Community health service center site selection method and system in urban high-density environment

Abstract

The invention relates to a community health service center site selection method and system in an urban high-density environment, and the method comprises the steps: obtaining the data of a target region in the urban high-density environment, and forming a first sample data set; arranging the first sample data set to obtain a second sample data set; obtaining community health service center site selection related indexes according to the second sample data set, and further obtaining a third sample data set; sampling the third sample data set to obtain a fourth sample data set, and obtaining a community health service center site selection model through parameter tuning and feature combination preferential selection; and inputting the characteristic value data set of the third sample data set into the model to obtain alternative points, and determining a service center site selection point set capable of covering the most demand points by using a maximum coverage model. According to the method, the site selection accuracy and the service coverage rate of the health service center can be effectively improved.

Description

Technical Field

[0001] The present disclosure relates to the technical field of urban planning big data information processing, and in particular to a method and system for site selection of community health service centers in a high-density urban environment. Background Art

[0002] Community health service centers are the foundation of the urban medical and health service system and the "gatekeepers" of people's health. Setting up public health units based on the 15-minute living circle of the community, strengthening the planning and construction of community health service centers, and optimizing the level of grassroots health service facilities and resource allocation are of great significance to meeting the people's health service needs and improving the health level of residents. Existing studies mainly propose site selection and planning strategies for community health service centers, focusing on qualitative discussions on factors affecting site selection or using ArcGIS for basic analysis of basic geographic information such as surrounding traffic network and land use efficiency. Quantitative research and application of multi-source data systems based on the study area are not sufficient. In addition, in the existing site selection methods, the existing measurement factors are mainly based on accessibility, population density, service scope, etc., which are not comprehensive for the site selection of community health service centers in the high-density urban built environment, and cannot consider multi-dimensional factors such as population characteristics, natural environment, resident accessibility, built environment and urban functions; in the quantitative research of site selection decision-making, existing studies mostly use subjective analytic hierarchy process (AHP) as a multi-objective decision-making method, lacking the systematic and objective use of multi-dimensional multi-source data, considering multiple machine learning algorithms and combining them with the maximum coverage model for modeling site selection, and exploring the application of model parameter and feature selection and combination optimization.

[0003] Therefore, it is necessary to propose a site selection method for community health service centers in high-density urban environments to effectively improve the accuracy of site selection and service coverage of community health service centers. Summary of the invention

[0004] In order to solve the problem that the traditional site selection method does not accurately select the location of the service center, resulting in low health service coverage, etc., this paper proposes a site selection method for community health service centers in a high-density urban environment, which comprehensively considers the five dimensions of population characteristics, accessibility, built environment, natural environment, and urban functions, and uses "multi-source data + machine learning + maximum coverage model" combined with "algorithm optimization + parameter tuning + feature combination optimization" to make more systematic and accurate decisions on the site selection of community health service centers.

[0005] According to one aspect of the present disclosure, a method for selecting a site for a community health service center in a high-density urban environment is provided, comprising: S10, obtaining network open source data, population grid data, mobile phone signaling data, point of interest / surface of interest data, night light data, building unit data, normalized difference vegetation index data, PM2.5 concentration data and urban road data of the target area in a high-density urban environment to form a first sample data set; S20, sorting and correcting the data in the first sample data set to obtain a second sample data set; S30. Calculate the site selection related indicators of the community health service center according to the second sample data set, and normalize the related indicators to obtain a third sample data set, wherein the related indicators include one or more of the following indicators: population density index, population structure index, general hospital accessibility index, public transportation convenience index, subway convenience index, residential area accessibility index, building density index, floor area ratio index, night light index, normalized vegetation index, surrounding function density index, surrounding function diversity index, general hospital index and residential area index; S40, sampling according to the target value of the samples in the third sample data set to obtain a fourth sample data set, based on the fourth sample data set, using the optimal machine learning algorithm to construct a preliminary model for the site selection of community health service centers, using grid search to perform parameter tuning, and outputting the best parameter combination through K-fold cross validation to obtain a preliminary model for the site selection of community health service centers; S50, evaluating the preliminary model for site selection of community health service centers, calculating the contribution of relevant indicators in the preliminary model for site selection of community health service centers, optimizing feature selection and combination by Monte Carlo simulation, and obtaining the best feature combination for the preliminary model for site selection of community health service centers; S60, based on the optimal feature combination, using the optimal machine learning algorithm to construct a community health service center site selection model, perform parameter tuning on the community health service center site selection model, and output the optimal parameter combination through K-fold cross validation to obtain the community health service center site selection model under the optimal feature combination and the optimal parameter combination training; S70, evaluating the obtained community health service center site selection model, and calculating the contribution of relevant indicators in the community health service center site selection model; S80. Export and save the community health service center site selection model, input the eigenvalue data set in the third sample data set in the target area into the community health service center site selection model, obtain the community health service center site selection candidate points, determine the number of community health service centers to be added according to the population size of the target area, and use the maximum coverage model to determine the set of community health service center site selection points that can cover the most demand points among the community health service center site selection candidate points.

[0006] Preferably, arranging and correcting the data in the first sample data set includes: Input a first sample data set, and perform random value assignment processing on the mobile phone signaling data in the first sample data set; use a line chart to draw the data of each feature in the first sample data set to determine whether there are outliers in the data of each feature that deviate far from the normal range, and delete the outliers; deduplicate the same data in the first sample set; and check and correct the urban road data in the first sample data set.

[0007] Preferably, the community health service center site selection related indicators calculated according to the second sample data set specifically include: Convert building unit surface data into geometric center point data of the building unit surface; According to the population grid data, the population in the grid is counted to obtain the population density index; According to the mobile phone signaling data, the user age structure distribution is counted to obtain the population structure index; Based on the data of points of interest / areas of interest, calculate the accessibility index of general hospitals, bus convenience index, subway convenience index and residential accessibility index; Calculate building density index and floor area ratio index based on building unit data; Calculate the night light index indicator based on the night light data; According to the normalized vegetation index data, a grid layer containing normalized vegetation index information is obtained, and the normalized vegetation index index is calculated; Based on the data of points of interest / areas of interest, the surrounding functional density index, surrounding functional diversity index, general hospital index and residential area index are calculated.

[0008] Preferably, the relevant indicators are normalized to obtain a third sample data set, which is expressed as: , In the formula, X is the original data, For normalized data, and are the maximum and minimum values ​​in the original data respectively.

[0009] Preferably, the fourth sample data set is obtained by sampling the target values ​​of the samples in the third sample data set, and based on the fourth sample data set, an optimal machine learning algorithm is used to construct a preliminary model for site selection of community health service centers, including: Importing the third sample data set into the table, oversampling the samples with a target value of 1 in the third sample data set, undersampling the samples with a target value of 0, sorting the samples with target values ​​of 1 and 0, and generating a fourth sample data set; Use 80% of the fourth sample data set as a training set and 20% as a test set; A variety of machine learning algorithms were used to test the model construction, and the optimal machine learning algorithm with the best goodness of fit and root mean square error was determined based on the test results. The optimal machine learning algorithm was selected to construct a preliminary model for the site selection of community health service centers.

