Pension service institution supplementary site selection method based on multi-source data and machine learning

Abstract

The invention discloses a pension service institution supplementary site selection method based on multi-source data and machine learning, and relates to the technical field of pension service institution site selection. According to the method, site selection is performed on the multi-source data of the pension service through machine learning, so that the labor cost and the dependence on manual analysis are reduced, and the automation degree of site selection is improved; according to the method, multiple machine learning models are optimized, the optimal machine learning model is selected for site selection, the application range is wide, changing requirements can be dynamically responded, it is ensured that service coverage of the old-age care facility is matched with actual requirements, and the method adapts to dynamic changes of city and population structures.

Description

Technical Field

[0001] The present invention relates to the technical field of site selection for elderly care service institutions, and in particular to a supplementary site selection method for elderly care service institutions based on multi-source data and machine learning. Background Art

[0002] The existing methods for site selection of elderly care service institutions have the following technical problems: First, the cost is high. Traditional site selection relies on manual research and planning, which involves a lot of manpower and time costs. Especially when dealing with multiple factors (such as population density, traffic accessibility, medical resources, etc.), the existing technology lacks efficient data analysis tools, which increases planning costs. Second, the site selection efficiency is low. The existing methods only consider limited factors, and it is difficult to dynamically adjust the site selection strategy in real time. Especially in the context of rapid urban development, site selection decisions often lag behind actual needs. In addition, the structure of the site selection model is complex. The existing site selection methods usually rely on experience-based decisions. The model construction is complex and difficult to automate. It is impossible to effectively integrate multi-source heterogeneous data, such as demographic, geographic and traffic data. Finally, the existing technologies are mostly static analysis, and it is difficult to dynamically adjust the site selection strategy according to changes in urban development and population mobility. Summary of the invention

[0003] The purpose of the present invention is to provide a method for supplementing the site selection of elderly care service institutions based on multi-source data and machine learning to improve the above-mentioned technical problems.

[0004] In order to achieve the above-mentioned object of the invention, the embodiment of the present invention provides the following technical solutions:

[0005] A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning, comprising:

[0006] S1. Real-time acquisition of multi-source data on elderly care services in the area to be selected and pre-processing to obtain pre-processed multi-source data on elderly care services;

[0007] S2, constructing a grid dataset of elderly care services and processing it using feature engineering to obtain a supplementary feature dataset of elderly care services;

[0008] S3. Construct a machine learning model for the location selection of elderly care service institutions, and optimize the machine learning model for the location selection of elderly care service institutions using K-fold cross validation and grid random search algorithms to obtain the optimal machine learning model for the location selection of elderly care service institutions;

[0009] S4, inputting the supplementary feature data set of elderly care services into the optimal elderly care service institution location selection machine learning model to obtain the supplementary initial location selection result of the elderly care service institution;

[0010] S5. Determine the prediction accuracy probability based on the additional initial site selection result of the elderly care service institution and the elderly care service multi-source data;

[0011] S6, judging whether the prediction accuracy probability is less than a threshold; if so, returning to S3; otherwise, entering S7;

[0012] S7. Based on the demand for elderly care services to be sited, the initial site selection results for the elderly care service institution are divided to obtain priority site selection areas and common site selection areas, thereby completing the supplementary site selection for the elderly care service institution.

[0013] Furthermore, the multi-source data of elderly care services in the area to be selected includes location information of existing elderly care service institutions, multi-dimensional POI data, multi-dimensional AOI data, economic data, and population data;

[0014] The location information of the existing elderly care service institution is the latitude and longitude of the existing elderly care service institution;

[0015] The multi-dimensional POI data includes catering POI data, shopping POI data, transportation POI data, financial POI data, medical POI data, education POI data, life service POI data, accommodation POI data, cultural and sports entertainment POI data, government and social group POI data, scenic spot POI data, enterprise and institution POI data and public facility POI data;

[0016] The multivariate AOI data includes building AOI data, cultivated land AOI data, green space AOI data, water body AOI data, and road AOI data;

[0017] The economic data includes house price grid data and per capita GDP grid data;

[0018] The population data includes data on the elderly population and data on the permanent population;

[0019] The preprocessing includes data cleaning, formatting, missing value interpolation, outlier processing and data coding.

