Verification of novel integrations of swarm intelligence algorithms into deep learning neural network for flood susceptibility mapping

•Examination of hybrid models that combine several swarm intelligenceoptimizers with deep learning neural network.•Comparison of performance of proposed models with benchmarked classifiers and algorithms.•Potential application of proposed models in flood susceptibility mapping is investigated. This...

Full description

Saved in:
Bibliographic Details
Published in:Journal of hydrology (Amsterdam) Vol. 581; p. 124379
Main Authors: Bui, Quang-Thanh, Nguyen, Quoc-Huy, Nguyen, Xuan Linh, Pham, Vu Dong, Nguyen, Huu Duy, Pham, Van-Manh
Format: Journal Article
Language:English
Published: Elsevier B.V 01.02.2020
Subjects:
case studies
climatic factors
data collection
Deep Learning Neural Network
digital elevation models
Flash flood
hydrology
mountains
normalized difference vegetation index
rain
remote sensing
rivers
statistical analysis
streams
support vector machines
Swarm Intelligence Optimization Algorithm
topography
Vietnam
Vietnam
Deep Learning Neural Network
Swarm Intelligence Optimization Algorithm
Flash flood
ISSN:0022-1694, 1879-2707
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•Examination of hybrid models that combine several swarm intelligenceoptimizers with deep learning neural network.•Comparison of performance of proposed models with benchmarked classifiers and algorithms.•Potential application of proposed models in flood susceptibility mapping is investigated. This study proposed and compared several novel hybrid models that combined swarm intelligence algorithms and Deep Learning Neural Network for flood susceptibility mapping. Lai Chau, a province in the northwest mountainous region of Vietnam was chosen as a case study since it had recently undergone severe flashflood in 2018. For this purpose, numerical predictor variables such as topographically derived factors (Digital Elevation Model, Aspect, Slope, Curvature, Topographic Wetness Index), climatic variables (Rain), and hydrological variables (stream density, stream power index, distance to river) and multiple remote sensing indices (Normalized Difference Vegetation Index, Normalized Difference Buildup Index) were used. These predictor variables were selected because they are globally collectible and reproducible. The performances of these models were evaluated by using common statistical indicators, namely Root Mean Square Error, Mean Absolute Error, Overall Accuracy and Area under Receiving Operating Characteristics, and the statistical test of differences. The results showed that the proposed swarm intelligence models outperformed benchmarked methods, namely Particle Swarm Optimization, Support Vector Machine, Random Forest in almost all comparing indicators. It is suggested that proposed models are more robust than the classifiers, which were used for benchmarking and they are good alternatives for flood susceptibility mapping given the availability of dataset.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 23
ISSN:0022-1694
1879-2707
DOI:10.1016/j.jhydrol.2019.124379