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

Methods for event time series prediction and anomaly detection

Liu, Siqi (2020) Methods for event time series prediction and anomaly detection. Doctoral Dissertation, University of Pittsburgh. (Unpublished)

[img]
Preview
PDF
Primary Text

Download (1MB) | Preview

Abstract

Event time series are sequences of events occurring in continuous time. They arise in many real-world problems and may represent, for example, posts in social media, administrations of medications to patients, or adverse events, such as episodes of atrial fibrillation or earthquakes. In this work, we study and develop methods for prediction and anomaly detection on event time series. We study two general approaches. The first approach converts event time series to regular time series of counts via time discretization. We develop methods relying on (a) nonparametric time series decomposition and (b) dynamic linear models for regular time series. The second approach models the events in continuous time directly. We develop methods relying on point processes. For prediction, we develop a new model based on point processes to combine the advantages of existing models. It is flexible enough to capture complex dependency structures between events, while not sacrificing applicability in common scenarios. For anomaly detection, we develop methods that can detect new types of anomalies in continuous time and that show advantages compared to time discretization.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Liu, Siqisiqiliu@cs.pitt.edu
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairHauskrecht, Milosmilos@pitt.edu
Committee MemberKovashka, Adrianakovashka@cs.pitt.edu
Committee MemberLitman, Dianedlitman@pitt.edu
Committee MemberPoczos, Barnabasbapoczos@cs.cmu.edu
Date: 20 August 2020
Date Type: Publication
Defense Date: 22 June 2020
Approval Date: 20 August 2020
Submission Date: 4 August 2020
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 143
Institution: University of Pittsburgh
Schools and Programs: School of Computing and Information > Computer Science
Degree: PhD - Doctor of Philosophy
Thesis Type: Doctoral Dissertation
Refereed: Yes
Uncontrolled Keywords: Time Series; Event Sequence; Anomaly Detection; Probabilistic Model; Bayesian Inference; Nonparametric; Point Process; Gaussian Process
Date Deposited: 20 Aug 2020 18:43
Last Modified: 20 Aug 2020 18:43
URI: http://d-scholarship.pitt.edu/id/eprint/39469

Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item