ANALYZING DETERMINANTS OF URBAN VIBRANCY – A BIG DATA APPROACH ON CONNECTING BUILT ENVIRONMENT, SOCIAL ACTIVITY, AND IMAGES OF PLACES

This dissertation includes three self-contained and interrelated papers on built environment, urban vibrancy and vibrant places. Paper 1: Built environment and social activity: Exploring the relationship between place pattern and tweeting in Chicago Understanding the relationships between human activities and the built environment are central to urban planning. The increase in readily available, location-based social media data offers scholars important new data for understanding this relationship. This study examines the relationship between the spatial distribution of geotagged tweets and key characteristics of the built environment at the census block group level in Chicago. First, we performed a hotspot analysis to ascertain the distribution of tweets in the study area. Then, we employed a count regression model with Twitter message counts by census block group as the dependent variable to test the significance and magnitude of the associations between the built environment and tweeting. Then, we standardized the coefficients to compare the variables’ effects on tweeting. The analysis found that the built environment significantly influenced tweeting and provides empirical statistical evidence to guide urban planners’ placemaking decisions. Paper 2: Built environment and social activity: Exploring the relationship between place pattern and tweeting in the U.S. This study focuses on the relationship between the spatial distribution of geotagged tweets and the key characteristics of built environments as well as the pattern of the features at the census block group (CBG) level within the contiguous U.S. First, we employed a count regression model, setting the Twitter message density by CBG as the dependent variable to test the significance and magnitude of the associations between the built environment and tweeting behavior. Then, we utilized a combined research framework based on cluster analysis, hotspot analysis, and standardized score to explore the built environment pattern of the most vibrant tweeting areas. Results revealed that the built environment significantly influenced tweeting behavior. We further discovered four different built environment pattern types that provides empirical statistical evidence to guide urban planners’ placemaking decisions. Paper3: Built environment and vibrancy perception: Applications of deep learning and computer vision techniques in streetview. This study shows that street view image data from digital platforms such as Google Maps can further improve our understanding of the built environment patterns in cities. Compared with traditional human environment audit methods, combining deep learning and computer vision techniques efficiently provides finer resolution information with greater environmental detail. This will allow for many larger-scale street view studies to be conducted in the future. Using online survey data on vibrancy perception, the findings indicate that factors such as street parking have a strong positive influence on vibrancy perception while the proportion of sky has a strongly negative association. The study identifies six different patterns of street view landscape and shows that the complex relationship between the built environment and vibrancy perception is better analyzed using pattern discovery methods rather than global regression. A further regression analysis based on each street view pattern confirms that the context is crucial in determining both the significance and magnitude of the built environment factors on urban vibrancy perception.Doctor of Philosoph