Research on the Relationship between Urban Spatial Form and Street Attractiveness and Design Suggestions Based on Multi-source Big Data and Deep Learning Methods —— A Case Study of Representative Blocks in Hong Kong

Taking Hong Kong’s representative blocks as the research object, this study explores the relationship between various types of urban spatial form indicators and street attractiveness by establishing a quantitative analysis model based on multi-source big data. The study is based on web data crawling, using LDM model and sliding window method to determine the scope of the experiment, constructing a map of Hong Kong’s gastronomy, and calculating the attractiveness of each block as the dependent variable of the model. Using SDNA, spatial statistics of POI data, semantic segmentation and other methods, a macro and meso-micro scale urban form evaluation index system is constructed, and each index is used as the independent variable. The results show that (1) accessibility, population density and street attractiveness have a non-linear relationship, and functional mixing degree is positively correlated with attractiveness (2). Among the streetscape indicators, green view index and closure have an inverted U-shaped relationship with block attractiveness, while walkability and imageability are basically positively correlated with attractiveness (3). With regard to the streets where shops with different consumption characteristics gather, we propose shop location and street design suggestions to enhance the attractiveness of the blocks from the perspective of their macroscopic characteristics and interface perceptions.