[0020] Furthermore, the elderly care service raster dataset is constructed and processed by feature engineering to obtain an elderly care service feature dataset, including:

[0021] S2-1, determining the grid size based on the regional characteristics of the area to be selected;

[0022] S2-2, based on the grid size, the pre-processed elderly care service multi-source data is processed using the fishing net analysis of ArcGIS software, and the spatial attributes of the pre-processed elderly care service multi-source data are given through the spatial connection function of ArcGIS software to construct the elderly care service raster dataset;

[0023] S2-3. Use feature engineering to extract features from the elderly care service raster dataset to obtain the elderly care service supplementary feature dataset.

[0024] Furthermore, the machine learning model for site selection of elderly care service institutions includes a decision tree model, a random forest model, an extreme gradient boosting model, a gradient boosting decision tree model, a support vector machine model and a K nearest neighbor model.

[0025] Furthermore, the K-fold cross validation and grid random search algorithm optimizes the elderly care service institution location selection machine learning model to obtain the optimal elderly care service institution location selection machine learning model, including:

[0026] S3-1, obtain multi-source data of historical elderly care services, and process them using the same method as S1 and S2 to obtain a supplementary feature data set of historical elderly care services;

[0027] S3-2, dividing the historical elderly care service supplementary feature data set into K equal subsets;

[0028] S3-3, randomly select a subset as the validation set, and the remaining K-1 subsets as the training set;

[0029] S3-4, randomly select a subset from the training set;

[0030] S3-5, respectively inputting the training set selected in S3-4 into the decision tree model, the random forest model, the extreme gradient boosting model, the gradient boosting decision tree model, the support vector machine model and the K nearest neighbor model for training, to obtain the corresponding trained decision tree model, random forest model, extreme gradient boosting model, gradient boosting decision tree model, support vector machine model and K nearest neighbor model;

[0031] S3-6, calculating the performance indicators corresponding to the trained decision tree model, random forest model, extreme gradient boosting model, gradient boosting decision tree model, support vector machine model and K nearest neighbor model;

[0032] S3-7, repeat S3-3 to S3-6 until K performance indicators of each machine learning model are obtained;

[0033] S3-8. Calculate the first average performance index corresponding to each machine learning model, and sort them in descending order, and select the machine learning models corresponding to the top two or top three first average performance indicators as the location selection potential models for elderly care service institutions;

[0034] S3-9. Use a grid random search algorithm to optimize the parameters of the elderly care service institution site selection potential model to obtain the optimal elderly care service institution site selection machine learning model.

[0035] Furthermore, the grid random search algorithm is used to optimize the parameters of the elderly care service institution site selection potential model to obtain the optimal elderly care service institution site selection machine learning model, including:

[0036] S3-9-1. Randomly select one of the above-mentioned elderly care service institution location potential models;

[0037] S3-9-2. Define the hyperparameter space of the selected elderly care service institution location potential model, and set the parameters to be optimized of the selected elderly care service institution location potential model and their corresponding value ranges;

[0038] S3-9-3. Selecting the optimal hyperparameter combination of the selected elderly care service institution location potential model using a genetic algorithm;

[0039] S3-9-4, repeat S3-9-1 to S3-9-3 until the optimal hyperparameter combination corresponding to each of the elderly care service institution site selection potential models is obtained;

[0040] S3-9-5. Calculate the second average performance index of each of the elderly care service institution site selection potential models;

[0041] S3-9-6. Based on the second average performance indicator, determine the optimal elderly care service institution location selection machine learning model.

[0042] Furthermore, the method of selecting the optimal hyperparameter combination of the selected elderly care service institution location potential model by using a genetic algorithm includes:

[0043] Randomly generate a population of hyperparameter combinations of the selected elderly care service institution location potential model;

[0044] Applying each hyperparameter combination in the hyperparameter combination population to the selected elderly care service institution location potential model and performing training;

[0045] Using a cross-validation method to evaluate the performance indicators of the trained elderly care service institution location potential model corresponding to each of the hyperparameter combinations, and obtain the corresponding fitness;

[0046] Arrange the fitness values ​​in descending order, and select the hyperparameter combination with the top 10% fitness value as the parent generation;

[0047] Calculate the SHAP value corresponding to each feature in the historical elderly care service supplementary feature data set; select core features based on each of the SHAP values ​​and retain them;

[0048] Using an adaptive crossover strategy to exchange the core features, and generating a new generation of offspring based on the parent generation;

[0049] Perform mutation exploration on the offspring of the new generation to obtain updated offspring, i.e., new hyperparameter combinations;

[0050] Repeat the above process until the preset number of iterations M is reached, and M new hyperparameter combinations are obtained; calculate the fitness of the new hyperparameter combination, select the new hyperparameter combination with the largest fitness and apply it to the selected elderly care service institution site selection potential model.