[0010] Preferably, the maximum coverage model is used to determine a set of community health service center site selection points that can cover the most demand points among the community health service center site selection candidate points. The maximum coverage model is expressed as: , The constraint formula of the maximum coverage model is: , , , In the formula, max S is the sum of the covered demand points, i As demand points, residential areas are taken as demand points. j Select alternative sites for community health service centers. Is the demand point covered? To cover the demand points i The set of candidate points, m is the number of alternative points, M Limit the number of locations for community health service centers. Whether to use the alternative point as the site selection point.

[0011] According to one aspect of the present disclosure, a community health service center site selection system in a high-density urban environment is provided, comprising: The data acquisition module acquires network open source data, population grid data, mobile phone signaling data, point of interest / surface of interest data, night light data, building unit data, normalized vegetation index data, PM2.5 concentration data and urban road data of the target area in a high-density urban environment to form a first sample data set; A second sample set acquisition module, which sorts and corrects the data in the first sample set to obtain a second sample set; The relevant indicator processing module calculates the community health service center site selection related indicators according to the second sample data set, and normalizes the related indicators to obtain a third sample data set, wherein the related indicators include one or more of the following indicators: population density index, population structure index, general hospital accessibility index, public transportation convenience index, subway convenience index, residential area accessibility index, building density index, floor area ratio index, night light index, normalized vegetation index, surrounding function density index, surrounding function diversity index, general hospital index and residential area index; A preliminary model building module for site selection of community health service centers, sampling a fourth sample data set according to the target value of the sample in the third sample data set, constructing a preliminary model for site selection of community health service centers based on the fourth sample data set using an optimal machine learning algorithm, using grid search to perform parameter tuning, and outputting the best parameter combination through K-fold cross validation to obtain a preliminary model for site selection of community health service centers; Optimizing the feature selection and combination module, evaluating the preliminary model for site selection of community health service centers, and calculating the contribution of relevant indicators in the preliminary model for site selection of community health service centers, optimizing the feature selection and combination by Monte Carlo simulation, and obtaining the best feature combination for the preliminary model for site selection of community health service centers; A community health service center site selection model construction module, based on the optimal feature combination, uses the optimal machine learning algorithm to construct a community health service center site selection model, performs parameter tuning on the community health service center site selection model, and outputs the optimal parameter combination through K-fold cross validation to obtain a community health service center site selection model under the optimal feature combination and the optimal parameter combination training; A model evaluation module is used to evaluate the obtained community health service center site selection model and calculate the contribution of relevant indicators in the community health service center site selection model; The site selection point set determination module exports and saves the community health service center site selection model, inputs the characteristic value data set in the third sample data set in the target area into the community health service center site selection model, obtains the community health service center site selection candidate points, determines the number of community health service centers to be added according to the population size of the target area, and uses the maximum coverage model to determine the community health service center site selection point set that can cover the most demand points among the community health service center site selection candidate points.

[0012] According to one aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to: execute the above-mentioned method for site selection of community health service centers in a high-density urban environment.

[0013] According to one aspect of the present disclosure, a computer-readable storage medium is provided, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the method for selecting a site for a community health service center in a high-density urban environment is implemented.

[0014] Compared with the prior art, the beneficial effects of the present disclosure are: 1) This paper proposes a method for site selection of community health service centers in a high-density urban environment, obtains multi-source data in five dimensions of population characteristics, accessibility, built environment, natural environment, and urban functions in the target area under a high-density urban built environment, and obtains a fourth sample data set by performing multiple processing on the data, so as to construct a systematic, objective and comprehensive feature indicator system for the site selection model of community health service centers based on machine learning.

[0015] 2) This paper uses five machine learning regression algorithms, LightGBM, XGBoost, CatBoost, NGBoost and random forest, to build and test the model, and obtains the optimal machine learning algorithm with the best R-squared (goodness of fit) and RMSE (root mean square error), and uses network search to tune parameters. Based on the contribution of each feature obtained by calculating the shap value, Monte Carlo simulation is used to optimize feature selection and combination, and then a community health service center site selection model is obtained. This process can effectively improve the accuracy of community health service center site selection.

[0016] 3) The present invention inputs the characteristic value data set in the third sample data set in the target area into the community health service center site selection model to obtain alternative site selection points for the community health service center, and uses the maximum coverage model to determine a set of community health service center site selection points that can cover the most demand points among the alternative site selection points for the community health service center. This process can ensure the coverage rate of community health services in the target area.

[0017] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

[0018] Further features and aspects of the present disclosure will become apparent from the following detailed description of exemplary embodiments with reference to the attached drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings herein are incorporated into the specification and constitute a part of the specification. These drawings illustrate embodiments consistent with the present disclosure and are used to illustrate the technical solutions of the present disclosure together with the specification.

[0020] Figure 1 Shows the flow chart of the site selection method for community health service centers in high-density urban environments; Figure 2A schematic diagram showing the contribution of each feature of the preliminary model for site selection of a community health service center in an embodiment of the present disclosure is shown; Figure 3 A schematic diagram of the cross-validation score of the model under different feature combinations in the embodiment of the present disclosure is shown; Figure 4 A schematic diagram showing the contribution of each feature of the community health service center site selection model in the embodiment of the present disclosure is shown; Figure 5 A schematic diagram of a set of site selection points for community health service centers that can cover the most demand points in an embodiment of the present disclosure is shown; Figure 6 The structural block diagram of the community health service center site selection system in a high-density urban environment is shown. DETAILED DESCRIPTION

[0021] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. The same reference numerals in the accompanying drawings represent elements with the same or similar functions. Although various aspects of the embodiments are shown in the accompanying drawings, the drawings are not necessarily drawn to scale unless otherwise specified.

[0022] The word “exemplary” is used exclusively herein to mean “serving as an example, example, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.

[0023] The term "and / or" herein is only a description of the association relationship of the associated objects, indicating that there may be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the term "at least one" herein represents any combination of at least two of any one or more of a plurality of. For example, including at least one of A, B, and C can represent including any one or more elements selected from the set consisting of A, B, and C.

[0024] In addition, in order to better illustrate the present disclosure, numerous specific details are given in the following specific embodiments. It should be understood by those skilled in the art that the present disclosure can also be implemented without certain specific details. In some examples, methods, means, components and circuits well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present disclosure.

