolution of 50m was finally obtained. Figure 4 .

[0119] In step S6, the population spatialization prediction data obtained based on the multi-ring buffer model is combined with Tencent location data to extract dynamic high-precision population distribution and structural characteristics. The specific steps are as follows:

[0120] (1) This example uses Python crawler technology to collect Tencent location data for Guangzhou from November 27, 2020, to December 5, 2020, with an interval of 10 minutes between each collection. The collected data includes four fields: time, longitude, latitude, and number of positioning times. The data accuracy is 0.05 arc seconds, which is converted to a distance unit of approximately 5 km × 5 km.

[0121] (2) Using the weight coefficient method, find the Tencent location data closest to any 50×50m grid in the study area, compare its location data in any hour with the average location data of 24 hours, and obtain the 24-hour location data change coefficient. Multiply the predicted 50×50m grid population data by this coefficient to obtain the 24-hour dynamic change 50×50m grid population data.

[0122] The population spatial distribution results are integrated with Tencent location data to obtain the population distribution spatial characteristics of Guangzhou’s main urban area at different times. Figure 5 shown.

[0123] Example 2

[0124] A computer storage medium stores a computer program, which, when executed by a processor, implements the steps of the method for estimating the spatiotemporal distribution of population on an hourly scale in embodiment 1.

[0125] Example 3

[0126] A computer device includes a storage medium, a processor, and a computer program stored in the storage medium and executable by the processor. When the computer program is executed by the processor, the steps of the method for estimating the spatiotemporal distribution of population on an hourly scale in Example 1 are implemented.

[0127] Obviously, the above embodiments of the present invention are merely examples for the purpose of clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Those skilled in the art will appreciate that other variations or modifications can be made based on the above description. It is not necessary and impossible to enumerate all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the scope of protection of the claims of the present invention.

Claims

1. A method for estimating the spatiotemporal distribution of population at the hourly scale, characterized in that: The following steps are involved: S1: Obtain census data, POI data, building area and height data, road network data, and Tencent location data for each administrative unit in the target area, perform coordinate conversion on the above data, and match and spatially connect them with the boundaries of each administrative unit in the target area; S2, based on the processing result of step S1, using the population correlation of the POI data to find the population density peak point of each administrative unit; S3, by calculating the probability density of the two-dimensional Gaussian distribution of the population density peak point of each administrative unit, the preliminary population spatial distribution based on the two-dimensional Gaussian model is realized; S4, establishing a multi-ring buffer zone for each administrative unit in the target area with the population density peak point of each administrative unit as the center; obtaining the population number, building area data and height data, road network length data, and various POI data in each buffer zone based on the processing results of step S1; S5, based on the preliminary population spatial distribution obtained in step S3 and using the processing result of step S4 as an input variable, a random forest model is used to obtain population prediction data at a preset spatial resolution; S6, fusing the population prediction data obtained in step S5 with the Tencent location data to achieve urban population distribution at a preset spatial resolution and temporal resolution; In step S6, the population prediction data obtained in step S5 is fused with the Tencent location data in the following manner: Using the weight coefficient method, find the Tencent location data closest to any grid in the target area, compare the positioning data of Tencent location data in any time period with the average positioning data of the preset time length, obtain the positioning data variation coefficient within the preset time length, multiply the population prediction data at the preset spatial resolution by the positioning data variation coefficient, and obtain the urban population distribution at the preset spatial resolution and time resolution.

2. The method for estimating the spatiotemporal distribution of population at the hourly scale according to claim 1, characterized in that: The step S2 includes the following processes: Calculating the TF-IDF value of the POI data in each administrative unit within the target area; performing Pearson correlation analysis on the calculated TF-IDF value and the census data to obtain the correlation coefficient of each type of POI in the POI data; The POI type with the highest correlation coefficient is taken as the POI type with the highest population correlation, and the distribution centroid of this type of POI point is taken as the population density peak point of the corresponding administrative unit.

3. The method for estimating the spatiotemporal distribution of population at the hourly scale according to claim 2, characterized in that: The TF-IDF value is calculated using the following formula: in, is the TF-IDF value of the i-th type of POI in the j-th administrative unit, n ij is the number of POIs of type i in the jth administrative unit, n k,j is the number of POIs of type k in the jth administrative unit, |D| is the total number of census units, |{j:t i ∈d j }| is the number of census units containing the i-th type of POI.

4. The method for estimating the population spatiotemporal distribution at the hourly scale according to claim 2, characterized in that: The Pearson correlation analysis was performed according to the following formula: Among them, TI ij is the TF-IDF value of the i-th type of POI in the j-th administrative unit, PD j is the population of the j-th administrative unit.

5. The method for estimating the spatiotemporal distribution of population at the hourly scale according to claim 2, wherein: In step S2, the POI type with the highest correlation coefficient is used as the POI type with the highest population correlation, and the distribution centroid of the POI type is used as the population density peak point of the corresponding administrative unit. The distribution centroid of the POI type is obtained by the following formula: Among them, x c ,y c are the longitude and latitude of the centroid of the administrative unit, N is the number of POIs of the type with the highest population correlation in the administrative unit, and x i ,y i are the longitude and latitude of the i-th POI point with the highest population correlation in the administrative unit.

6. The method for estimating the population spatiotemporal distribution at the hourly scale according to claim 1, characterized in that: In step S3, the probability density of the two-dimensional Gaussian distribution is calculated for the population density peak point of each administrative unit according to the following formula: Among them, f i (x,y) is the population distribution probability at position (x,y) in the ith census block, and x and y follow normal distributions respectively. and ρ is the correlation coefficient between and, μ x ,μ y is the longitude and latitude of the peak population density point of the census block, Dist(loc(x,y),loc(μ x ,μ y )) is the distance between the census block location (x, y) and the peak population density point.

7. The method for estimating the spatiotemporal distribution of population at the hourly scale according to claim 1, characterized in that: In step S5, the population prediction data is corrected according to the following formula: Among them, Pop adjust Population forecast data after correction cal is the population prediction data output by the random forest model, N census is the census data in the corresponding administrative unit, N cal is the sum of the population prediction data output by the random forest model in the corresponding administrative unit.

8. The method for estimating the spatiotemporal distribution of population at the hourly scale according to claim 1, characterized in that: The specific gravity coefficient method is implemented according to the following formula: Pop ik =W ij ·P k ; Among them, TP ij is the number of positioning times of Tencent location data in the jth grid in the i-th period, W ij is the proportion of the positioning times in the jth grid in the i-th period to the total area, n is the number of regional grids, P k is the population prediction data in the kth grid obtained in step S5, Pop ik is the final population count in the kth grid in the i-th period.

9. A computer storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for estimating the spatiotemporal distribution of population at the hourly scale as claimed in any one of claims 1 to 8 are implemented.

Images