Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

A Deep Learning Based Approach for Time-series Modeling

Alghamdi, Dhaifallah (2022) A Deep Learning Based Approach for Time-series Modeling. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

This is the latest version of this item.

[img]
Preview
PDF
Download (13MB) | Preview

Abstract

Big data has evolved as a new research domain in the digital era in which we live today. This domain deals with the study of huge datasets with numerous different features, whose volumes are rapidly snowballing with time. These types of datasets can be produced by different autonomous sources, including scientific experiments, engineering applications, government records, financial transactions, etc. The availability of big data is of a great value because of the opportunity this provides for making better-informed decisions, but it also requires advanced analytical tools to derive important insights for these decisions. This is the main reason that artificial intelligence (AI) and machine learning (ML) have gained immense popularity in recent years.
Time-series forecasting is an important application area for machine learning. It is important because there are so many prediction problems from various application domains that involve a time component. However, the temporal dimension also makes time-series problems more challenging to handle as opposed to many other prediction tasks. For this purpose, the goal of this dissertation is to design end-to-end frameworks and build advanced models for time-series forecasting that are based on deep learning. The discussed frameworks in this dissertation share three important characteristics: 1) the ability to generate forecasts for multiple steps ahead in the future, 2) the ability to provide estimates of uncertainty associated with these forecasts, and 3) the flexibility to incorporate exogenous factors. Our approach is to harness the encoder-decoder architecture to learn from historical data and capture important relationships embedded in the time-series, and to then use this knowledge to generate forecasts for multiple steps in the future along with estimates on the uncertainty associated with these forecasts.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Alghamdi, Dhaifallahdha8@pitt.edudha80000-0002-1369-0042
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairRajgopal, Jayantj.rajgopal@pitt.eduj.rajgopal0000-0001-7730-8749
Committee MemberZeng, Bobzeng@pitt.edu
Committee MemberBidkhori, Hodabidkhori@pitt.edu
Committee MemberBarati, Masoudmasoud.barati@pitt.edu
Date: 6 September 2022
Date Type: Publication
Defense Date: 24 May 2022
Approval Date: 6 September 2022
Submission Date: 16 June 2022
Access Restriction: 1 year -- Restrict access to University of Pittsburgh for a period of 1 year.
Number of Pages: 145
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Industrial Engineering
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Deep Learning, Time-series
Date Deposited: 06 Sep 2022 16:12
Last Modified: 06 Sep 2023 05:15
URI: http://d-scholarship.pitt.edu/id/eprint/43239

Available Versions of this Item


Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item