Abstract
Recent advancement on the internet of things technology has opened up various opportunities in collecting day-to-day human activity data for various smart city applications. Location-based human activity is one of the important aspects in smart city applications as most of the applications are designed based on the objective of providing a better livable space for urban residents. Therefore, mining insights about human activity helps to better understand the relationship between human and space in a smart city context. However, the deployment of insight mining about human activity often requires domain knowledge and thoughtful design of implementation. By implementing data fusion into insights mining, choosing the ideal data sources for performing such a task is challenging. In this thesis, the insights mining method is designed to obtain insights about human activity leveraging data fusion in three different scales, which are the building-level, point of interest, and region of interest. Through careful evaluation and identifying the constraint of existing smart city applications, the design of location-centric human activity monitoring has been deployed. First, the deployment framework of human activity monitoring at a point of interest is proposed and aims to capture human activity by considering aspects such as sensor fusion, data quality, and data collection longevity. Physical sensor fusion technique has demonstrated fruitful capturing of human activity while providing some interesting findings. Next, to study region of interests using a location-centric approach, the segmentation of the active and non-active point of interests helps to understand potential factors that drive public space utilization. Through the merging of cluster result and static features, potential factors that drive public space utilization were identified. Lastly, the deviated behaviour contrasted to the normal human activity in a buildinglevel has been explored. Leveraging the hybrid norm generated through the fusion of normal activity and group activity, different types of abnormal human activity can be further analysed and categorize