Back to search

Cornell University

Time Series, Clustering, Network Models and Tensor Analysis

Abstract

dc:description.abstract

High dimensional data comes in many forms and with modern technologies we are able to capture and store ever larger amounts of data. These increasing amounts of high dimensional data lead to a need for statistical methods which are able analyze and digest such data in a computationally feasible manner. Toward this end, this thesis develops methods for analyzing some of the types of high dimensional data that are observed in practice: time series, networks, and higher-order tensors with an emphasis on clustering methods applied to these data types.First we look at analyzing corpora of time series data. We develop a model-based approach to clustering time series data using Autoregressive Moving Average (ARMA) models. Using these models we are able to not only identify groups of related time series, but also develop goodness of fit tests for models applied to the resulting clusters. Next, we look at analyzing sets of networks. To analyze these, we apply Hutchinson Trace Estimation to estimate quantiles of the spectrum of a graph to use as an embedding. With these embeddings we show how sets of graphs can be clustered using both the K-means algorithm and dynamic time warping. Finally, we look at analyzing higher-order tensor data via a decomposition which interpolates between the CP-decomposition and the Tucker decomposition. This decomposition, which we call the grouped-tucker decomposition, groups together the different dimensions of the tensor and allows for different behavior within the groups and between the groups. We also develop a method for identifying the groups via a hierarchical clustering inspired algorithm.

Degree

thesis:*
Name thesis:degree_name
Ph. D., Statistics
Level thesis:degree_level
Doctor of Philosophy
Discipline thesis:degree_discipline
Statistics
Grantor
Cornell University
Year dc:date.issued
2025

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Hoare, Derek
Chair dc:contributor.chair
  • Matteson, David
Committee members dc:contributor.committeemember
  • Wells, Martin
  • Basu, Sumanta

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language.iso
en

Identifiers

dc:identifier.*
Dc Identifier Bibid
17423419
Dc Identifier Other
ProQuest Submission ID: 15267
ProQuest Publication ID: 32172812

Chain of custody

source
Harvested from
Cornell University
Base URL
ecommons.cornell.edu/server/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
citation

Hoare, Derek. Time Series, Clustering, Network Models and Tensor Analysis. Doctor of Philosophy thesis, Cornell University, 2025. https://hdl.handle.net/1813/121048