Methods to window data to differentiate between markov models

Jason M. Schwier, Richard R. Brooks, Christopher Griffin

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

In this paper, we consider how we can detect patterns in data streams that are serial Markovian, where target behaviors are Markovian, but targets may switch from one Markovian behavior to another. We want to reliably and promptly detect behavior changes. Traditional Markov-model-based pattern detection approaches, such as hidden Markov models, use maximum likelihood techniques over the entire data stream to detect behaviors. To detect changes between behaviors, we use statistical pattern matching calculations performed on a sliding window of data samples. If the window size is very small, the system will suffer from excessive false-positive rates. If the window is very large, change-point detection is delayed. This paper finds both necessary and sufficient bounds on the window size. We present two methods of calculating window sizes based on the state and transition structures of the Markov models. Two application examples are presented to verify our results. Our first example problem uses simulations to illustrate the utility of the proposed approaches. The second example uses models extracted from a database of consumer purchases to illustrate their use in a real application.

Original languageEnglish (US)
Article number5593913
Pages (from-to)650-663
Number of pages14
JournalIEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics
Volume41
Issue number3
DOIs
StatePublished - Jun 1 2011

Fingerprint

Databases
Pattern matching
Hidden Markov models
Maximum likelihood
Switches

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Software
  • Medicine(all)
  • Information Systems
  • Human-Computer Interaction
  • Computer Science Applications
  • Electrical and Electronic Engineering

Cite this

@article{609b62e792e342bcaf3c396c7246d010,
title = "Methods to window data to differentiate between markov models",
abstract = "In this paper, we consider how we can detect patterns in data streams that are serial Markovian, where target behaviors are Markovian, but targets may switch from one Markovian behavior to another. We want to reliably and promptly detect behavior changes. Traditional Markov-model-based pattern detection approaches, such as hidden Markov models, use maximum likelihood techniques over the entire data stream to detect behaviors. To detect changes between behaviors, we use statistical pattern matching calculations performed on a sliding window of data samples. If the window size is very small, the system will suffer from excessive false-positive rates. If the window is very large, change-point detection is delayed. This paper finds both necessary and sufficient bounds on the window size. We present two methods of calculating window sizes based on the state and transition structures of the Markov models. Two application examples are presented to verify our results. Our first example problem uses simulations to illustrate the utility of the proposed approaches. The second example uses models extracted from a database of consumer purchases to illustrate their use in a real application.",
author = "Schwier, {Jason M.} and Brooks, {Richard R.} and Christopher Griffin",
year = "2011",
month = "6",
day = "1",
doi = "10.1109/TSMCB.2010.2076325",
language = "English (US)",
volume = "41",
pages = "650--663",
journal = "IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics",
issn = "1083-4419",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
number = "3",

}

Methods to window data to differentiate between markov models. / Schwier, Jason M.; Brooks, Richard R.; Griffin, Christopher.

In: IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, Vol. 41, No. 3, 5593913, 01.06.2011, p. 650-663.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Methods to window data to differentiate between markov models

AU - Schwier, Jason M.

AU - Brooks, Richard R.

AU - Griffin, Christopher

PY - 2011/6/1

Y1 - 2011/6/1

N2 - In this paper, we consider how we can detect patterns in data streams that are serial Markovian, where target behaviors are Markovian, but targets may switch from one Markovian behavior to another. We want to reliably and promptly detect behavior changes. Traditional Markov-model-based pattern detection approaches, such as hidden Markov models, use maximum likelihood techniques over the entire data stream to detect behaviors. To detect changes between behaviors, we use statistical pattern matching calculations performed on a sliding window of data samples. If the window size is very small, the system will suffer from excessive false-positive rates. If the window is very large, change-point detection is delayed. This paper finds both necessary and sufficient bounds on the window size. We present two methods of calculating window sizes based on the state and transition structures of the Markov models. Two application examples are presented to verify our results. Our first example problem uses simulations to illustrate the utility of the proposed approaches. The second example uses models extracted from a database of consumer purchases to illustrate their use in a real application.

AB - In this paper, we consider how we can detect patterns in data streams that are serial Markovian, where target behaviors are Markovian, but targets may switch from one Markovian behavior to another. We want to reliably and promptly detect behavior changes. Traditional Markov-model-based pattern detection approaches, such as hidden Markov models, use maximum likelihood techniques over the entire data stream to detect behaviors. To detect changes between behaviors, we use statistical pattern matching calculations performed on a sliding window of data samples. If the window size is very small, the system will suffer from excessive false-positive rates. If the window is very large, change-point detection is delayed. This paper finds both necessary and sufficient bounds on the window size. We present two methods of calculating window sizes based on the state and transition structures of the Markov models. Two application examples are presented to verify our results. Our first example problem uses simulations to illustrate the utility of the proposed approaches. The second example uses models extracted from a database of consumer purchases to illustrate their use in a real application.

UR - http://www.scopus.com/inward/record.url?scp=79957475703&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=79957475703&partnerID=8YFLogxK

U2 - 10.1109/TSMCB.2010.2076325

DO - 10.1109/TSMCB.2010.2076325

M3 - Article

C2 - 20923740

AN - SCOPUS:79957475703

VL - 41

SP - 650

EP - 663

JO - IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics

JF - IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics

SN - 1083-4419

IS - 3

M1 - 5593913

ER -