[0051] Furthermore, the performance indicators include accuracy, precision, recall, and F1 score; the first average performance indicator and the second average performance indicator are both average values ​​of accuracy, precision, recall, and F1 score, that is:

[0052]

[0053] in, Indicates The average performance index, Indicates the total number of indicators, Indicates The accuracy value, Indicates The precision value, Indicates The recall value, Indicates F1 score value; The value of is 1 or 2. When it is 1, it means calculating the first average performance index. , , , Respectively represent the accuracy value, precision value, recall value, and F1 score value of the machine learning model; when When it is 2, it means calculating the second average performance index. , , , They respectively represent the accuracy value, precision value, recall value, and F1 score value of the location selection potential model of the elderly care service institution.

[0054] Furthermore, determining the prediction accuracy probability based on the additional initial site selection result of the elderly care service institution and the elderly care service multi-source data includes:

[0055] S5-1, using ArcGIS software to link the supplementary initial site selection results of the elderly care service institution and the supplementary feature data set of the elderly care service in S2-3 to form a visualized site selection prediction raster data;

[0056] S5-2, determining the number of overlaps between the additional initial site selection results of the elderly care service institution and the elderly care service multi-source data;

[0057] S5-3, according to the formula:

[0058]

[0059] Get the probability of accurate prediction ;in, represents the number of grid overlaps between the additional initial site selection result of the elderly care service institution in the visualized site selection prediction grid data and the existing elderly care service structure in the to-be-selected site selection area, The number of grids representing the existing elderly care service structures in the area to be selected.

[0060] Furthermore, based on the demand for elderly care services to be sited, the initial site selection results of the elderly care service institution are divided to obtain priority site selection areas and common site selection areas, including:

[0061] Determine and delete the grids of the existing elderly care service institutions in the additional initial site selection result of the elderly care service institutions to obtain an updated additional initial site selection result of the elderly care service institutions;

[0062] Determine the demand for elderly care services at the site to be selected, that is, divide the elderly area and non-elderly area according to the elderly population data;

[0063] Spatial fitting is performed on the elderly areas, non-elderly areas and the updated initial site selection results of the elderly care service institutions to obtain the priority site selection area and the general site selection area.

[0064] The beneficial effects of the present invention are:

[0065] The present invention uses machine learning to select sites for multi-source data of elderly care services, thereby reducing labor costs and dependence on manual analysis and improving the degree of automation in site selection. It optimizes multiple machine learning models and selects the optimal machine learning model for site selection. It has a wide range of applications and can dynamically respond to changing needs, ensuring that the coverage of elderly care facility services matches actual needs and adapts to dynamic changes in urban and population structures. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0067] Figure 1 is a flow chart of a method in an embodiment of the present invention;

[0068] Figure 2 The diagram is a schematic diagram of the division process of supplementing the initial site selection results for the elderly care service agency in an embodiment of the present invention. DETAILED DESCRIPTION

[0069] 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 only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0070] See also Figure 1 This embodiment provides a method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning, which includes:

[0071] S1. Real-time acquisition of multi-source data on elderly care services in the area to be selected and pre-processing to obtain pre-processed multi-source data on elderly care services;

[0072] The multi-source data of elderly care services in the area to be selected includes location information of existing elderly care service institutions, multi-dimensional POI data, multi-dimensional AOI data, economic data, and population data;

[0073] The location information of the existing elderly care service institution is the latitude and longitude of the existing elderly care service institution;

[0074] The multivariate POI data includes catering POI data (such as restaurants, coffee shops, fast food restaurants, bars, etc.), shopping POI data (such as supermarkets, department stores, convenience stores, shopping malls, etc.), transportation POI data (such as bus stops, subway stations, railway stations, airports, parking lots, etc.), financial POI data (such as banks, ATMs, self-service banks, securities companies, etc.), medical POI data (such as hospitals, clinics, key points, emergency centers, etc.), education POI data (schools, universities, training centers, libraries, etc.), life service POI data (such as laundries, POI data of hotels, guesthouses, youth hostels, etc.), cultural and entertainment POI data (such as cinemas, theaters, stadiums, amusement parks, museums, etc.), government and social groups POI data (such as government agencies, police stations, community service centers, public welfare organizations, etc.), scenic spots POI data (such as tourist attractions, parks, historical sites, nature reserves, etc.), corporate institutions POI data (such as companies, factories, office buildings, entrepreneurial parks, etc.) and public facilities POI data (such as toilets, garbage stations, fire hydrants, etc.);

[0075] The multivariate AOI data includes building AOI data, cultivated land AOI data, green space AOI data, water body AOI data, and road AOI data;

[0076] The economic data includes house price grid data and per capita GDP grid data;

[0077] The population data includes data on the elderly population and data on the permanent population;

[0078] Integrate the location information of existing elderly care service institutions, multi-source POI data, multi-source AOI data, economic data, and population data as multi-source heterogeneous data for site selection, comprehensively consider the impact of various factors on the site selection of elderly care service institutions, improve the accuracy of site selection of subsequent machine learning models, and meet actual needs.

