Deep Probabilistic Modeling for Coupled Multivariate Time Series

Yang, Yang

Deep Probabilistic Modeling for Coupled Multivariate Time Series

Yang, Yang

Permalink

Publication Type:: Thesis
Issue Date:: 2023

Open Access

Copyright Clearance Process

Recently Added
In Progress
Open Access

This item is open access.

Adobe PDF

Download thesisAdobe PDF (6.15 MB)

View statistics

Full metadata record

Field	Value	Language
dc.contributor.author	Yang, Yang
dc.date.accessioned	2023-10-03T21:59:18Z
dc.date.available	2023-10-03T21:59:18Z
dc.date.issued	2023
dc.identifier.uri	http://hdl.handle.net/10453/172448
dc.description	University of Technology Sydney. Faculty of Engineering and Information Technology.	en_US.UTF-8
dc.description.abstract	Some complex intelligent systems such as for tackling the COVID-19 pandemic involve coupled Multivariate Time Series (MTSs), where both target variables (such as COVID-19 infected, confirmed, and recovered cases) and external factors (such as virus mutation and infectivity, vaccination, and government intervention influence) are coupled. Forecasting such MTSs with multiple external factors needs to model interactions within and between MTSs and handle their uncertainty, heterogeneity, and dynamics. However, existing shallow to deep MTSs modelers, including regressors, deep recurrent neural networks such as DeepAR, deep state space models, and deep factor models cannot jointly characterize these issues in a probabilistic manner across coupled MTSs. Therefore, it raises two main research problems: (1) How to conduct robust probabilistic forecasting for COVID-19 using coupled MTSs with multiple external factors? (2) how to explicitly model intra- and inter-MTS couplings and effectively handle volatile covariates of coupled MTSs? To tackle the first problem, Chapter 3 proposes an end-to-end deep probabilistic cross-MTS learning network (MTSNet). It incorporates a tensor input consisting of scaled targeting and external MTSs. It then vertically and horizontally stacks long-short memory networks for encoding and decoding target MTSs and enhances uncertainty modeling, generalization, and forecasting robustness by residual connection, variational zoneout, and probabilistic forecasting. The tensor input is projected to a probability distribution for target MTS forecasting. MTSNet outperforms the State-of-the-Art (SOTA) deep probabilistic MTS networks in forecasting COVID-19 confirmed cases and Intensive Care Unit (ICU) patient numbers for six countries by involving virus mutation, vaccination, government interventions, and infectivity. To tackle the second problem, Chapter 4 proposes Deep Spectral Copula Mechanisms (DSCM). Specifically, DSCM incorporates a Singular Spectral Analysis (SSA) module to reduce the volatility of multiple covariates. It applies an intra-MTS coupling module to explicitly model the temporal couplings within a single set of multivariate time series and transforms target variables into joint probability distributions via Gaussian copula transformation to establish inter-MTS couplings across multiple multivariate time series. Substantial experiments on COVID-19 time-series data from multiple countries indicate the superiority of DSCM over state-of-the-art approaches.	en_US.UTF-8
dc.format	Thesis (MAnalytics)
dc.language.iso	en_US	en_US.UTF-8
dc.relation	https://opus.lib.uts.edu.au/bitstream/10453/172448/1/thesis.pdf
dc.rights	info:eu-repo/semantics/openAccess
dc.rights	The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.
dc.rights	© 2023 Yang Yang
dc.rights	au.edu.uts.lib/nph
dc.title	Deep Probabilistic Modeling for Coupled Multivariate Time Series	en_US.UTF-8
dc.type	Thesis
utslib.copyright.status	open_access	*

Abstract:

Some complex intelligent systems such as for tackling the COVID-19 pandemic involve coupled Multivariate Time Series (MTSs), where both target variables (such as COVID-19 infected, confirmed, and recovered cases) and external factors (such as virus mutation and infectivity, vaccination, and government intervention influence) are coupled. Forecasting such MTSs with multiple external factors needs to model interactions within and between MTSs and handle their uncertainty, heterogeneity, and dynamics. However, existing shallow to deep MTSs modelers, including regressors, deep recurrent neural networks such as DeepAR, deep state space models, and deep factor models cannot jointly characterize these issues in a probabilistic manner across coupled MTSs. Therefore, it raises two main research problems: (1) How to conduct robust probabilistic forecasting for COVID-19 using coupled MTSs with multiple external factors? (2) how to explicitly model intra- and inter-MTS couplings and effectively handle volatile covariates of coupled MTSs? To tackle the first problem, Chapter 3 proposes an end-to-end deep probabilistic cross-MTS learning network (MTSNet). It incorporates a tensor input consisting of scaled targeting and external MTSs. It then vertically and horizontally stacks long-short memory networks for encoding and decoding target MTSs and enhances uncertainty modeling, generalization, and forecasting robustness by residual connection, variational zoneout, and probabilistic forecasting. The tensor input is projected to a probability distribution for target MTS forecasting. MTSNet outperforms the State-of-the-Art (SOTA) deep probabilistic MTS networks in forecasting COVID-19 confirmed cases and Intensive Care Unit (ICU) patient numbers for six countries by involving virus mutation, vaccination, government interventions, and infectivity. To tackle the second problem, Chapter 4 proposes Deep Spectral Copula Mechanisms (DSCM). Specifically, DSCM incorporates a Singular Spectral Analysis (SSA) module to reduce the volatility of multiple covariates. It applies an intra-MTS coupling module to explicitly model the temporal couplings within a single set of multivariate time series and transforms target variables into joint probability distributions via Gaussian copula transformation to establish inter-MTS couplings across multiple multivariate time series. Substantial experiments on COVID-19 time-series data from multiple countries indicate the superiority of DSCM over state-of-the-art approaches.

Please use this identifier to cite or link to this item:

http://hdl.handle.net/10453/172448