地理学报 ›› 2019, Vol. 74 ›› Issue (8): 1637-1649.doi: 10.11821/dlxb201908011

• 人口与区域发展 • 上一篇    下一篇

基于众源数据挖掘的中国饮食口味与慢性病的空间关联

李瀚祺1,贾鹏2,3,费腾1   

  1. 1. 武汉大学资源与环境科学学院,武汉 430079
    2. 荷兰特文特大学空间健康研究中心,荷兰 恩斯赫德 7514
    3. 空间生命历程流行病学国际合作网络,荷兰 恩斯赫德 7514
  • 收稿日期:2018-11-26 修回日期:2019-06-04 出版日期:2019-08-25 发布日期:2019-08-07
  • 作者简介:李瀚祺(1994-), 女, 硕士生, 研究方向为健康地理学、空间数据分析。E-mail: <email>lihanqitongxue@126.com</email>

Geographical association between dietary tastes and chronic diseases in China:An exploratory study using crowdsourcing data mining techniques

LI Hanqi1,JIA Peng2,3,FEI Teng1   

  1. 1. School of Resource and Environmental Sciences, Wuhan University, Wuhan 430079, China
    2. GeoHealth Initiative, Department of Earth Observation Science, Faculty of Geo-Information Science and Earth Observation (ITC), University of Twente, Enschede, 7514, The Netherlands
    3. International Initiative on Spatial Lifecourse Epidemiology (ISLE), 7514 Enschede, The Netherlands
  • Received:2018-11-26 Revised:2019-06-04 Online:2019-08-25 Published:2019-08-07

摘要:

慢性病是全球最主要的死亡原因,在所有慢性病风险因素中,不健康饮食因素居于首位,也是影响最广泛的风险因素。尽管已有许多关于饮食行为的研究,但在饮食口味与慢性病关联方面尚缺乏定量研究。鉴于此,利用众源网络菜谱数据,提取菜系中多维口味信息,结合不同地区分类的餐饮类兴趣地点(POI)数据,定量分析不同地区人群口味偏好;使用地理探测器方法,从空间分异性角度建立7种口味与出血性卒中、胰腺癌、上呼吸道感染3种慢性病的关联,得到饮食口味对慢性病空间分布的解释能力度量值。结果表明:在7种口味中,过咸是出血性卒中的首要口味风险因子;一定程度的甜是胰腺癌的首要口味风险因子,且甜的程度与胰腺癌风险并非呈简单线性关系;过辛是上呼吸道感染的首要口味风险因子,三者在统计上均表现显著。本文首次提出了基于众源数据挖掘的潜在健康风险因素定量研究方法,可以应用于病因的探索性分析,并有助于公共卫生部门制定相应的干预措施。

关键词: 慢性病, 风险因素, 饮食口味, 众源数据, 地理探测器

Abstract:

Chronic diseases are the main cause for death in the world. Among all risk factors concerning chronic diseases, those related to an unhealthy diet are most important. Although much research was done on dietary behavior, there are only few quantitative studies on the relationship between dietary taste and chronic diseases. In this article, a taste dataset of the major categories of Chinese cuisine is established based on crowdsourced data from Chinese recipe websites. For a quantitative analysis of people's taste in different regions, additionally the locations of restaurants by category (using their respective points of interest) are integrated. Using the software Geodetector, these regional taste preferences are then correlated with the three chronic diseases, hemorrhagic stroke, pancreatic cancer, and upper respiratory tract infection. For all the three diseases, the results indicate very salty, moderate sweet and very spicy food as the primary risk factors. Also, the degree of sweetness is not linear with the risk of pancreatic cancer. These results are statistically significant. In this study, a quantitative method on discovering potential health risk factors based on mining of crowdsourced data is proposed for the first time. This method can be applied before disease-related experiments to filter potential factors, and it is helpful for the public health department to make quick corresponding intervention policies.

Key words: chronic disease, risk factor, dietary tastes, crowdsourcing data mining, Geodetector