[0079] The preprocessing includes data cleaning, formatting, missing value interpolation, outlier processing and data coding.

[0080] S2, constructing a grid dataset of elderly care services and processing it using feature engineering to obtain a supplementary feature dataset of elderly care services;

[0081] In S2, it is necessary to unify the coordinate system of the multi-source data of the elderly care services in the area to be selected, and combine the location information of the longitude and latitude of the multi-source data of the elderly care services in the area to be selected with the x and y fields of the ArcGIS software, and then use the ArcGIS software fishing net analysis to construct Raster data.

[0082] The elderly care service raster dataset is constructed and processed by feature engineering to obtain an elderly care service feature dataset, including:

[0083] S2-1, determining the grid size based on the regional characteristics of the area to be selected;

[0084] S2-2, based on the grid size, the pre-processed elderly care service multi-source data is processed using the fishing net analysis of ArcGIS software, and the spatial attributes of the pre-processed elderly care service multi-source data are given through the spatial connection function of ArcGIS software to construct the elderly care service raster dataset;

[0085] S2-3. Feature engineering is used to extract features from the elderly care service grid dataset. For example, whether there is an elderly care service structure in the unit grid, if yes, the label is set to 1, otherwise the label is set to 0; whether there is a certain type of POI data in the unit grid, if yes, the label is set to 1, otherwise the label is set to 0; whether there is a certain type of AOI data in the unit grid, if yes, the label is set to 1, otherwise the label is set to 0; extract the POI density in the unit grid (the number of a certain type of POI / grid area), the AOI ratio in the unit grid (a certain type of AOI area / grid area), the average housing price level in the unit grid, the per capita GDP level in the unit grid and other features to obtain the supplementary feature dataset of the elderly care service. .in, Indicates whether there is a pension service institution in the unit grid. Represents all multi-source big data information in the cell grid.

[0086] S3. Construct a machine learning model for the location selection of elderly care service institutions, and optimize the machine learning model for the location selection of elderly care service institutions using K-fold cross validation and grid random search algorithms to obtain the optimal machine learning model for the location selection of elderly care service institutions;

[0087] The machine learning model for site selection of elderly care service institutions includes a decision tree model, a random forest model, an extreme gradient boosting model, a gradient boosting decision tree model, a support vector machine model and a K nearest neighbor model.

[0088] The K-fold cross validation and grid random search algorithm optimizes the elderly care service institution location selection machine learning model to obtain the optimal elderly care service institution location selection machine learning model, including:

[0089] S3-1, obtain multi-source data of historical elderly care services, and process them using the same method as S1 and S2 to obtain a supplementary feature data set of historical elderly care services;

[0090] S3-2, dividing the historical elderly care service supplementary feature data set into K equal subsets, namely ;

[0091] S3-3. Randomly select a subset As validation sets, the remaining K-1 subsets As a training set;

[0092] S3-4, randomly select a subset from the training set;

[0093] S3-5, respectively inputting the training set selected in S3-4 into the decision tree model, the random forest model, the extreme gradient boosting model, the gradient boosting decision tree model, the support vector machine model and the K nearest neighbor model for training, to obtain the corresponding trained decision tree model, random forest model, extreme gradient boosting model, gradient boosting decision tree model, support vector machine model and K nearest neighbor model;

[0094] S3-6, calculating the performance indicators corresponding to the trained decision tree model, random forest model, extreme gradient boosting model, gradient boosting decision tree model, support vector machine model and K nearest neighbor model;

[0095] S3-7, repeat S3-3 to S3-6 until K performance indicators of each machine learning model are obtained;

[0096] Record the number of samples correctly predicted as positive examples by each machine learning model (TP), the number of samples correctly predicted as negative examples by the model (TN), the number of negative samples incorrectly predicted as positive examples by the model (FP), and the number of positive samples incorrectly predicted as negative examples by the model (FN), and calculate the corresponding accuracy, precision, recall, and F1 score.

