@Article{ZhangQWFLWOGR:2022:ImTrMa,
author = "Zhang, Tingyu and Quevedo, Renata Pacheco and Wang, Huanyuan and
Fu, Quan and Luo, Dan and Wang, Tao and Oliveira, Guilherme Garcia
de and Guasselli, Laurindo Antonio and Renn{\'o}, Camilo
Daleles",
affiliation = "{Key Laboratory of Degraded and Unused Land Consolidation
Engineering} and {Instituto Nacional de Pesquisas Espaciais
(INPE)} and {Key Laboratory of Degraded and Unused Land
Consolidation Engineering} and Shaanxi Provincial Land Engineering
Construction Group Land Survey Planning, Design Institute Co and
Shaanxi Provincial Land Engineering Construction Group Land Survey
Planning, Design Institute Co and Shaanxi Provincial Land
Engineering Construction Group Land Survey Planning, Design
Institute Co and {Universidade Federal do Rio Grande do Sul
(UFRGS)} and {Universidade Federal do Rio Grande do Sul (UFRGS)}
and {Instituto Nacional de Pesquisas Espaciais (INPE)}",
title = "Improved tree-based machine learning algorithms combining with
bagging strategy for landslide susceptibility modeling",
journal = "Arabian Journal of Geosciences",
year = "2022",
volume = "15",
number = "2",
pages = "183",
keywords = "Landslide susceptibility · Decision tree · Logistic model tree ·
Reduced error pruning tree · Hybrid models ·, Bagging strategy.",
abstract = "Landslide is considered one of the most dangerous natural hazards.
Reasonable landslide susceptibility mapping can aid decision
makers in landslide prevention. For this reason, based on the feld
survey data of landslide in Chenggu County, Shaanxi Province,
China, 15 conditioning factors (altitude, slope, aspect, plan
curvature, profle curvature, SPI, TWI, distance to roads, distance
to rivers, distance to faults, rainfall, NDVI, soil, lithology,
and land use) were selected and quantifed by the certainty factor
index. Then, 184 landslides data were divided into training and
validation datasets according to the ratio of 7/3. Based on the
GIS platform, three hybrid tree-based models, namely decision tree
(DT), logistic model tree (LMT), and reduced error pruning tree
(REPT), were established. Additionally, the bagging method was
applied to build three baghybrid tree-based models: Bag-DT,
Bag-LMT, and Bag-REPT. Finally, the landslide susceptibility maps
were produced, and statistical indexes, seed cell area index and
the ROC curve, were used for model validation and comparison. The
results showed that the bagging method can signifcantly improve
the classifcation ability of hybrid models. Furthermore, the
BagREPT presented the best performance, with an accuracy value of
92.5%, being a suitable model for landslide susceptibility mapping
in the study area.",
doi = "10.1007/s12517-022-09488-3",
url = "http://dx.doi.org/10.1007/s12517-022-09488-3",
issn = "1866-7511",
label = "lattes: 7712719010541171 9 ZhangQWFLWOGR:2022:ImTrMa",
language = "en",
targetfile = "Zhang2022_Article_ImprovedTree-basedMachineLearn.pdf",
urlaccessdate = "25 jun. 2024"
}