Uncovering hidden spatial patterns by hidden Markov model

Ruihong Huang, Christina Kennedy

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

1 Scopus citations


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 publicationGeographic Information Science - 5th International Conference, GIScience 2008, Proceedings
Number of pages20
ISBN (Print)3540874720, 9783540874720
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)0302-9743
ISSN (Electronic)1611-3349


Other5th International Conference on Geographic Information Science, GIScience 2008
Country/TerritoryUnited States
CityPark City, UT


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

ASJC Scopus subject areas

  • Theoretical Computer Science
  • General Computer Science


Dive into the research topics of 'Uncovering hidden spatial patterns by hidden Markov model'. Together they form a unique fingerprint.

Cite this