研究内容

Research contents

（1）上海市小区生活圈内养老助餐空间可达性的现状测度。借助多要素修正的两步移动搜索方法，对上海市各小区的老年人口助餐需求和助餐供应能力之间的供求匹配程度进行计算；基于计算结果进行描述统计、多层级划定、研究单元敏感性分析以及设施缺失量计算；

（2）助餐可达性数值集聚区域的划定。使用集计模型进行多行政区级别的结果集计对比；借助空间自相关方法，对上海市养老助餐服务建设现状发展呈现集中不均衡的区域进行划定；

（3）养老助餐设施空间分布影响因素识别。借助机器学习技术结合城市多源大数据，对上海市当前养老助餐服务的分布现状的影响因素进行研究探讨；然后使用地理加权回归技术对影响要素的空间异质性进行解读。

(1) The current measurement of the space accessibility of assisted meals for the aged in the residential community in Shanghai. Based on the two-step mobile search method with multi-factor modification, the matching degree of supply and demand between the elderly population's dietary assistance demand and dietary assistance supply capacity was calculated. Based on the calculation results, descriptive statistics, multi-level delineation, sensitivity analysis of research units and calculation of facility missing amount were carried out.

(2) Demarcation of cluster areas for the accessibility of auxiliary meals. The results of multi-district level were compared by the set model. With the help of spatial autocorrelation method, the paper delimits the areas with concentrated and unbalanced development of assisted meal service construction in Shanghai.

(3) Identification of factors affecting the spatial distribution of assisted meal facilities for the elderly. With the help of machine learning technology and multi-source big data in the city, the influential factors of the current distribution of assisted meals for the elderly in Shanghai were studied and discussed. Then the spatial heterogeneity of the influencing factors is interpreted by geographical weighted regression technique.

上海市社区养老助餐设施空间可达性

The spatial accessibility of community elderly care assisted meal facilities in Shanghai

1.研究尺度的差异：

1. Differences in research scales:

可达性会随着研究距离也就是生活圈大小发生变化，该研究首先探讨该问题，明确现在上海市比较充分供应老年助餐的空间距离分布，如下图展示了在不同研究单元尺度下的养老助餐服务设施可达性（以下简称可达性）数值的统计结果。从结果可知，不同研究单元尺度下，可达性数值的平均值和标准差均呈现先增后减趋势。最大值和最小值之间的数值差距较大，但是可达性数值的最大和最小值在不同单元尺度下几乎没有变化。

Accessibility will change with the research distance, that is, the size of the living circle. This study first discusses this issue and defines the spatial distance distribution of the elderly assisted meals in Shanghai. The figure below shows the statistical results of the accessibility of elderly assisted meals service facilities (hereinafter referred to as accessibility) at different research unit scales. It can be seen from the results that the mean and standard deviation of the reachability value increase first and then decrease at different research unit scales. The numerical difference between the maximum and minimum values is large, but the maximum and minimum values of the reachability values are almost unchanged at different cell scales.

2.不同生活圈尺度下的可达性低值小区的精准识别：

2. Accurate identification of low-reachability communities under different life circle scales:

  在可达性数值结果基础上，将 20 000多小区的可达性数值进行有意义的分层级划分，可以更便于评价助餐现状，同时也有利于低值小区的精准识别。

  借鉴北京市关于养老助餐设施的政策文件，其要求城市社区的助餐设施点至少服务 2000 老年人，即 5 个/万人，该文将该线划定为标准线。截至 2021年末，上海市总共有老年人口 542.22 万，养老助餐设施可获取到 1 722 个，由此可以计算平均比值为3.176 个/万人，我们将次数值划定为平均线；同时为了在标准线以上的部分对资源充足的小区进一步区分，参考了部分研究依据以及访谈的调研结果，将两倍及格线数值划定为优异线。

  在 ArcGIS 软件中将可达性数值计算结果分为 A、B、C、D 四个等级，分别是：A[10.001, 200]、B[5.001, 10.000]、C[3.177, 5.000]、D[0, 3.176]。其中，D 类低于平均线，C 类为低于标准线，B 类高于标准线，A 类可达性最好。不同研究单元尺度下，四类小区的空间分布在25min生活圈分布如下图所示。

Based on the reachability results, the reachability values of more than 20 000 cells were meaningfully stratified, which could be more convenient to evaluate the current situation of auxiliary meals, and also conducive to the accurate identification of low-value cells.

Drawing on Beijing's policy document on assisted meal facilities for the elderly, which requires that assisted meal facilities in urban communities serve at least 2,000 elderly people, that is, 5 per 10,000 people, the paper defines this line as a standard line. By the end of 2021, Shanghai has a total elderly population of 5.422 million, and 1 722 assisted meal facilities for the elderly can be obtained. Therefore, the average ratio can be calculated as 3.176 per 10,000 people, which is defined as the average line. At the same time, in order to further distinguish the communities with sufficient resources in the part above the standard line, we refer to some research bases and the survey results of interviews, and delimit the value of twice the pass line as the excellent line.

In ArcGIS software, the reachability numerical calculation results are divided into four grades: A[10.001, 200], B[5.001, 10.000], C[3.177, 5.000], D[0, 3.176]. Among them, Class D is lower than the average line, Class C is lower than the standard line, Class B is higher than the standard line, and class A has the best accessibility. Under different research unit scales, the spatial distribution of the four types of cells in the 25min life circle is shown in the figure below.

