研究方法

Research method

首先选取石景山区218个小区来自于九个街道，再应用20个相关性因子（主要是老年人口指标、老年服务圈质量指标、老年保健设施质量的指标）。在该研究中先进行皮尔森相关性分析，然后运用监督机器学习进行了回归分析和建模研究,进一步探讨影响因子在社区老年服务的质量水平。

研究区域

根据第七次人口普查，石景山区的常住人口为567851,60岁以上共有137768名占24.3%。而在有的社区例如八角社区，老龄化社区也是越来越多，有的小区的老龄化人口甚至达到了40%。

First, 218 residential districts in Shijingshan District were selected from nine streets, and then 20 correlation factors were applied (mainly elderly population index, elderly service circle quality index, elderly health care facility quality index). In this study, Pearson correlation analysis was first carried out, and then supervised machine learning was used to conduct regression analysis and modeling research to further explore the level of influencing factors in the quality of community elderly services.

Study area

According to the 7th census, the permanent population of Shijingshan District was 567,851, with a total of 137,768 or 24.3% aged 60 or above. In some communities, such as Bajiao community, there are more and more aging communities, and the aging population of some communities even reaches 40%.

数据类型

data type

1. 病床覆盖率：yanglao.com.cn

2. 设施可访问性：采用Gaussian 2-step Floating Catchment Area (G2SFCA)，G2SFCA方法估计的可访问性老年保健设施在两个阶段,分别取决于供给和需求的地方。在这篇文章中,供需双方为老年设施和社区

3. 街道连接性：采用阻抗大小研究街道可访问性,并使用O-D成本矩阵方法计算，数据来源于百度地图。

4. 老年人服务区指数：经过试验发现15分钟步行圈比单纯的用行政区更加精准。

5. 功能密度：主要的功能种类包括:老年保健设施、医疗设施、餐饮、公园、公共设施、购物设施,交通电台、教育设施、生活服务设施、体育设施、政府机构和住房。
   

6. 视觉感知：通过腾讯地图的api接口获取了1346个位置点聚集到5384街景图片，分析其绿视率和蓝天率。运用语义分割cnn的方法达到这个目的。

7. 老年人口指标：包括老年人密度、占比和死亡率。

1. Bed coverage: yanglao.com.cn

2. Facility accessibility: Using the Gaussian 2-step Floating Catchment Area (G2SFCA), the G2SFCA method estimates the accessibility of elderly care facilities in two stages, depending on supply and demand, respectively. In this article, supply and demand sides for senior facilities and communities

3. Street connectivity: Impedance size was used to study street accessibility, and O-D cost matrix method was used to calculate the data from Baidu map.

4. Elderly service area index: The experiment found that the 15-minute walking circle is more accurate than the simple administrative area.

5. Functional density: The main functional types include: elderly health care facilities, medical facilities, catering, parks, public facilities, shopping facilities, transportation stations, educational facilities, living service facilities, sports facilities, government agencies and housing.

6. Visual perception: The api interface of Tencent Map obtained 1346 location points gathered to 5384 street view pictures, and analyzed its green vision rate and blue sky rate. The method of semantic segmentation of cnn is used to achieve this goal. 

7. Indicators of the elderly population: including the density, proportion and mortality rate of the elderly.

机器学习

machine learning

将样本分成3份，70%的是训练集、15%是验证集、15%是因子归一化分析。

The sample is divided into 3 parts, 70% is the training set, 15% is the verification set, 15% is the factor normalization analysis.

研究结果

Research results

首先将各个街道的数据和gis地图相关联，得到各种数据在不同街道的差异性感受。如图可知，B和F有最大比例的老年人群，B和E的老年人密度最高，G和C的死亡率最高。相比南部地区，石景山北部地区有更好的老年人服务设施。在这之中，B的服务设施覆盖性最好，另外绿视率和蓝天绿西部更高是因为有更好的山水自然环境。

其次对养老设施的覆盖率进行分析，发现运用步行圈比行政区划更加合理。经过分析发现养老服务水平高的地方主要是政府的中心。养老服务设施水平和死亡率并不完全是反比例关系，例如G服务水平很高但是死亡率也是最高的。

接着进行因子相关性分析，结果表明老年服务的覆盖圈,老年保健设施密度和街头环境会显著影响老年人的健康

First of all, the data of each street is associated with the gis map, and the different feelings of various data in different streets are obtained. As can be seen from the figure, B and F have the largest proportion of elderly people, B and E have the highest density of elderly people, and G and C have the highest mortality rate. Compared with the southern area, the northern area of Shijingshan has better service facilities for the elderly. Among them, B has the best coverage of service facilities, in addition, the green vision rate and blue sky green West are higher because there are better mountains and rivers natural environment.

Secondly, it analyzes the coverage rate of elderly care facilities and finds that the use of walking circle is more reasonable than administrative division. After analysis, it is found that places with high levels of elderly care services are mainly government centers. There is not a completely inverse relationship between the level of service facilities for the elderly and the mortality rate. For example, the service level of G is very high, but the mortality rate is also the highest.

The results of factor correlation analysis showed that the coverage circle of geriatric service, the density of geriatric health care facilities and street environment would significantly affect the health of the elderly