Date of Award

Spring 5-5-2023

Level of Access Assigned by Author

Open-Access Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Spatial Information Science and Engineering

Advisor

Kate Beard-Tisdale

Second Committee Member

Silvia Nittel

Third Committee Member

Torsten Hahmann

Additional Committee Members

Kathleen Bell

Sean Smith

Abstract

Similarity or distance measures between data objects are applied frequently in many fields or domains such as geography, environmental science, biology, economics, computer science, linguistics, logic, business analytics, and statistics, among others. One area where similarity measures are particularly important is in the analysis of spatiotemporal event sequences and associated environs or settings. This dissertation focuses on developing a framework of modeling, representation, and new similarity measure construction for sequences of spatiotemporal events and corresponding settings, which can be applied to different event data types and used in different areas of data science. The first core part of this dissertation presents a matrix-based spatiotemporal event sequence representation that unifies punctual and interval-based representation of events. This framework supports different event data types and provides support for data mining and sequence classification and clustering. The similarity measure is based on the modified Jaccard index with temporal order constraints and accommodates different event data types. This approach is demonstrated through simulated data examples and the performance of the similarity measures is evaluated with a k-nearest neighbor algorithm (k-NN) classification test on synthetic datasets. These similarity measures are incorporated into a clustering method and successfully demonstrate the usefulness in a case study analysis of event sequences extracted from space time series of a water quality monitoring system. This dissertation further proposes a new similarity measure for event setting sequences, which involve the space and time in which events occur. While similarity measures for spatiotemporal event sequences have been studied, the settings and setting sequences have not yet been considered. While modeling event setting sequences, spatial and temporal scales are considered to define the bounds of the setting and incorporate dynamic variables along with static variables. Using a matrix-based representation and an extended Jaccard index, new similarity measures are developed to allow for the use of all variable data types. With these similarity measures coupled with other multivariate statistical analysis approaches, results from a case study involving setting sequences and pollution event sequences associated with the same monitoring stations, support the hypothesis that more similar spatial-temporal settings or setting sequences may generate more similar events or event sequences. To test the scalability of STES similarity measure in a larger dataset and an extended application in different fields, this dissertation compares and contrasts the prospective space-time scan statistic with the STES similarity approach for identifying COVID-19 hotspots. The COVID-19 pandemic has highlighted the importance of detecting hotspots or clusters of COVID-19 to provide decision makers at various levels with better information for managing distribution of human and technical resources as the outbreak in the USA continues to grow. The prospective space-time scan statistic has been used to help identify emerging disease clusters yet results from this approach can encounter strategic limitations imposed by the spatial constraints of the scanning window. The STES-based approach adapted for this pandemic context computes the similarity of evolving normalized COVID-19 daily cases by county and clusters these to identify counties with similarly evolving COVID-19 case histories. This dissertation analyzes the spread of COVID-19 within the continental US through four periods beginning from late January 2020 using the COVID-19 datasets maintained by John Hopkins University, Center for Systems Science and Engineering (CSSE). Results of the two approaches can complement with each other and taken together can aid in tracking the progression of the pandemic. Overall, the dissertation highlights the importance of developing similarity measures for analyzing spatiotemporal event sequences and associated settings, which can be applied to different event data types and used for data mining, sequence classification, and clustering.

Download

Share