Traditional occupant behavior modeling has been studied at the building level, and it has become an important factor in the investigation of building energy consumption. However, studies modeling occupant behaviors at the urban scale are still limited. Recent work has revealed that urban big data can enable occupant behavior modeling at the urban scale – however, utilizing the existing data sources and modeling methods in building science to model urban scale occupant behaviors can be quite challenging. Beyond building science, urban scale human behaviors have been studied in several different domains using more advanced modeling methods, including Stochastic Modeling, Neural Networks, Reinforcement Learning, Network Modeling, etc. This paper aims to bridge the gap between data sources and modeling methodologies in building science by borrowing from other domains. Based on a comprehensive review, we 1) identify the modeling challenges of the current approaches in building science, 2) discuss the modeling requirements and data sources both in building science and other domains, 3) review the current modeling methods in building science and other domains, and 4) summarize available performance evaluation metrics for evaluating the modeling methods. Finally, we present future opportunities in building science with enhanced data sources and modeling methods from other domains.