Comparison of Neuro-Fuzzy, Genetic Algorithm, Artificial Neural Network and Multivariate Regression for Prediction of Soil Salinity (Case study: Ardakan City)

Document Type : Research Paper

Authors

1 Assistant Professor Faculty of Agriculture & Natural Resources, University of Ardakan

2 Assistant Professor Department of Soil Science, University College of Agriculture & Natural Resources, University of Tehran

3 Professor Department of Soil Science, University College of Agriculture & Natural Resources, University of Tehran

4 Professor Faculty of Agricultural Engineering & Technology, University of Tehran

5 Assistant Professor Research Institute of Agriculture & Natural Resources, Isfahan

6 Academic Member of Soil Salinity National Center, Yazd

7 GIS & RS Expert of Soil Salinity National Center, Yazd

Abstract

In recent years, alternative methods have been used for estimation of soil salinity. Therefore, at present research, 600 soil samples collected from Ardakan in central Iran. Then EM38 and terrain parameters such as wetness index, land index and curvature as readily measured properties and soil salinity (0-30 and 0-100) as predicted variables were measured. After that, the data set was divided into two subsets for calibration (80%) and testing (20%) of the models. For predicting of mentioned parameters, ANFIS, GA, ANNs and MLR were applied. In order to evaluate models, some evaluation parameters such as root mean square, average error, average standard error and coefficient of determination were used. Results showed that the ANFIS model gives better estimation than the other techniques for all characteristics whereas this model increased accuracy of predictions about 17 and 11% for EC30 and EC100 respectability. After ANFIS model, GA and ANN had better accuracy than multivariate regression.

Keywords

Files

History

  • Receive Date: 23 September 2011
  • Revise Date: 19 August 2013
  • Accept Date: 17 April 2012
  • First Publish Date: 22 June 2013

Share

How to cite

Statistics