This thesis introduces a cyberGIS and machine learning framework for urban analytics. Due to the rapid urbanization and global changes, it is critical to understand urban environments and the complexity in the urban systems. The framework bridges the gap between heterogeneous geospatial big data and the urban complex system, proposing a novel approach for urban analytics. Applied across three thesis chapters focused on the urban heat issue, the framework aims to model and predict urban heat with fine spatiotemporal granularity, (near) real-time, and high precision using heterogeneous urban big data.
The first chapter showcases the integration of cyberGIS and machine learning for predicting Urban Heat Island in Chicago, achieving high spatiotemporal granularity, achieving modeling at 10 minutes temporal granularity.
The second chapter aims to conduct (near) real-time evaluation and mapping of human sentiments of heat exposure using Location-based Social Media data using keywork-based natural language processing algorithm and backend supercomputer.
The third chapter introduces a video machine learning framework for urban spatiotemporal analysis, showcasing advantages such as integrated factors, applicability to diverse urban issues, handling of heterogeneous geospatial data, adaptable spatiotemporal granularity, and very high precision, which is effectively demonstrated in predicting urban heat dynamics.
Overall, these chapters highlight the achievements of the proposed cyberGIS and machine learning framework for urban analytics, offering fine spatiotemporal granularity, real-time application, and high accuracy. This innovative urban analytics framework contributes to the understanding of urban heat dynamics and provides effective tools for general urban analytics.