3.各小区生活圈内养老助餐设施缺失量计算：

3. Calculation of the missing amount of assisted meal facilities for the aged in the living circle of each community:

是对于政府来说，明确助餐设施缺失数量十分关键。研究依托 15 min 阈值下的助餐可达性分布现状，计算推衍各小区欲达各分割线所需养老助餐设施数量。按照 15 min 养老圈的建设标准，欲达平均线，目前缺失各小区需要补充数量主要在 1~8 个之间（上下四分位数），最多的需要补充 12 个养老服务设施，整体来看，各小区需要补充的数量标准差为 3.38 左右，差异较小；欲达标准线，各小区需要补充助餐设施数量主要在 2 至 14 个之间（上下四分位数），最多的需要补充 20 个助餐设施，整体来看，各小区需要补充的数量标准差为 5.79 左右；欲达养老助餐服务可达性优异线，各小区需要补充助餐设施数量主要在 5 个到 29 个之间（上下四分位数），最多的需要补充43 个养老服务设施，整体来看，各小区需要补充的数量平均值为 12.4 个，标准差为 12.42 左右，差异较明显。

It is crucial for the government to identify the number of missing feeding facilities. Based on the distribution of assisted meal accessibility under the threshold of 15 min, the study calculated and deduced the number of assisted meal facilities for the aged in each community to reach each division line. According to the construction standard of 15-min elderly care circle, in order to reach the average line, at present, the number of missing communities needs to be supplemented mainly ranges from 1 to 8 (upper and lower quartiles), and the most need to supplement 12 elderly care service facilities. On the whole, the standard deviation of the number of supplementary communities needs to be about 3.38, and the difference is small. In order to reach the target, the number of supplementary feeding facilities in each community mainly ranges from 2 to 14 (upper and lower quartiles), and the most needs to supplement 20 feeding facilities. On the whole, the standard deviation of the number of supplementary facilities in each community is about 5.79. In order to reach the excellent line of accessibility of assisted meal service for the elderly, the number of assisted meal facilities in each community mainly ranges from 5 to 29 (upper and lower quartiles), and the most needs to supplement 43 elderly service facilities. On the whole, the average number of supplementary facilities in each community is 12.4, and the standard deviation is about 12.42, with obvious differences.

城市多维度相关影响因素的建模和要素识别结果

Modeling and element identification results of multi-dimensional correlation influencing factors in cities

为了探究哪些因子最为影响老年助餐机构，选用随机森林进行建模和相关性分析，选用以下要素因子：养老机构可达性、长者照护之家可达性、日间照料中心可达性、护理站可达性、医疗可达性、道路交叉口密度、公交密度、平均建筑体型系数、平均建筑高度、容积率、建筑密度、建筑高度标准差、餐饮密度、景点密度、用地混合度、工厂密度、酒店密度、人口密度、平均房价、人均收入和房租均值 21 项指标。这些特征影响因子作为预测变量建立机器学习模型。解释结果如下

In order to explore which factors most affect feeding institutions for the elderly, random forest was selected for modeling and correlation analysis, and the following factors were selected: Accessibility of elderly care institutions, accessibility of elderly care homes, accessibility of day care centers, accessibility of nursing stations, accessibility of medical services, density of road intersections, density of public transport, average building size coefficient, average building height, plot ratio, building density, standard deviation of building height, catering density, scenic spot density, land mix, factory density, hotel density, population density, average Housing price, per capita income and rent average 21 indicators. These feature influence factors are used as predictive variables to build machine learning models. Explain the results as follows

在上海市中心城区，养老助餐可达性的特征影响因子，排列前四项依次为：日间照料中心可达性、养老机构可达性、长者照护之家可达性、护理站可达性。从全局影响方向来看，助餐可达性与养老机构可达性为负向相关，其余三者是正向相关。

可以明确的是，日间照料中心、长者照护之家、护理站都同属于社区养老设施，从数据上来看，这些设施通常与养老助餐设施呈现集中分布。而养老机构属于区域级的养老设施，由于占地面积相对较大和提供长期护理服务的原因，因此通常远离居民区。而养老助餐设施目前主要是为社区居家养老服务所设立，因此这样的结果是符合认知的。

而养老机构可达性在模型中的影响因素较大，也说明目前其布局与社区养老设施的布局是有负向的内在关联。

In the central urban area of Shanghai, the top four influencing factors of the accessibility of assisted meals for the elderly are: accessibility of day care centers, accessibility of elderly care institutions, accessibility of elderly care homes, and accessibility of nursing stations. From the perspective of global influence, the accessibility of assisted meals is negatively correlated with the accessibility of pension institutions, while the other three are positively correlated.

It is clear that day care centres, elderly care homes and nursing stations are all community aged care facilities, and from the data point of view, these facilities are usually concentrated with elderly assisted meal facilities. Elderly care institutions belong to the regional level of elderly care facilities, because of the relatively large area and the provision of long-term care services, so usually far away from residential areas. At present, the elderly meal assistance facilities are mainly set up for community home care services, so this result is in line with cognition.

The accessibility of pension institutions has a large influence factor in the model, which also indicates that the current layout of pension institutions has a negative internal correlation with the layout of community pension facilities.