Design and development of an intelligent extension for mapping landslide susceptibility using artificial neural network

Abstract

Amongst the indirect and quantitative methods which have been propounded for assessing landslide susceptibility, artificial neural network and especially multilayer perceptron dominated research activities. It is due to its high power in solving nonlinear separable problems and the capability of generalization. This study deals with designing a model for systematic usage of multilayer perceptron network to solve existing challenges on choosing input patterns and target outputs of spatial data. This model accompanies with a modified Backpropagation (BP) as a learning algorithm. The designed model was applied to create an extension in ArcGIS® in order to reach an intelligent decision-making tool. The Landslide Susceptibility Map (LSM) was then generated for an expanded landscape in Mazandaran, Iran, using the extension while the landslide's data and criteria maps were produced in small-scale. Statistical results of landslides which were happened in different domains of susceptibility showed the overall accuracy equivalent to 98.2% in hazard approximating. © 2009 Springer Berlin Heidelberg.