Date of Award


Document Type

Open Access Dissertation




College of Arts and Sciences

First Advisor

Diansheng Guo


Data on spatial mobility have become increasingly available with the wide use of location-aware technologies such as GPS and smart phones. The analysis of movements is involved in a wide range of domains such as demography, migration, public health, urban study, transportation and biology.

A movement data set consists of a set of moving objects, each having a sequence of sampled locations as the object moves across space. The locations (points) in different trajectories are usually sampled independently and trajectory data can become very big such as billions of geotagged tweets, mobile phone records, floating vehicles, millions of migrants, etc. Movement data can be analyzed to extract a variety of information such as point of interest or hot spots, flow patterns, community structure, and spatial interaction models. However, it remains a challenging problem to analyze and map large mobility data and understand its embedded complex patterns due to the massive connections, complex patterns and constrained map space to display.

My research focuses on the development of scalable and effective computational and visualization approaches to help derive insights from big geographic mobility data, including both origin-destination (OD) data and trajectory data. Specifically, my research contribution has two components: (1) flow clustering and flow mapping of massive flow data, with applications in mapping billions of taxi trips (Chapter 2 and Chapter 3); and (2) time series analysis of mobility, with applications in urban event detection (Chapter 4).

Flow map is the most common approach for visualizing spatial mobility data. However, a flow map quickly becomes illegible as the data size increases due to the massive intersections and overlapping flows in the limited map space. It remains a challenging research problem to construct flow maps for big mobility data, which demands new approaches for flow pattern extraction and cartographic generalization. I have developed new cartographic generalization approaches to flow mapping, which extract high-level flow patterns from big data through hierarchical flow clustering, kernel-based flow smoothing, and flow abstraction. My approaches represent a significant breakthrough that enables effective flow mapping of big data to discover complex patterns at multiple scales and present a holistic view of high-level information.

The second area of my research focuses on the time series analysis of urban mobility data, such as taxi trips and geo-social media check-ins, to facilitate scientific understanding of urban dynamics and environments. I have developed new approaches to construct location-based time series from mobility data and decompose each mobility time series into three components, i.e. long-term trend, seasonal periodicity pattern and anomalies, from which urban events, land use types, and changes can be inferred. Specifically, I developed time series decomposition method for urban event detection, where an event is defined as a time series anomaly deviating significantly from its regular trend and periodicity.

Included in

Geography Commons