[0025] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0026] Example 1 Based on the above ideas, the present invention proposes a method for site selection of community health service centers in a high-density urban environment. Figure 1 A flow chart showing a method for selecting a site for a community health service center in a high-density urban environment. The method comprises: S10, obtaining network open source data, population grid data, mobile phone signaling data, point of interest / surface of interest data, night light data, building unit data, normalized difference vegetation index data, PM2.5 concentration data and urban road data of the target area in a high-density urban environment to form a first sample data set; S20, sorting and correcting the data in the first sample data set to obtain a second sample data set; S30. Calculate the site selection related indicators of the community health service center according to the second sample data set, and normalize the related indicators to obtain a third sample data set, wherein the related indicators include one or more of the following indicators: population density index, population structure index, general hospital accessibility index, public transportation convenience index, subway convenience index, residential area accessibility index, building density index, floor area ratio index, night light index, normalized vegetation index, surrounding function density index, surrounding function diversity index, general hospital index and residential area index; S40, sampling according to the target value of the samples in the third sample data set to obtain a fourth sample data set, based on the fourth sample data set, using the optimal machine learning algorithm to construct a preliminary model for the site selection of community health service centers, using grid search to perform parameter tuning, and outputting the best parameter combination through K-fold cross validation to obtain a preliminary model for the site selection of community health service centers; S50, evaluating the preliminary model for site selection of community health service centers, calculating the contribution of relevant indicators in the preliminary model for site selection of community health service centers, optimizing feature selection and combination by Monte Carlo simulation, and obtaining the best feature combination for the preliminary model for site selection of community health service centers; S60, based on the optimal feature combination, using the optimal machine learning algorithm to construct a community health service center site selection model, perform parameter tuning on the community health service center site selection model, and output the optimal parameter combination through K-fold cross validation to obtain the community health service center site selection model under the optimal feature combination and the optimal parameter combination training; S70, evaluating the obtained community health service center site selection model, and calculating the contribution of relevant indicators in the community health service center site selection model; S80, exporting and saving the community health service center site selection model, inputting the characteristic value data set in the third sample data set in the target area into the community health service center site selection model, obtaining candidate site selection points for the community health service center, determining the number of community health service centers to be added according to the population size of the target area, and using the maximum coverage model to determine a set of community health service center site selection points that can cover the most demand points among the candidate site selection points for the community health service center; The present disclosure provides a method for selecting a site for a community health service center in a high-density urban environment. The method targets a target area of ​​Yuexiu District, Guangzhou, and includes the following steps: S10. Obtain network open source data, population grid data, mobile phone signaling data, point of interest / area of ​​interest data, night light data, building unit data, normalized vegetation index data, PM2.5 concentration data and urban road data of the target area in a high-density urban environment to form a first sample data set.

[0027] Furthermore, step S10 specifically includes: S101. Obtain open source data from the Internet: obtain the information table of Guangzhou Yuexiu District Community Health Service Center from the website of Guangzhou Municipal Health Commission, and obtain basic administrative maps from the National Geographic Information Public Service Platform; S102. Obtain population grid data: Obtain national population grid data with 100-meter accuracy from the worldpop open source; S103, obtaining mobile phone signaling data: applying for relevant mobile phone signaling data from China Unicom Smart Footprint; S104, obtaining point of interest / area of ​​interest data: obtaining POI data information of Guangzhou City on Baidu Map based on Python web crawler technology, the data includes leisure and activity facilities, cultural and sports facilities, transportation ancillary facilities, catering service facilities, hotel accommodation facilities, commercial facilities, government agencies, etc.; S105, obtaining urban road data: obtaining road line elements and road name and type data from the OSM open source wiki map; S106, obtaining building unit data: obtaining building unit surface data and height data from Guangzhou building CAD data; S107. Obtain night light data: Obtain global night light data with an accuracy of 500 meters from the National Earth System Science Data Center shared service platform; S108. Obtain PM2.5 concentration data: Obtain national PM2.5 concentration data with an accuracy of 1 km from the National Earth System Science Data Center shared service platform; S109. Obtaining Normalized Difference Vegetation Index data: Obtaining national 30-meter-precision Normalized Difference Vegetation Index data from the National Science and Technology Infrastructure Platform-National Ecological Science Data Center; S20: Arrange and correct the data in the first sample data set to obtain a second sample data set.

[0028] In this embodiment, the data in the first sample data set is sorted and corrected, including: Input a first sample data set, and perform random value assignment processing on the mobile phone signaling data in the first sample data set; use a line chart to draw the data of each feature in the first sample data set to determine whether there are outliers in the data of each feature that deviate far from the normal range, and delete the outliers; deduplicate the same data in the first sample set; and check and correct the urban road data in the first sample data set.

[0029] In this embodiment, step S20 specifically includes: S201, data input: input the collected community health service center information into an Excel table, read the information in the Excel table by writing a Python program, obtain the GCJ-02 geographic coordinates of the community health service centers in batches in the geocoding application of the AutoNavi open platform, and then convert the GCJ-02 geographic coordinates into WGS-1984 geographic coordinates. After data correction and cleaning, obtain the coordinate list of the community health service centers; S202, random value processing: if the population of a single grid in the mobile phone signaling data is less than 5, it is uniformly represented by "<5". Therefore, this embodiment performs random value processing on this part of the data and takes a random value between 0 and 5; S203, outlier deletion: Draw a line chart for the data value of each feature, determine whether there is an outlier that deviates far from the normal range in each feature, and delete the outlier; S204, duplicate value processing: performing a deduplication operation on the same data in the first sample set; S205, urban road data processing: remove the data of highways, expressways, viaducts, tunnels and other non-walkable roads, build a road network buffer through ArcGIS software and extract the road network centerline through vectorization method to build a single-line road network, perform topological check on road data, and correct double-line overlap and pseudo-node errors; S206, sorting to obtain a second sample data set: sorting the preprocessed first sample data set and visualizing it in ArcGIS to obtain a second sample data set.

[0030] Further, step S206 specifically includes: S20601. Use the "Add XY Data" tool in ArcGIS software to set the X field to "lon" and the Y field to "lat" for the community health service center coordinate data, urban road data, building unit data, point of interest / area of ​​interest data, and mobile phone signaling data, export the data to Shapefile files, and visualize them in ArcGIS software; S20602. Use the "Create NetCDF Raster Layer" tool in ArcGIS software for PM2.5 concentration data, set the X field to "lon", set the Y field to "lat", export the raster file, further define the raster file as WGS-1984 geographic coordinates and convert it to a TIFF file and visualize it in ArcGIS software; S20603. Visualize the TIFF files of Normalized Difference Vegetation Index (NDVI) data and night light data in ArcGIS software.

[0031] S30. Calculate the relevant indicators for site selection of community health service centers based on the second sample data set, and normalize the relevant indicators to obtain a third sample data set, wherein the relevant indicators include one or more of the following indicators: population density index, population structure index, general hospital accessibility index, public transportation convenience index, subway convenience index, residential area accessibility index, building density index, floor area ratio index, night light index, normalized vegetation index, surrounding function density index, surrounding function diversity index, general hospital index and residential area index.

