多源大数据下的北京市居民就餐活动与城市空间关系探究

融合了位置和文本信息的社交媒体数据为城市空间研究提供了有力的数据支撑,也为发现其背后隐藏的人类行为时空模式及规律提供了可能.本文构建了文本信息挖掘技术与空间分析有机结合的居民时空行为研究框架,利用2017年微博数据,采用BERT与Fast.AI结合的文本分类模型,结合LDA主题模型进行文本主题挖掘,对北京居民日常就餐活...

STUDY ON THE RELATIONSHIP BETWEEN DINING ACTIVITIES OF BEIJING RESIDENTS AND URBAN SPACE BASED ON MULTI-SOURCE BIG DATA

This paper constructs a residential spatiotemporal behavior research framework that combines text information mining technology and spatial analysis. By obtaining the Sina Weibo data of Beijing residents in 2017, using the text classification model combining BERT and fast.AI, combining with the LDA model for text theme mining. Analyze the spatial pattern of residents' daily dining activities, and use spatial analysis methods and the Geodetector to explore its influencing factors. The study concluded that the residents' dining activities can be divided into 4 categories of topics, namely friends gathering, daily catering, general catering and special catering. The spatial analysis found that the four types of theme dining activities are mainly distributed within the Third Ring Road, forming a hierarchical distribution pattern centered on the Workers' Stadium-Chaowai-CBD business district. Meanwhile, various theme dining activities also have the common characteristics of dense distribution along important business districts, famous blocks, popular attractions, and large shopping malls. Residents' dining choices have the strongest consistency with the spatial distribution of catering service facilities. It is found that there is a spatial co-location model between residents' dining activities and the urban spatial structure combined with POI data.