eprintid: 5794
rev_number: 2
eprint_status: archive
userid: 1
dir: disk0/00/00/57/94
datestamp: 2023-11-09 16:17:32
lastmod: 2023-11-09 16:17:32
status_changed: 2023-11-09 16:03:54
type: article
metadata_visibility: show
creators_name: Siddiqui, F.I.
creators_name: Pathan, D.M.
creators_name: Osman, S.B.A.B.S.
creators_name: Pinjaro, M.A.
creators_name: Memon, S.
title: Comparison between regression and ANN models for relationship of soil properties and electrical resistivity
ispublished: pub
keywords: algorithm; artificial neural network; comparative study; cost analysis; data set; electrical resistivity; geophysical method; numerical model; prediction; regression analysis; soil property; soil structure
note: cited By 12
abstract: Precise determination of engineering properties of soil is essential for proper design and successful construction of any structure. The conventional methods for determination of engineering properties are invasive, costly, and time-consuming. Geoelectrical survey is a very attractive tool for delineating subsurface properties without soil disturbance. Proper correlations of various soil parameters with electrical resistivity of soil will bridge the gap between geotechnical and geophysical engineering and also enable geotechnical engineers to estimate geotechnical parameters from electrical resistivity data. The regression models of relationship between electrical resistivity and various soil properties used in the current research for the purpose of comparison with artificial neural network (ANN) models were adopted from the work of Siddiqui and Osman (Environ Earth Sci 70:259�26, 2013). In order to obtain better relationships, ANN modeling was done using same data as regression analysis. The neural network models were trained using single input (electrical resistivity) and single output (i.e., moisture content, plasticity index, and friction angle). Twenty (20) multilayer feedforward (MLFF) networks were developed for each properties, ten (10) each for two different learning algorithms, Levenberg-Marquardt (LM) and scaled conjugate gradient (SCG). The numbers of neurons in hidden layer were experimented from 1 to 10. Best network with particular learning algorithm and optimum number of neuron in hidden layer presenting lowest root mean square error (RMSE) was selected for prediction of various soil properties. ANN models show better prediction results for all soil properties. © 2014, Saudi Society for Geosciences.
date: 2015
publisher: Springer Verlag
official_url: https://www.scopus.com/inward/record.uri?eid=2-s2.0-84939573680&doi=10.1007%2fs12517-014-1637-y&partnerID=40&md5=41f8e4ccd080cb8c9300c44d59e74fbf
id_number: 10.1007/s12517-014-1637-y
full_text_status: none
publication: Arabian Journal of Geosciences
volume: 8
number: 8
pagerange: 6145-6155
refereed: TRUE
issn: 18667511
citation:   Siddiqui, F.I. and Pathan, D.M. and Osman, S.B.A.B.S. and Pinjaro, M.A. and Memon, S.  (2015) Comparison between regression and ANN models for relationship of soil properties and electrical resistivity.  Arabian Journal of Geosciences, 8 (8).  pp. 6145-6155.  ISSN 18667511