[0032] In this embodiment, step S30 specifically includes: S301, using the "element to point" tool in ArcGIS software to convert the building unit surface data into the geometric center point data of the building unit surface; S302, using the service area analysis tool of the "Network Analysis" toolbar of ArcGIS software, constructing a network dataset with urban road data, assigning a walking speed of residents to 60 m / min, and using the community health service center as a supporting facility of the 15-minute community life circle, with the road network distance of residents walking for 15 minutes as its service radius, constructing a 900 m service area of ​​the geometric center point of each building unit in the target area as a 15-minute walking service area based on the road network data, and calculating the area of ​​the 900 m service area of ​​the geometric center point of each building unit; Furthermore, step S302 specifically includes: S30201. Create a new feature dataset in ArcGIS software, and import the processed urban road data in the second sample dataset into the new feature dataset; S30202. Create a new network dataset in the feature dataset and set the value of the cost attribute to the "Shape_Length" field; S30203. Use the service area analysis tool on the "Network Analysis" toolbar, set the facility point to the geometric center point of each building unit in the target area, set the impedance in the analysis settings to length (meters), set the cutoff value to 900, set the multiple facility point option to "overlap", and solve and run to obtain the 900m service area of ​​the geometric center point of each building unit in the target area; S303. Create a 100m*100m grid of the target area based on the population grid data. Use ArcGIS's "Display Zoning Statistics as Table" tool to obtain a population table for each grid and connect it to the grid to obtain a grid layer containing population information. Use ArcGIS's "Spatial Join" tool to count the total population in the grids within the 900m service area, and divide it by the area of ​​the 900m service area to obtain the population density, which is used as the population density index within the 900m service area. S304. Calculate the ratio of the number of residents to the number of working population, the ratio of the number of children and the ratio of the elderly population in each grid according to the mobile phone signaling data, and use the "Spatial Join" tool of ArcGIS to calculate the average ratio of the number of residents to the number of working population, the ratio of the number of children and the ratio of the elderly population in the grids within the 900m service area as the population structure indicator within the 900m service area; S305. Based on the point of interest data, the nearest facility point tool of the "Network Analysis" tool of ArcGIS is used to construct a network dataset with urban road data. All the points of interest of general hospitals in the target area are taken as facility points, and the geometric center points of all building units in the target area are taken as event points. The distance from the geometric center point of each building unit to the nearest general hospital is calculated as the accessibility index of the general hospital. S306. Based on the POI data, the nearest facility point tool of the “Network Analysis” tool of ArcGIS is used to construct a network dataset with urban road data. All bus stops in the target area are taken as facility points, and the geometric center points of all building units in the target area are taken as event points. The distance from the geometric center point of each building unit to the nearest bus stop is calculated as a bus convenience index. S307. Based on the POI data, the nearest facility point tool of the “Network Analysis” tool of ArcGIS is used to construct a network dataset with urban road data. All subway stations in the target area are used as facility points, and all geometric center points of building units in the target area are used as event points. The distance from the geometric center point of each building unit to the nearest subway station is calculated as a subway convenience index. S308. Based on the POI data, the OD cost matrix tool of the "Network Analysis" tool of ArcGIS is used to construct a network dataset with urban road data, with the geometric center points of all building units in the target area as the starting point and all residential areas in the target area as the destination point. The cutoff value is set to 900m, and the sum of the distances from the geometric center points of each building unit to all residential areas within the 900m service area is calculated, and divided by the number of residential areas within the 900m service area as the residential area accessibility index; S309. Use the “buffer analysis” tool in ArcGIS software to construct a 300m buffer zone around the geometric center of each building unit; S310, based on the building unit data, using the “Spatial Join” tool of ArcGIS to count the sum of the ground floor areas of all building units within a 300m buffer zone of the geometric center of each building unit, and dividing it by the area of ​​the 300m buffer zone to obtain the building density as a building density index; S311. According to the building unit data, add a "building area" field in the attribute table, calculate the product of the ground floor area and the number of floors of each building unit to obtain the building area of ​​each building unit, use the "spatial connection" tool of ArcGIS to count the total building area of ​​all building units within the 300m buffer zone of the geometric center point of each building unit, and divide it by the area of ​​the 300m buffer zone to obtain the floor area ratio as the floor area ratio indicator; S312. Create a 100m*100m grid of the target area based on the night light data, obtain the night light index table of each grid through the ArcGIS "Display Zoning Statistics as Table" tool and connect to the grid to obtain a grid layer containing night light index information; use ArcGIS's "Spatial Join" tool to calculate the average night light index in the grids within the 900m service area as the night light index indicator; S313. Create a 30m*30m grid in the target area based on the NDVI data. Obtain the NDVI data table of each grid using the ArcGIS "Display Zonal Statistics as Table" tool and connect it to the grid to obtain a grid layer containing NDVI data information. Use the ArcGIS "Spatial Join" tool to calculate the average value of the NDVI data in the grid within the 900m service area of ​​the geometric center of each building unit as the NDVI indicator. S314. Based on the POI data, use the “Spatial Join” tool of ArcGIS to count the number of all POIs of each type within the 900m service area of ​​the geometric center of each building unit as the peripheral functional density indicator; S315. Create a 200m*200m grid in the target area based on the POI data, use ArcGIS's "Generate Neighbor Table" tool to search for POI data within the 900m buffer of the grid, and export the POI neighbor table to Excel to calculate the POI mixing degree FD around each grid. The calculation formula is: , Where FD is the POI mixing degree around each grid, For Type i The proportion of points of interest in all points of interest, m is the number of POI types; The “Spatial Join” tool of ArcGIS was used to calculate the average POI mixing degree in the grid within the 900m service area of ​​the geometric center of each building unit as an indicator of surrounding functional diversity; S316. Based on the POI data, use the “Spatial Join” tool of ArcGIS to count the number of POIs of general hospitals within the 900m service area of ​​the geometric center of each building unit as the general hospital index; S317. Based on the POI data, use the “Spatial Join” tool of ArcGIS to count the number of POIs in residential areas within the 900m service area of ​​the geometric center of each building unit as the residential area index; S318. The above indicators are summarized and organized, and the indicator system table is shown in Table 1.

[0033]

[0034] S319, normalize the above indicators, and the calculation formula is: , In the formula, X is the original data, For normalized data, and are the maximum and minimum values ​​in the original data respectively.

[0035] S320. Based on the open source data on the Internet, the target value of the geometric center point of the building unit where the community health service center is located is assigned to 1, and the target values ​​of the geometric center points of other building units are assigned to 0, as the target value data of the community health service center characteristics.

[0036] S321. Summarize and organize the above data to obtain a third sample data set, and divide the third sample data set into a characteristic value data set and a target value data set, wherein the characteristic value data set includes population density, ratio of resident population to working population, proportion of children and adolescents, proportion of elderly population, accessibility of general hospitals, convenience of public transportation, convenience of subway, accessibility of residential areas, distance from walkable roads, floor area ratio, building density, night light index, PM2.5 concentration, normalized vegetation index data, surrounding function density, surrounding function diversity (POI mixing degree), number of general hospitals in the service area, and number of residential communities in the service area, represented by the letter X, and ; The target value data set includes the target value data of the community health service center characteristics, represented by the letter y.

[0037] S40. A fourth sample data set is obtained by sampling according to the target values ​​of the samples in the third sample data set. Based on the fourth sample data set, an optimal machine learning algorithm is used to construct a preliminary model for the site selection of community health service centers. Grid search is used for parameter tuning. The optimal parameter combination is output through K-fold cross validation to obtain a preliminary model for the site selection of community health service centers.

[0038] In this embodiment, step S40 specifically includes: S401. Import the third sample data set into an Excel table. Since the number of building units where community health service centers are located in the target area is significantly smaller than the number of other building units, in order to avoid data set imbalance, the samples with a target value of 1 in the third sample data set are oversampled, and the samples with a target value of 0 are undersampled.

[0039] Furthermore, step S401 specifically includes: S40101. Duplicate 50 samples with a target value of 1 to expand the original 24 samples with a target value of 1 to 1200 as oversampled samples.

[0040] S40102. For all samples with a target value of 0, use Excel's RAND function to generate random numbers in a new column, and use Excel's sorting function to sort the random numbers. Select the first 2,400 samples as under-sampling samples in a ratio of 1:2 between the number of samples with a target value of 1 and 0.

[0041] S40103. Arrange samples whose target values ​​are 1 and 0 to generate a fourth sample data set.

[0042] S402, using the sklearn library in Python, calling the model_selection module to use 80% of the fourth sample data set as a training set and 20% as a test set.

