Date of Award

12-1-2020

Language

English

Document Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

College/School/Department

Department of Computer Science

Content Description

1 online resource (viii, 127 pages) : illustrations (some color)

Dissertation/Thesis Chair

Petko Bogdanov

Committee Members

Charalampos Chelmis, Feng Chen

Keywords

network analysis, optimization, time series analysis, Data mining, Time-series analysis, System analysis, Electronic data processing

Subject Categories

Computer Sciences

Abstract

The rate at which data is generated in modern applications has created an unprecedented demand for novel methods to effectively and efficiently extract insightful patterns. Methods aware of known domain-specific structure in the data tend to be advantageous. In particular, a joint temporal and networked view of observations offers a holistic lens to many real-world systems. Example domains abound: activity of social network users, gene interactions over time, a temporal load of infrastructure networks, and others. Existing analysis and mining approaches for such data exhibit limited quality and scalability due to their sensitivity to noise, missing observations, and the need to explore a large search space. In this talk, I will discuss my thesis work on novel data mining algorithms designed to take advantage of known structure in networked samples, time series, and evolving networks. For networked samples, I propose methods to learn discriminative subgraphs in both supervised and unsupervised settings. I will also discuss several solutions employing temporal structures, such as period- icity and smoothness, in time series with implications on downstream applications such as forecasting and anomaly detection. I will also discuss techniques for mining communities in dynamic networks by fusing the notions of network and temporal structures.

Download

Share

COinS