Date of Award

5-1-2021

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School/Department

Department of Computer Science

Content Description

1 online resource (vii, 85 pages) : illustrations (some color), map.

Dissertation/Thesis Chair

Petko Bogdanov

Committee Members

Evangelos Papalexakis, Mariya Zheleva, Charalampos Chelmis

Keywords

Data Mining, Dynamic Networks, Machine Learning, Data mining, Time-series analysis, Temporal databases, Spatial data mining

Subject Categories

Computer Sciences

Abstract

A variety of dynamic systems can be broken down into potentially overlapping subcomponents with varying temporal behavior, ranging from communities in networks, to clusters of trajectories in spatiotemporal data, to co-evolving subsets within multivariate time series. Using explicit regularization on various temporal behaviors within a tensor factorizationframework, I demonstrate means to mine these subgroups along with their temporal activities, as well as how that yields information about the overall systems. Additionally, I adapt this notion of temporal communities to the spatiotemporal setting to develop a reinforcement learning approach for optimizing co-ordinated communication between independent agents.

Download

Share

COinS