[0043] S403. Use five machine learning algorithms, including LightGBM, XGBoost, CatBoost, NGBoost, and random forest, to build models and test them respectively. Select the optimal machine learning algorithm with the best R-squared (goodness of fit) and RMSE (root mean square error) to build a preliminary model for the site selection of community health service centers.

[0044] Furthermore, step S403 specifically includes: S40301. Use the LightGBM library in Python to call the LGBMRegressor method to build a LightGBM regression model. The model parameters are as follows: params_LightGBM{learning_rate, boosting_type, objective, metric, num_leaves, verbose, seed, n_jobs, feature_fraction, bagging_fraction, bagging_freq} Among them, learning_rate represents the learning rate, boosting_type represents the boosting method, objective represents the loss function, metric represents the evaluation indicator, num_leaves represents the number of leaf nodes in each decision tree, seed represents the random seed, feature_fraction represents the subsampling of features, bagging_fraction and bagging_freq are used to use the bagging method.

[0045] In this embodiment, a set of initial parameters of the LightGBM algorithm are defined: params_lgb = {learning_rate: 0.02, boosting_type: gbdt, objective:mse, metric: rmse, num_leaves: 127, verbose: -1,seed: 42, n_jobs: -1,feature_fraction: 0.8, bagging_fraction: 0.9, bagging_freq: 4} The MSE (mean square error) of the model built by the LightGBM algorithm is 0.09825647850032115, the RMSE (root mean square error) is 0.3134588944348543, the MAE (mean absolute error) is 0.24379498997769858, and the R-squared (goodness of fit) is 0.9051377686179084.

[0046] S40302. Use the XGBoost library in Python and call the XGBRegressor method to build an XGBoost regression model. The model parameters are as follows: params_XGBoost{learning_rate, booster, objective, max_leaves, verbosity, seed, nthread, colsample_bytree, subsample, early_stopping_rounds, eval_metric} Among them, learning_rate represents the learning rate, booster represents the boosting method, objective represents the loss function, max_leaves represents the number of leaf nodes of each decision tree, verbosity is used to control the detail of the output information, seed represents the random seed, nthread represents the number of threads for parallel computing, colsample_bytree represents the proportion of features randomly selected for each decision tree, subsample represents the proportion of samples randomly selected at each iteration, and eval_metric represents the evaluation indicator.

[0047] In this embodiment, a set of initial parameters of the XGBoost algorithm is defined: params_XGBoost={learning_rate: 0.02, booster: gbtree, objective: reg:squarederror, max_leaves: 127, verbosity: 1, seed: 42, nthread: -1, colsample_bytree: 0.6, subsample: 0.7, early_stopping_rounds: 100, eval_metric: rmse} The model built by the XGBoost algorithm has an MSE (mean square error) of 0.14852525526041624, a RMSE (root mean square error) of 0.385389744622786, a MAE (mean absolute error) of 0.3201391679162975, and an R-squared (goodness of fit) of 0.8566055150190248.

[0048] S40303. Use the sklearn library in Python to call the RandomForestRegressor method to build a random forest regression model. The model parameters are as follows: params_RandomForest{n_estimators, max_depth, max_features, min_samples_leaf, random_state} Among them, n_estimators represents the number of decision trees, max_depth represents the maximum depth of the decision tree, max_features represents the number of features that need to be considered for optimal segmentation, min_samples_leaf represents the minimum number of samples at the leaf node, and random_state represents the random seed.

[0049] In this embodiment, a set of initial parameters of the random forest algorithm is defined: params_RandomForest={n_estimators=1000, max_depth=8, max_features=10, min_samples_leaf=7, random_state=42} The model built by the random forest algorithm has an MSE (mean square error) of 0.03637238490928531, a RMSE (root mean square error) of 0.19071545534980983, a MAE (mean absolute error) of 0.1369673173794692, and an R-squared (goodness of fit) of 0.8420415687901608.

[0050] S40304. Use the NGBoost library in Python to call the NGBRegressor method to build an NGBoost regression model. The model parameters are as follows: params_ngb{learning_rate, n_estimator, verbose, random_state, natural_gradient} Among them, learning_rate represents the learning rate, n_estimator represents the number of decision trees, verbose is used to control the detail of the output information, random_state represents the random seed, and natural_gradient represents the use of the natural gradient method.

[0051] In this embodiment, a set of initial parameters of the NGBoost algorithm is defined: params_ngb = {learning_rate: 0.02, n_estimators: 1000, verbose: False, random_state: 42, natural_gradient: True} The MSE (mean square error) of the model built by the NGBoost algorithm is 0.09544001627059343, the RMSE (root mean square error) is 0.3089336761678685, the MAE (mean absolute error) is 0.2235466115673266, and the R-squared (goodness of fit) is 0.9078569368172296.

[0052] S40305. Use the catboost library in Python to call the CatboostRegressor method to build a CatBoost regression model. The model parameters are as follows: params_Catboost{learning_rate, iterations, depth, eval_metric,random_seed, verbose} Among them, learning_rate represents the learning rate, iterations represents the number of decision trees, depth represents the depth of the decision tree, eval_metric represents the evaluation metric, random_seed represents the random seed, and verbose is used to control the detail of the output information.

[0053] In this embodiment, a set of CatBoost algorithm initial parameters are defined: params_Catboost={learning_rate: 0.02, iterations: 1000, depth: 8, eval_metric: RMSE, random_seed: 42, verbose: 500} The MSE (mean square error) of the model built by the CatBoost algorithm is 0.0302320406812082, the RMSE (root mean square error) is 0.17387363423247412, the MAE (mean absolute error) is 0.09504162218148243, and the R-squared (goodness of fit) is 0.9508123181084268.

[0054] S40306. The R-squared (goodness of fit) and RMSE (root mean square error) of the models constructed by the above five machine learning algorithms are shown in Table 2.

[0055]

[0056] According to the R-squared (goodness of fit) and RMSE (root mean square error) of the above five machine learning algorithms, the CatBoost algorithm with the best R-squared (goodness of fit) and RMSE (root mean square error) was selected to construct the community health service center location selection model.

[0057] S404. Use the CatBoost model constructed in step S40305 as the community health service center site selection initialization model, define the parameter grid of the grid search, tune the learning_rate, depth, and iterations parameters in the CatBoost model respectively, and use the 5-fold cross-validation method to find the parameter value corresponding to the model with the highest score in the cross-validation to form the best parameter combination, and obtain the preliminary model of the community health service center site selection under the training of the best parameter combination.

[0058] Further, step S404 specifically includes: S40401. Define the parameter grid for grid search as follows: param_grid{learning_rate:[0.01,0.02,0.05,0.1],depth:[4,6,8,10,12],iterations:[800,1000,1200,1400,1600,2000]} S40402. Use the sklearn library in Python to call the GridSearchCV method in the model_selection module to perform grid search and K-fold cross validation. The parameters of the GridSearchCV method are as follows: GridSearchCV(estimator=model, param_grid=param_grid, scoring=neg_root_mean_squared_error, cv=5, n_jobs=-1, verbose=1) Among them, scoring=neg_root_mean_squared_error means that the negative value of RMSE (root mean square error) is used as the evaluation indicator, and cv=5 means 5-fold cross validation.

