Effective and efficient techniques for analyzing spatio-temporal sensor data from the urban enviornment are paramount, particularly in addressing these key growth areas in urbanization: human mobility, transportation, and energy consumption. One main challenge in spatio-temporal analytics of large scale sensor data is to discover meaningful correlations among thousands of sensors. It is important to observe and learn the context from which the data is generated in, particularly when dealing with heterogenous highdimensional data from buildings, cities, and urban areas. 

I will firstly introduce our Information-Gain based temporal segmentation techniques that can be used for discovering transitions in human mobility data, extracting temporal features from multiple different sensor data, finding change points in data streams, and summarising temporal patterns. I will also briefly present a recent paper on a Bayesian Non Parametric technique to discover both contexts (eg social contexts, physical activities) and also groups of users that have similar observable contexts (which may indicate that they belong to a social group). 

I will then present a new model of spatio-temporal interval data (which are generally found in infrastructure sensor data e.g. parking sensor, WiFi data in shopping malls, etc). Clustering this data is useful for hot-region detection across different times of day. This paper presents a new approach to evaluate clustering methods across spatial, temporal, and data domains, and propose new similarity and balance metrics to evaluate these clusters. 

Lastly, I will introduce a couple of domain applications of our research for smarter cities and smarter buildings, including human mobility analysis, intelligent transportation, indoor analytics, and energy consumption prediction.