Uncovering hidden spatial patterns by hidden Markov model

Ruihong Huang, Christina Kennedy

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

Many spatial data mining and spatial modeling approaches use Euclidean distance in modeling spatial dependence. Although meaningful and convenient, Euclidean distance has weaknesses. These include providing an over simplified representation of spatial dependence, being limited to certain spatial pattern and symmetrical relationships, being unable to account for cross-class dependencies, and unable to work with categorical especially multinomial data. This paper introduces Hidden Markov Model (HMM) as an attractive approach to uncovering hidden spatial patterns. The HMM assumes that a hidden state (factor or process) generates observable symbols (indicators). This doubly embedded stochastic approach uncovers hidden states based on observed symbol sequences using two integrated sets of probabilities, transition probability and emission probability. As an alternative to Euclidean distance based approaches, the HMM measures spatial dependency by transition probabilities and cross-class correlation better capturing geographic context. HMM works with data of any measurement scale and dimension. To demonstrate the method, we assume urban spatial structure as a hidden spatial factor underlying single family housing unit prices in Milwaukee, Wisconsin, we then use the HMM to uncover four hidden spatial states from home sale prices.

Original languageEnglish (US)
Title of host publicationLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Pages70-89
Number of pages20
Volume5266 LNCS
DOIs
StatePublished - 2008
Event5th International Conference on Geographic Information Science, GIScience 2008 - Park City, UT, United States
Duration: Sep 23 2008Sep 26 2008

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume5266 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Other

Other5th International Conference on Geographic Information Science, GIScience 2008
CountryUnited States
CityPark City, UT
Period9/23/089/26/08

Fingerprint

Spatial Pattern
Hidden Markov models
Markov Model
Euclidean Distance
Spatial Dependence
Transition Probability
Sale price
Spatial Data Mining
Spatial Modeling
Spatial Structure
Set theory
Categorical
Data mining
Sales
Unit
Alternatives
Modeling
Demonstrate
Class

Keywords

  • Data mining
  • GIS
  • Hidden Markov model
  • Spatial modeling

ASJC Scopus subject areas

  • Computer Science(all)
  • Theoretical Computer Science

Cite this

Huang, R., & Kennedy, C. (2008). Uncovering hidden spatial patterns by hidden Markov model. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 5266 LNCS, pp. 70-89). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5266 LNCS). https://doi.org/10.1007/978-3-540-87473-7-5

Uncovering hidden spatial patterns by hidden Markov model. / Huang, Ruihong; Kennedy, Christina.

Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5266 LNCS 2008. p. 70-89 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 5266 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Huang, R & Kennedy, C 2008, Uncovering hidden spatial patterns by hidden Markov model. in Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). vol. 5266 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 5266 LNCS, pp. 70-89, 5th International Conference on Geographic Information Science, GIScience 2008, Park City, UT, United States, 9/23/08. https://doi.org/10.1007/978-3-540-87473-7-5
Huang R, Kennedy C. Uncovering hidden spatial patterns by hidden Markov model. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5266 LNCS. 2008. p. 70-89. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-540-87473-7-5
Huang, Ruihong ; Kennedy, Christina. / Uncovering hidden spatial patterns by hidden Markov model. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Vol. 5266 LNCS 2008. pp. 70-89 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{dd76d180092c45f2810d05f9f24471f3,
title = "Uncovering hidden spatial patterns by hidden Markov model",
abstract = "Many spatial data mining and spatial modeling approaches use Euclidean distance in modeling spatial dependence. Although meaningful and convenient, Euclidean distance has weaknesses. These include providing an over simplified representation of spatial dependence, being limited to certain spatial pattern and symmetrical relationships, being unable to account for cross-class dependencies, and unable to work with categorical especially multinomial data. This paper introduces Hidden Markov Model (HMM) as an attractive approach to uncovering hidden spatial patterns. The HMM assumes that a hidden state (factor or process) generates observable symbols (indicators). This doubly embedded stochastic approach uncovers hidden states based on observed symbol sequences using two integrated sets of probabilities, transition probability and emission probability. As an alternative to Euclidean distance based approaches, the HMM measures spatial dependency by transition probabilities and cross-class correlation better capturing geographic context. HMM works with data of any measurement scale and dimension. To demonstrate the method, we assume urban spatial structure as a hidden spatial factor underlying single family housing unit prices in Milwaukee, Wisconsin, we then use the HMM to uncover four hidden spatial states from home sale prices.",
keywords = "Data mining, GIS, Hidden Markov model, Spatial modeling",
author = "Ruihong Huang and Christina Kennedy",
year = "2008",
doi = "10.1007/978-3-540-87473-7-5",
language = "English (US)",
isbn = "3540874720",
volume = "5266 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
pages = "70--89",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

}

TY - GEN

T1 - Uncovering hidden spatial patterns by hidden Markov model

AU - Huang, Ruihong

AU - Kennedy, Christina

PY - 2008

Y1 - 2008

N2 - Many spatial data mining and spatial modeling approaches use Euclidean distance in modeling spatial dependence. Although meaningful and convenient, Euclidean distance has weaknesses. These include providing an over simplified representation of spatial dependence, being limited to certain spatial pattern and symmetrical relationships, being unable to account for cross-class dependencies, and unable to work with categorical especially multinomial data. This paper introduces Hidden Markov Model (HMM) as an attractive approach to uncovering hidden spatial patterns. The HMM assumes that a hidden state (factor or process) generates observable symbols (indicators). This doubly embedded stochastic approach uncovers hidden states based on observed symbol sequences using two integrated sets of probabilities, transition probability and emission probability. As an alternative to Euclidean distance based approaches, the HMM measures spatial dependency by transition probabilities and cross-class correlation better capturing geographic context. HMM works with data of any measurement scale and dimension. To demonstrate the method, we assume urban spatial structure as a hidden spatial factor underlying single family housing unit prices in Milwaukee, Wisconsin, we then use the HMM to uncover four hidden spatial states from home sale prices.

AB - Many spatial data mining and spatial modeling approaches use Euclidean distance in modeling spatial dependence. Although meaningful and convenient, Euclidean distance has weaknesses. These include providing an over simplified representation of spatial dependence, being limited to certain spatial pattern and symmetrical relationships, being unable to account for cross-class dependencies, and unable to work with categorical especially multinomial data. This paper introduces Hidden Markov Model (HMM) as an attractive approach to uncovering hidden spatial patterns. The HMM assumes that a hidden state (factor or process) generates observable symbols (indicators). This doubly embedded stochastic approach uncovers hidden states based on observed symbol sequences using two integrated sets of probabilities, transition probability and emission probability. As an alternative to Euclidean distance based approaches, the HMM measures spatial dependency by transition probabilities and cross-class correlation better capturing geographic context. HMM works with data of any measurement scale and dimension. To demonstrate the method, we assume urban spatial structure as a hidden spatial factor underlying single family housing unit prices in Milwaukee, Wisconsin, we then use the HMM to uncover four hidden spatial states from home sale prices.

KW - Data mining

KW - GIS

KW - Hidden Markov model

KW - Spatial modeling

UR - http://www.scopus.com/inward/record.url?scp=56749181990&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=56749181990&partnerID=8YFLogxK

U2 - 10.1007/978-3-540-87473-7-5

DO - 10.1007/978-3-540-87473-7-5

M3 - Conference contribution

SN - 3540874720

SN - 9783540874720

VL - 5266 LNCS

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 70

EP - 89

BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

ER -