[0059] S40403. Substitute the eigenvalue dataset X and the target value dataset y in the training set into the CatBoost regression model for training. Use the 5-fold cross-validation method to divide the eigenvalue dataset X and the target value dataset in the training set into 5 subsets. In each validation process, 4 subsets are used to train the model, and the remaining 1 subset is used to test the model. The validation process is repeated 5 times, and a different subset is selected as the test set each time. Each parameter combination in the parameter grid is evaluated by using the 5-fold cross-validation method, and the optimal parameter combination is output, and a preliminary model for the site selection of community health service centers under the optimal parameter combination training is obtained.

[0060] In this embodiment, the optimal parameter combination is {depth:10, iterations:2000, learning_rate:0.02}. Under this parameter combination, the best RMSE in 5-fold cross validation is 0.08446205914955598.

[0061] S50. Evaluate the preliminary model for site selection of community health service centers, calculate the contribution of relevant indicators in the preliminary model for site selection of community health service centers, and use Monte Carlo simulation to optimize feature selection and combination to obtain the best feature combination of the preliminary model for site selection of community health service centers.

[0062] In this embodiment, step S50 specifically includes: S501, call the metrics module in the sklearn library, substitute the test set into the preliminary model for site selection of community health service centers obtained in S40 for model evaluation, and evaluate the R-squared (goodness of fit) of the preliminary model as a comparison benchmark value for Monte Carlo simulation optimization feature selection and combination.

[0063] In this embodiment, the R-squared (goodness of fit) of the preliminary model for site selection of community health service centers is 0.9606862717050595.

[0064] S502, call shap (model prediction result interpretation method based on Shapley value theory) to calculate the shap value of each feature in the test set and interpret the contribution of each feature in the community health service center site selection model. The contribution (shap value) of each feature in the preliminary model of community health service center site selection is as follows: Figure 2 shown.

[0065] S503. Based on the contribution (shap value) of each feature, call the abs module in the numpy library to calculate the average absolute shap value of each feature, obtain the feature label, call the argsort module in the numpy library to sort the average absolute shap value of each feature and correspond it to the feature label, re-sort the feature value data set X according to the model feature contribution, and obtain a new feature value data set X'.

[0066] S504, calling the cross_val_score of the model_selection module in the sklearn library, performing Monte Carlo simulation and cross validation on the new feature value dataset X' and the target value dataset y, setting the number of Monte Carlo simulations to 20, and defining the cv_scores variable to record the cross validation scores.

[0067] S505. Use the best parameter combination obtained in step S405 to set the model for Monte Carlo simulation. In each simulation, the data set is re-divided to achieve the purpose of random sampling, effectively deal with the uncertainty in the data, and in each simulation, the number of features is gradually increased from large to small according to the feature contribution and cross-validation is performed. The number of cross-validation iterations cv is set to 5, the evaluation effect scoring is 'r2' (goodness of fit), and the average cross-validation score is calculated.

[0068] S506, draw a cross-validation score line graph of the model under different feature combinations, such as Figure 3 As shown, the number and names of the features that performed best in the Monte Carlo simulation were obtained.

[0069] In this embodiment, the number of best-performing features is 17, which are accessibility to general hospitals, distance to walkable roads, accessibility to residential areas, number of residential communities, plot ratio, public transportation convenience, proportion of elderly population, public transportation convenience, number of general hospitals, proportion of resident population to working population, population density, night light index, PM2.5 concentration, proportion of children and adolescents, surrounding functional density, surrounding functional diversity, and NDVI (normalized vegetation index). The cv_score (measured by goodness of fit (R-squared)) of the best feature combination is 0.9677246949050842.

[0070] S60. Based on the optimal feature combination, the optimal machine learning algorithm is used to construct a community health service center site selection model, parameters of the community health service center site selection model are optimized, and the optimal parameter combination is output through K-fold cross validation to obtain the community health service center site selection model under the optimal feature combination and the optimal parameter combination training.

[0071] In this embodiment, step S60 specifically includes: S601. Use the best feature combination obtained in step S50 to screen the feature values ​​in the fourth sample data set, and screen out 17 features in the combination as a new feature value data set X_2. The target value data set is the target value data including the community health service center features, represented by the letter y.

[0072] S602. Use the preliminary model parameter combination of the community health service center as the initialization CatBoost model, define the parameter grid of the grid search, tune the learning_rate, depth, and iterations parameters in the CatBoost model respectively, use the 5-fold cross-validation method to find the parameter value corresponding to the model with the highest score in the cross-validation, form the best parameter combination, and obtain the community health service center site selection model under the training of the best parameter combination.

[0073] Furthermore, step S602 specifically includes: S60201. Define the parameter grid for grid search as follows: param_grid{learning_rate:[0.01,0.02,0.05,0.1],depth:[4,6,8,10,12],iterations:[800,1000,1200,1400,1600,2000]} S60202. Use the sklearn library in Python to call the GridSearchCV method in the model_selection module to perform grid search and K-fold cross validation. The parameters of the GridSearchCV method are as follows: GridSearchCV(estimator=model, param_grid=param_grid, scoring=neg_root_mean_squared_error, cv=5, n_jobs=-1, verbose=1) Among them, scoring=neg_root_mean_squared_error means that the negative value of RMSE (root mean square error) is used as the evaluation indicator, and cv=5 means 5-fold cross validation.

[0074] S60203. Substitute the eigenvalue dataset X_2 and target value dataset y in the training set into the CatBoost regression model for training, use the 5-fold cross-validation method to evaluate each parameter combination in the parameter grid, output the optimal parameter combination, and obtain the community health service center site selection model under the optimal feature combination and the optimal parameter combination training.

[0075] In this embodiment, the optimal parameter combination is {depth:10, iterations:2000, learning_rate:0.05}. Under this parameter combination, the best RMSE in 5-fold cross validation is 0.08292214473600664.

[0076] S70, evaluating the obtained community health service center site selection model, and calculating the contribution of relevant indicators in the community health service center site selection model; In this embodiment, step S70 specifically includes: S701, calling the metrics module in the sklearn library, substituting the test set into the community health service center site selection model obtained in step S60 for model evaluation, and calculating the output evaluation indicators as follows: MSE (mean square error), RMSE (root mean square error), MAE (mean absolute error), R-squared (goodness of fit) In this embodiment, the MSE (mean square error) of the community health service center site selection model is 0.004625254626564098, the RMSE (root mean square error) is 0.07800922456964274, the MAE (mean absolute error) is 0.02818615846819666, and the R-squared (goodness of fit) is 0.9699133885066853, which proves the accuracy of the CatBoost model in predicting the site selection of community health service centers.

[0077] S702: Call shap (a model prediction result interpretation method based on Shapley value theory) to calculate the shap value of each feature in the test set and interpret the contribution of each feature of the preliminary model.

[0078] The schematic diagram of the contribution of each feature of the community health service center location model is as follows: Figure 4 As shown in the figure, we can see that the ten characteristics of accessibility to general hospitals, distance to walkable roads, accessibility to residential areas, number of residential communities, proportion of elderly population, floor area ratio, subway convenience, bus convenience, population density, and number of general hospitals have the greatest contribution.

[0079] S80. Export and save the community health service center site selection model, input the eigenvalue data set in the third sample data set in the target area into the community health service center site selection model, obtain the community health service center site selection candidate points, determine the number of community health service centers to be added according to the population size of the target area, and use the maximum coverage model to determine the set of community health service center site selection points that can cover the most demand points among the community health service center site selection candidate points.