[0097] S3-8. Calculate the first average performance index corresponding to each machine learning model, and sort them in descending order, and select the machine learning models corresponding to the top two or top three first average performance indicators as the location selection potential models for elderly care service institutions;

[0098] S3-9. Use a grid random search algorithm to optimize the parameters of the elderly care service institution site selection potential model to obtain the optimal elderly care service institution site selection machine learning model.

[0099] The method of optimizing the parameters of the elderly care service institution site selection potential model by using a grid random search algorithm to obtain the optimal elderly care service institution site selection machine learning model includes:

[0100] S3-9-1. Randomly select one of the above-mentioned elderly care service institution location potential models;

[0101] S3-9-2. Define the hyperparameter space of the selected elderly care service institution location potential model , set the parameters to be optimized of the selected elderly care service institution location potential model and their corresponding value ranges; wherein, , , They represent the first hyperparameter, the second hyperparameter and the hyperparameters, , , They represent the first hyperparameter, the second hyperparameter and the The possible value range of each hyperparameter.

[0102] S3-9-3. Selecting the optimal hyperparameter combination of the selected elderly care service institution location potential model using a genetic algorithm;

[0103] The method of selecting the optimal hyperparameter combination of the selected elderly care service institution location potential model by using a genetic algorithm comprises:

[0104] Randomly generate a population of hyperparameter combinations of the selected elderly care service institution location potential model;

[0105] Applying each hyperparameter combination in the hyperparameter combination population to the selected elderly care service institution location potential model and performing training;

[0106] Using a cross-validation method to evaluate the performance indicators of the trained elderly care service institution location potential model corresponding to each of the hyperparameter combinations, and obtain the corresponding fitness;

[0107] Arrange the fitness values ​​in descending order, and select the hyperparameter combination with the top 10% fitness value as the parent generation;

[0108] Calculate the SHAP value corresponding to each feature in the historical elderly care service supplementary feature data set; select core features based on each of the SHAP values ​​and retain them;

[0109] Using an adaptive crossover strategy to exchange the core features, and generating a new generation of offspring based on the parent generation;

[0110] Perform mutation exploration on the offspring of the new generation to obtain updated offspring, i.e., new hyperparameter combinations;

[0111] Repeat the above process until the preset number of iterations M is reached, and M new hyperparameter combinations are obtained; calculate the fitness of the new hyperparameter combination, select the new hyperparameter combination with the largest fitness and apply it to the selected elderly care service institution site selection potential model.

[0112] S3-9-4, repeat S3-9-1 to S3-9-3 until the optimal hyperparameter combination corresponding to each of the elderly care service institution site selection potential models is obtained;

[0113] S3-9-5. Calculate the second average performance index of each of the elderly care service institution site selection potential models;

[0114] S3-9-6. Based on the second average performance indicator, determine the optimal elderly care service institution location selection machine learning model.

[0115] The performance indicators include accuracy, precision, recall, and F1 score; the first average performance indicator and the second average performance indicator are both the average values ​​of accuracy, precision, recall, and F1 score, that is:

[0116]

[0117] in, Indicates The average performance index, Indicates the total number of indicators, Indicates The accuracy value, Indicates The precision value, Indicates The recall value, Indicates F1 score value; The value of is 1 or 2. When it is 1, it means calculating the first average performance index. , , , Respectively represent the accuracy value, precision value, recall value, and F1 score value of the machine learning model; when When it is 2, it means calculating the second average performance index. , , , They respectively represent the accuracy value, precision value, recall value, and F1 score value of the location selection potential model of the elderly care service institution.

[0118] Machine learning algorithms are used to analyze multi-source data, explore potential site selection needs, and optimize site selection plans. K-fold cross validation and grid search algorithms are used to select the optimal machine learning model twice, and then the optimal machine learning model is used for site selection. This can explore potential site selection needs and provide the optimal restriction plan. When applied in different situations, the machine learning model that best suits the current situation can be selected, and dynamic updates and adjustments based on real-time data can ensure that site selection decisions can adapt to changes in urban development and population demand. Genetic algorithms are used to adjust the parameters of the optimal machine learning model, and the best hyperparameter combination is selected for application, which can improve the accuracy of site selection, ensure the reasonable distribution of elderly care service institutions, and reduce resource waste and uneven service problems.

