Toward smarter cities : exploring data-driven approaches to human-land interaction by artificial urban intelligence

Abstract

The rapid development of advanced technologies and the advent of the big data era have ushered in unprecedented opportunities for understanding and improving urban spaces. As urbanization continues to reshape our world, there is an increasing need to study human-land interactions and their impacts on urban environments. Artificial urban intelligence, a domain-specific application of artificial intelligence techniques for urban-related tasks, plays a crucial role in addressing this need. Emphasizing human-land interaction in urban applications is essential to developing smarter cities that are more sustainable, efficient, and adaptable. This thesis aims to contribute to the body of knowledge in urban environment comprehension, human mobility understanding, and location recommendation by investigating a series of challenges and limitations of existing methodologies, and proposing novel frameworks and techniques to overcome these obstacles. In the first chapter, we provide an introduction to the background and scope of the research, highlighting the significance of human-land interaction in artificial urban intelligence applications. The second chapter reviews the current state of the art, examining the methods employed in urban environment comprehension and human mobility understanding. In the third chapter, we propose a novel multi-graph framework called Region2Vec for urban region representation learning. The framework captures inter-region relations through human mobility, geographical contextual information via neighborhood data, and intra-region information using Point of Interest (POI) side information in knowledge graphs. Experiments on real-world datasets demonstrate the effectiveness of Region2Vec, consistently outperforming state-of-the-art baselines in various tasks and metrics. The fourth chapter is divided into two parts: human mobility analysis and location recommendation. We use tourist travel patterns as a case study and employ trip chains to model and discover fixed patterns. In the location recommendation section, we propose a novel Temporal Prompt-based and Geography-aware (TPG) framework, which excels in interval prediction on various real-world datasets. In the final chapter, we provide a conclusion for the thesis, summarizing the key findings and contributions made to the field of artificial urban intelligence. The proposed techniques hold great potential for further development and application in the pursuit of smarter cities and more intelligent urban environments.