[0080] In this embodiment, step S80 specifically includes: S801. Call the joblib library to save the community health service center site selection model and export it to a pkl format file.

[0081] S802: Input the characteristic value data set in the third sample data set in the target area into the model to obtain the predicted target value result. Connect the predicted target value result table to the building unit geometric center point data in ArcGIS and visualize it.

[0082] In this embodiment, the predicted target value result range is [-0.176866, 1.003224]. The building units with predicted target values ​​> 0.2 are selected as the site selection candidates for the community health service center.

[0083] S803. According to the permanent population data of Yuexiu District, Guangzhou, the permanent population of Yuexiu District is 960,000. Based on the calculation of setting up one community health service center for every 30,000 people, Yuexiu District needs to set up 32 community health service centers. According to the 15-minute walking distance, 900m is set as the service radius of the community health service center. Under the condition that the number and service radius of community health service centers are known, the maximum coverage model is used to determine the set of community health service center locations that can cover the most demand points.

[0084] The maximum coverage model is used to determine a set of community health service center site selection points that can cover the most demand points among the community health service center site selection candidate points. The maximum coverage model is expressed as: , The constraint formula of the maximum coverage model is: , , , In the formula, max S is the sum of the covered demand points, i As demand points, residential areas are taken as demand points. j Select alternative sites for community health service centers. Whether the demand point is covered, To cover the demand points iThe set of candidate points, m is the number of alternative points, M Limit the number of locations for community health service centers. Whether to use the alternative point as the site selection point.

[0085] The purpose of the maximum coverage model is to maximize the coverage of demand points and max S. Use the "Location Matching" tool in the ArcGIS "Network Analysis" toolbar, take the community health service center site selection points as facility points, and take the existing community health service centers as required options, take the residential points in the target area as demand points, and take the road data as line data. Set the impedance to 900m and the number of facility points to 32. Run the solution to generate a set of community health service center site selection points that can cover the most demand points, such as Figure 5 shown.

[0086] By using the site selection method for community health service centers in a high-density urban environment in the embodiment, the site selection accuracy of the health service center can be effectively improved, thereby improving the coverage rate of health services.

[0087] Example 2 As another aspect of the embodiment of the present disclosure, a community health service center site selection system 100 in a high-density urban environment is also provided. Figure 6 As shown, including: Data acquisition module 1, obtains network open source data, population grid data, mobile phone signaling data, point of interest / surface of interest data, night light data, building unit data, normalized vegetation index data, PM2.5 concentration data and urban road data of the target area in a high-density urban environment to form a first sample data set; A second sample set acquisition module 2, which sorts and corrects the data in the first sample set to obtain a second sample set; The relevant indicator processing module 3 calculates the relevant indicators for the site selection of the community health service center according to the second sample data set, and normalizes the relevant indicators to obtain a third sample data set, wherein the relevant indicators include one or more of the following indicators: population density index, population structure index, general hospital accessibility index, public transportation convenience index, subway convenience index, residential area accessibility index, building density index, floor area ratio index, night light index, normalized vegetation index, surrounding function density index, surrounding function diversity index, general hospital index and residential area index; A preliminary model building module 4 for site selection of community health service centers is used to obtain a fourth sample data set according to the target value of the sample in the third sample data set, and based on the fourth sample data set, an optimal machine learning algorithm is used to build a preliminary model for site selection of community health service centers, and grid search is used to perform parameter tuning, and the optimal parameter combination is output through K-fold cross validation to obtain a preliminary model for site selection of community health service centers; Optimizing feature selection and combination module 5, evaluating the preliminary model of community health service center site selection, calculating the contribution of relevant indicators in the preliminary model of community health service center site selection, optimizing feature selection and combination by Monte Carlo simulation, and obtaining the best feature combination of the preliminary model of community health service center site selection; The community health service center site selection model construction module 6 constructs the community health service center site selection model based on the optimal feature combination and adopts the optimal machine learning algorithm, optimizes the parameters of the community health service center site selection model, and outputs the optimal parameter combination through K-fold cross validation to obtain the community health service center site selection model under the optimal feature combination and the optimal parameter combination training; Model evaluation module 7, evaluates the obtained community health service center site selection model, and calculates the contribution of relevant indicators in the community health service center site selection model; The site selection point set determination module 8 exports and saves the community health service center site selection model, inputs the characteristic value data set in the third sample data set in the target area into the community health service center site selection model, obtains the community health service center site selection candidate points, determines the number of community health service centers to be added according to the population size of the target area, and uses the maximum coverage model to determine the community health service center site selection point set that can cover the most demand points among the community health service center site selection candidate points; In the absence of any contradiction, the above modules in the system of the embodiment of the present disclosure can implement any implementation of the above method.

[0088] Based on the description of the above embodiments, it can be seen that the embodiments of the present disclosure can achieve the following technical effects: 1) This paper proposes a method for site selection of community health service centers in a high-density urban environment, obtains multi-source data in five dimensions of population characteristics, accessibility, built environment, natural environment, and urban functions in the target area under a high-density urban built environment, and obtains a fourth sample data set by performing multiple processing on the data, so as to construct a systematic, objective and comprehensive feature indicator system for the site selection model of community health service centers based on machine learning.

[0089] 2) This paper uses five machine learning regression algorithms, LightGBM, XGBoost, CatBoost, NGBoost and random forest, to build and test the model, and obtains the optimal machine learning algorithm with the best R-squared (goodness of fit) and RMSE (root mean square error), and uses network search to tune parameters. Based on the contribution of each feature obtained by calculating the shap value, Monte Carlo simulation is used to optimize feature selection and combination, and then a community health service center site selection model is obtained. This process can effectively improve the accuracy of community health service center site selection.

[0090] 3) The present invention inputs the characteristic value data set in the third sample data set in the target area into the community health service center site selection model to obtain alternative site selection points for the community health service center, and uses the maximum coverage model to determine a set of community health service center site selection points that can cover the most demand points among the alternative site selection points for the community health service center. This process can ensure the coverage rate of community health services in the target area.

[0091] The embodiment of the present disclosure also proposes an electronic device, comprising: a processor; a memory for storing instructions executable by the processor; wherein the processor is configured to implement the above-mentioned method for selecting a site for a community health service center in a high-density urban environment. The electronic device can be provided as a terminal, a server, or other forms of equipment.

[0092] The embodiment of the present disclosure also provides a computer-readable storage medium on which computer program instructions are stored, and when the computer program instructions are executed by a processor, the method for selecting a site for a community health service center in a high-density urban environment is implemented. The computer-readable storage medium may be a non-volatile computer-readable storage medium.

[0093] Those skilled in the art will understand that in the specific implementation method and system for site selection of community health service centers in high-density urban environments, the order in which the steps are written does not mean a strict order of execution and does not constitute any limitation on the implementation process. The specific order of execution of each step should be determined by its function and possible internal logic.

[0094] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of the module, program segment or instruction includes one or more executable instructions for realizing the specified logical function. In some alternative implementations, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs the specified function or action, or can be implemented with a combination of special hardware and computer instructions.