[0119] S4, inputting the supplementary feature data set of elderly care services into the optimal elderly care service institution location selection machine learning model to obtain the supplementary initial location selection result of the elderly care service institution;

[0120] S5. Determine the prediction accuracy probability based on the additional initial site selection result of the elderly care service institution and the elderly care service multi-source data;

[0121] The determining of the prediction accuracy probability based on the supplementary initial site selection result of the elderly care service institution and the elderly care service multi-source data includes:

[0122] S5-1, using ArcGIS software to link the supplementary initial site selection results of the elderly care service institution and the supplementary feature data set of the elderly care service in S2-3 to form a visualized site selection prediction raster data;

[0123] S5-2, determining the number of overlaps between the additional initial site selection results of the elderly care service institution and the elderly care service multi-source data;

[0124] S5-3, according to the formula:

[0125]

[0126] Get the probability of accurate prediction ;in, represents the number of grid overlaps between the additional initial site selection result of the elderly care service institution in the visualized site selection prediction grid data and the existing elderly care service structure in the to-be-selected site selection area, The number of grids representing the existing elderly care service structures in the area to be selected.

[0127] S6, determine whether the prediction accuracy probability is less than a threshold; if so, increase the number of iterations in the genetic algorithm by 50 and return to S3; otherwise, enter S7; wherein the threshold is 75%.

[0128] S7. Based on the demand for elderly care services to be sited, the initial site selection results for the elderly care service institution are divided to obtain priority site selection areas and common site selection areas, thereby completing the supplementary site selection for the elderly care service institution.

[0129] like Figure 2 As shown, based on the demand for elderly care services to be sited, the initial site selection results of the elderly care service institutions are divided to obtain priority site selection areas and common site selection areas, including:

[0130] Determine and delete the grids of the existing elderly care service institutions in the additional initial site selection result of the elderly care service institutions to obtain an updated additional initial site selection result of the elderly care service institutions;

[0131] Determine the demand for elderly care services at the site to be selected, that is, divide the elderly areas (where the proportion of the elderly population aged 65 and above reaches 7% of the total population) and non-elderly areas (where the proportion of the elderly population aged 65 and above does not reach 7% of the total population) according to the elderly population data;

[0132] Spatial fitting is performed on the elderly areas, non-elderly areas and the updated initial site selection results of the elderly service institutions to obtain the priority site selection area and the general site selection area. Based on the global grid created in step S2-1, the resident population density data and the elderly population density data are spatially connected, and the aging degree (elderly population / total population) in the grid is calculated. If the aging degree in the grid is ≥7%, it is assigned a value of 1, otherwise it is 0. The global grid connected with the aging degree is fitted with the supplementary grid of the elderly service institutions. The supplementary grid of the elderly service institutions that overlaps with the grid assigned with a value of 1 is the priority site selection area, otherwise it is the general site selection area.

[0133] A certain area was selected as the site selection area, and multi-source data of elderly care services were crawled and sorted on the Internet, and preprocessed. The unified coordinate system WGS 1984 World Mercator was specified through the projection conversion tool in ArcGIS software. A 1000m×1000m grid was constructed using fishing net analysis, with a grid number of 199696, and spatial data information was given to each unit grid through the spatial connection function.

[0134] After feature engineering extraction, the obtained elderly care service supplementary feature data set was randomly sampled with equal ratios of positive and negative samples, and the ratio of training set to test set was set to 7.5:2.5.

[0135] Through 15-fold cross validation, the decision tree model, random forest model, extreme gradient boosting model, K nearest neighbor model, gradient boosting decision tree model, and support vector machine were selected to identify the random forest model and gradient boosting decision tree as the potential models for the location selection of elderly care service institutions, as shown in Table 1. In Table 1, Decision Tree represents the decision tree model, RF represents the random forest model, XGBoost represents the extreme gradient boosting model, GBDT represents the gradient boosting decision tree model, KNN represents the K nearest neighbor model, and SVM represents the support vector machine.

[0136] Table 1

[0137]

[0138] The hyperparameters of the location potential model of elderly care service institutions were optimized through grid search. The accuracy, precision, recall and F1 score of the random forest model were better than those of the gradient boosting decision tree model. The random forest model with 150 trees with a depth of 10 and square root number of features randomly selected from all features for each split was determined as the optimal location model of elderly care service institutions. The random forest model performed better in balancing complexity, generalization ability and computational efficiency, and was suitable for the current data set.

[0139] The optimal elderly care service institution site selection model outputs the prediction results and forms a visual site selection prediction raster dataset in ArcGIS software. The prediction accuracy probability is determined by fitting the actual elderly care site selection distribution. The prediction accuracy probability is 89.72%, which exceeds 80% to verify the reliability of the model.

