Landslide susceptibility mapping using Naïve Bayes and Bayesian network models in Umyeonsan, Korea

Sunmin Lee, Moung Jin Lee, Hyung Sup Jung, Saro Lee

Research output: Contribution to journalArticlepeer-review

41 Scopus citations

Abstract

In recent years, machine learning techniques have been increasingly applied to the assessment of various natural disasters, including landslides and floods. Machine learning techniques can be used to make predictions based on the relationships among events and their influencing factors. In this study, a machine learning approaches were applied based on landslide location data in a geographic information system environment. Topographic maps were used to determine the topographical factors. Additional soil and forest parameters were examined using information obtained from soil and forest maps. A total of 17 factors affecting landslide occurrence were selected and a spatial database was constructed. Naïve Bayes and Bayesian network models were applied to predict landslides based on selected risk factors. The two models showed accuracies of 78.3 and 79.8%, respectively. The results of this study provide a useful foundation for effective strategies to prevent and manage landslides in urban areas.

Original languageEnglish
JournalGeocarto International
DOIs
StateAccepted/In press - 2019

Keywords

  • Bayesian Network
  • GIS
  • Landslide susceptibility
  • Naïve Bayes
  • Umyeonsan

Fingerprint

Dive into the research topics of 'Landslide susceptibility mapping using Naïve Bayes and Bayesian network models in Umyeonsan, Korea'. Together they form a unique fingerprint.

Cite this