[0095] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. The site selection method for community health service centers in high-density urban environments is characterized by: The steps include: S10, obtaining network open source data, population grid data, mobile phone signaling data, point of interest / surface of interest data, night light data, building unit data, normalized difference vegetation index data, PM2.5 concentration data and urban road data of the target area in a high-density urban environment to form a first sample data set; S20, sorting and correcting the data in the first sample data set to obtain a second sample data set; S30. Calculate the site selection related indicators of the community health service center according to the second sample data set, and normalize the related indicators to obtain a third sample data set, wherein the related indicators include one or more of the following indicators: population density index, population structure index, general hospital accessibility index, public transportation convenience index, subway convenience index, residential area accessibility index, building density index, floor area ratio index, night light index, normalized vegetation index, surrounding function density index, surrounding function diversity index, general hospital index and residential area index; S40, sampling according to the target value of the samples in the third sample data set to obtain a fourth sample data set, based on the fourth sample data set, using the optimal machine learning algorithm to construct a preliminary model for the site selection of community health service centers, using grid search to perform parameter tuning, and outputting the best parameter combination through K-fold cross validation to obtain a preliminary model for the site selection of community health service centers; S50, evaluating the preliminary model for site selection of community health service centers, calculating the contribution of relevant indicators in the preliminary model for site selection of community health service centers, optimizing feature selection and combination by Monte Carlo simulation, and obtaining the best feature combination for the preliminary model for site selection of community health service centers; S60, based on the optimal feature combination, using the optimal machine learning algorithm to construct a community health service center site selection model, perform parameter tuning on the community health service center site selection model, and output the optimal parameter combination through K-fold cross validation to obtain the community health service center site selection model under the optimal feature combination and the optimal parameter combination training; S70, evaluating the obtained community health service center site selection model, and calculating the contribution of relevant indicators in the community health service center site selection model; S80. Export and save the community health service center site selection model, input the eigenvalue data set in the third sample data set in the target area into the community health service center site selection model, obtain the community health service center site selection candidate points, determine the number of community health service centers to be added according to the population size of the target area, and use the maximum coverage model to determine the set of community health service center site selection points that can cover the most demand points among the community health service center site selection candidate points.

2. The method according to claim 1, characterized in that Arranging and correcting the data in the first sample data set includes: Input a first sample data set, and perform random value assignment processing on the mobile phone signaling data in the first sample data set; use a line chart to draw the data of each feature in the first sample data set to determine whether there are outliers in the data of each feature that deviate far from the normal range, and delete the outliers; deduplicate the same data in the first sample set; and check and correct the urban road data in the first sample data set.

3. The method according to claim 2, characterized in that The community health service center site selection related indicators calculated based on the second sample data set specifically include: Convert building unit surface data into geometric center point data of the building unit surface; According to the population grid data, the population in the grid is counted to obtain the population density index; According to the mobile phone signaling data, the user age structure distribution is counted to obtain the population structure index; Based on the data of points of interest / areas of interest, calculate the accessibility index of general hospitals, bus convenience index, subway convenience index and residential area accessibility index; Calculate building density index and floor area ratio index based on building unit data; Calculate the night light index indicator based on the night light data; According to the normalized vegetation index data, a grid layer containing normalized vegetation index information is obtained, and the normalized vegetation index data index is calculated; Based on the data of points of interest / areas of interest, the surrounding functional density index, surrounding functional diversity index, general hospital index and residential area index are calculated.

4. The method according to any one of claims 1 or 3, characterized in that: The relevant indicators are normalized to obtain a third sample data set, which is expressed as: ， In the formula, X is the original data, For normalized data, and are the maximum and minimum values ​​in the original data respectively.

5. The method according to claim 4, characterized in that The method of sampling the target values ​​of the samples in the third sample data set to obtain a fourth sample data set, and constructing a preliminary model for site selection of community health service centers using an optimal machine learning algorithm based on the fourth sample data set, including: Importing the third sample data set into the table, oversampling the samples with a target value of 1 in the third sample data set, undersampling the samples with a target value of 0, sorting the samples with target values ​​of 1 and 0, and generating a fourth sample data set; Use 80% of the fourth sample data set as a training set and 20% as a test set; A variety of machine learning algorithms were used to test the model construction, and the optimal machine learning algorithm with the best goodness of fit and root mean square error was determined based on the test results. The optimal machine learning algorithm was selected to construct a preliminary model for the site selection of community health service centers.

6. The method according to claim 1, characterized in that The maximum coverage model is used to determine a set of community health service center site selection points that can cover the most demand points among the community health service center site selection candidate points. The maximum coverage model is expressed as: ， The constraint formula of the maximum coverage model is: ， ， ， In the formula, max S is the sum of the covered demand points, i As demand points, residential areas are taken as demand points. j Select alternative sites for community health service centers. Is the demand point covered? To cover the demand points i The set of candidate points, m is the number of alternative points, M Limit the number of locations for community health service centers. Whether to use the alternative point as the site selection point.

7. The site selection system for community health service centers in high-density urban environments is characterized by: include: The data acquisition module acquires network open source data, population grid data, mobile phone signaling data, point of interest / surface of interest data, night light data, building unit data, normalized vegetation index data, PM2.5 concentration data and urban road data of the target area in a high-density urban environment to form a first sample data set; A second sample set acquisition module, which sorts and corrects the data in the first sample set to obtain a second sample set; The relevant indicator processing module calculates the community health service center site selection related indicators according to the second sample data set, and normalizes the related indicators to obtain a third sample data set, wherein the related indicators include one or more of the following indicators: population density index, population structure index, general hospital accessibility index, public transportation convenience index, subway convenience index, residential area accessibility index, building density index, floor area ratio index, night light index, normalized vegetation index, surrounding function density index, surrounding function diversity index, general hospital index and residential area index; A preliminary model building module for site selection of community health service centers, sampling a fourth sample data set according to the target value of the sample in the third sample data set, constructing a preliminary model for site selection of community health service centers based on the fourth sample data set using an optimal machine learning algorithm, using grid search to perform parameter tuning, and outputting the best parameter combination through K-fold cross validation to obtain a preliminary model for site selection of community health service centers; Optimizing the feature selection and combination module, evaluating the preliminary model for site selection of community health service centers, and calculating the contribution of relevant indicators in the preliminary model for site selection of community health service centers, optimizing the feature selection and combination by Monte Carlo simulation, and obtaining the best feature combination for the preliminary model for site selection of community health service centers; A community health service center site selection model construction module, based on the optimal feature combination, uses the optimal machine learning algorithm to construct a community health service center site selection model, performs parameter tuning on the community health service center site selection model, and outputs the optimal parameter combination through K-fold cross validation to obtain a community health service center site selection model under the optimal feature combination and the optimal parameter combination training; A model evaluation module is used to evaluate the obtained community health service center site selection model and calculate the contribution of relevant indicators in the community health service center site selection model; The site selection point set determination module exports and saves the community health service center site selection model, inputs the characteristic value data set in the third sample data set in the target area into the community health service center site selection model, obtains the community health service center site selection candidate points, determines the number of community health service centers to be added according to the population size of the target area, and uses the maximum coverage model to determine the community health service center site selection point set that can cover the most demand points among the community health service center site selection candidate points.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for selecting a site for a community health service center in a high-density urban environment as described in any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method for selecting a site for a community health service center in a high-density urban environment as described in any one of claims 1 to 6 is implemented.

Images