[0140] The present invention uses machine learning to select sites for multi-source data of elderly care services, thereby reducing labor costs and dependence on manual analysis and improving the degree of automation in site selection. It optimizes multiple machine learning models and selects the optimal machine learning model for site selection. It has a wide range of applications and can dynamically respond to changing needs, ensuring that the coverage of elderly care facility services matches actual needs and adapts to dynamic changes in urban and population structures.

[0141] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning, characterized in that: include: S1. Real-time acquisition of multi-source data on elderly care services in the area to be selected and pre-processing to obtain pre-processed multi-source data on elderly care services; S2, constructing a grid dataset of elderly care services and processing it using feature engineering to obtain a supplementary feature dataset of elderly care services; S3. Construct a machine learning model for the location selection of elderly care service institutions, and optimize the machine learning model for the location selection of elderly care service institutions using K-fold cross validation and grid random search algorithms to obtain the optimal machine learning model for the location selection of elderly care service institutions; S4, inputting the supplementary feature data set of elderly care services into the optimal elderly care service institution location selection machine learning model to obtain the supplementary initial location selection result of the elderly care service institution; S5. Determine the prediction accuracy probability based on the additional initial site selection result of the elderly care service institution and the elderly care service multi-source data; S6, judging whether the prediction accuracy probability is less than a threshold; if so, returning to S3; otherwise, entering S7; S7. Based on the demand for elderly care services to be sited, the initial site selection results for the elderly care service institution are divided to obtain priority site selection areas and common site selection areas, thereby completing the supplementary site selection for the elderly care service institution.

2. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 1, characterized in that: The multi-source data of elderly care services in the area to be selected includes location information of existing elderly care service institutions, multi-dimensional POI data, multi-dimensional AOI data, economic data, and population data; The location information of the existing elderly care service institution is the latitude and longitude of the existing elderly care service institution; The multi-dimensional POI data includes catering POI data, shopping POI data, transportation POI data, financial POI data, medical POI data, education POI data, life service POI data, accommodation POI data, cultural and sports entertainment POI data, government and social group POI data, scenic spot POI data, enterprise and institution POI data and public facility POI data; The multivariate AOI data includes building AOI data, cultivated land AOI data, green space AOI data, water body AOI data, and road AOI data; The economic data includes house price grid data and per capita GDP grid data; The population data includes data on the elderly population and data on the permanent population; The preprocessing includes data cleaning, formatting, missing value interpolation, outlier processing and data coding.

3. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 2, characterized in that: The elderly care service raster dataset is constructed and processed by feature engineering to obtain an elderly care service feature dataset, including: S2-1, determining the grid size based on the regional characteristics of the area to be selected; S2-2, based on the grid size, the pre-processed elderly care service multi-source data is processed using the fishing net analysis of ArcGIS software, and the spatial attributes of the pre-processed elderly care service multi-source data are given through the spatial connection function of ArcGIS software to construct the elderly care service raster dataset; S2-3. Use feature engineering to extract features from the elderly care service raster dataset to obtain the elderly care service supplementary feature dataset.

4. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 2, characterized in that: The machine learning model for site selection of elderly care service institutions includes a decision tree model, a random forest model, an extreme gradient boosting model, a gradient boosting decision tree model, a support vector machine model and a K nearest neighbor model.

5. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 4, characterized in that: The K-fold cross validation and grid random search algorithm optimizes the elderly care service institution location selection machine learning model to obtain the optimal elderly care service institution location selection machine learning model, including: S3-1, obtain multi-source data of historical elderly care services, and process them using the same method as S1 and S2 to obtain a supplementary feature data set of historical elderly care services; S3-2, dividing the historical elderly care service supplementary feature data set into K equal subsets; S3-3, randomly select a subset as the validation set, and the remaining K-1 subsets as the training set; S3-4, randomly select a subset from the training set; S3-5, respectively inputting the training set selected in S3-4 into the decision tree model, the random forest model, the extreme gradient boosting model, the gradient boosting decision tree model, the support vector machine model and the K nearest neighbor model for training, to obtain the corresponding trained decision tree model, random forest model, extreme gradient boosting model, gradient boosting decision tree model, support vector machine model and K nearest neighbor model; S3-6, calculating the performance indicators corresponding to the trained decision tree model, random forest model, extreme gradient boosting model, gradient boosting decision tree model, support vector machine model and K nearest neighbor model; S3-7, repeat S3-3 to S3-6 until K performance indicators of each machine learning model are obtained; S3-8. Calculate the first average performance index corresponding to each machine learning model, and sort them in descending order, and select the machine learning models corresponding to the top two or top three first average performance indicators as the location selection potential models for elderly care service institutions; S3-9. Use a grid random search algorithm to optimize the parameters of the elderly care service institution site selection potential model to obtain the optimal elderly care service institution site selection machine learning model.

6. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 5, characterized in that: The method of optimizing the parameters of the elderly care service institution site selection potential model by using a grid random search algorithm to obtain the optimal elderly care service institution site selection machine learning model includes: S3-9-1. Randomly select one of the above-mentioned elderly care service institution location potential models; S3-9-2. Define the hyperparameter space of the selected elderly care service institution location potential model, and set the parameters to be optimized of the selected elderly care service institution location potential model and their corresponding value ranges; S3-9-3. Selecting the optimal hyperparameter combination of the selected elderly care service institution location potential model using a genetic algorithm; S3-9-4, repeat S3-9-1 to S3-9-3 until the optimal hyperparameter combination corresponding to each of the elderly care service institution site selection potential models is obtained; S3-9-5. Calculate the second average performance index of each of the elderly care service institution site selection potential models; S3-9-6. Based on the second average performance indicator, determine the optimal elderly care service institution location selection machine learning model.

7. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 6, characterized in that: The method of selecting the optimal hyperparameter combination of the selected elderly care service institution location potential model by using a genetic algorithm comprises: Randomly generate a population of hyperparameter combinations of the selected elderly care service institution location potential model; Applying each hyperparameter combination in the hyperparameter combination population to the selected elderly care service institution location potential model and performing training; Using a cross-validation method to evaluate the performance indicators of the trained elderly care service institution location potential model corresponding to each of the hyperparameter combinations, and obtain the corresponding fitness; Arrange the fitness values ​​in descending order, and select the hyperparameter combination with the top 10% fitness value as the parent generation; Calculate the SHAP value corresponding to each feature in the historical elderly care service supplementary feature data set; select core features based on each of the SHAP values ​​and retain them; Using an adaptive crossover strategy to exchange the core features, and generating a new generation of offspring based on the parent generation; Perform mutation exploration on the offspring of the new generation to obtain updated offspring, i.e., new hyperparameter combinations; Repeat the above process until the preset number of iterations M is reached, and M new hyperparameter combinations are obtained; calculate the fitness of the new hyperparameter combination, select the new hyperparameter combination with the largest fitness and apply it to the selected elderly care service institution site selection potential model.

8. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 5 or 6, characterized in that: The performance indicators include accuracy, precision, recall, and F1 score; the first average performance indicator and the second average performance indicator are both the average values ​​of accuracy, precision, recall, and F1 score, that is: in, Indicates Average performance indicators, Indicates the total number of indicators, Indicates The accuracy value, Indicates The precision value, Indicates The recall value, Indicates F1 score value; The value of is 1 or 2. When it is 1, it means calculating the first average performance index. , , , Respectively represent the accuracy value, precision value, recall value, and F1 score value of the machine learning model; when When it is 2, it means calculating the second average performance index. , , , They respectively represent the accuracy value, precision value, recall value, and F1 score value of the location selection potential model of the elderly care service institution.

9. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 3, characterized in that: The determining of the prediction accuracy probability based on the supplementary initial site selection result of the elderly care service institution and the elderly care service multi-source data includes: S5-1, using ArcGIS software to link the supplementary initial site selection results of the elderly care service institution and the supplementary feature data set of the elderly care service in S2-3 to form a visualized site selection prediction raster data; S5-2, determining the number of overlaps between the additional initial site selection results of the elderly care service institution and the elderly care service multi-source data; S5-3, according to the formula: Get the probability of accurate prediction ;in, represents the number of grid overlaps between the additional initial site selection result of the elderly care service institution in the visualized site selection prediction grid data and the existing elderly care service structure in the to-be-selected site selection area, The number of grids representing the existing elderly care service structures in the area to be selected.

10. A method for selecting additional locations for elderly care service institutions based on multi-source data and machine learning according to claim 3, characterized in that: The additional initial site selection results of the elderly care service institution are divided based on the elderly care service demand to obtain priority site selection areas and common site selection areas, including: Determine and delete the grids of the existing elderly care service institutions in the additional initial site selection result of the elderly care service institutions to obtain an updated additional initial site selection result of the elderly care service institutions; Determine the demand for elderly care services at the site to be selected, that is, divide the elderly area and non-elderly area according to the elderly population data; Spatial fitting is performed on the elderly areas, non-elderly areas and the updated initial site selection results of the elderly care service institutions to obtain the priority site selection area and the general site selection area.

Images