Preparing of capability map for road construction using artificial neural network and GIS (case study: Arasbaran area)

The aim of the study was to provide an intelligent artificial neural networks-based method for modeling the capability of Arasbaran protected area for road crossing, in order to design, modify, and appropriate development of existing road network and communication routs in the region. First, using Analytical Hierarchy Process (AHP) and Weighted Linear Combination (WLC) method, and utilization the effective informative layers on routing, the suitability map of road construction was prepared to provide training samples in ArcGIS. In the following, Multilayer Perceptron (MLP) network was used to estimate the suitability value of road crossing. In order to evaluate the neural network’s model performance, the results were compared with the results of multivariate linear regression. According to the results, artificial neural network and statistical method of regression were shown to be useful in determining the suitability value of road crossing with coefficient of determination (R2) 0.908 and 0.901, root mean squared error (RMSE) 0.0385 and 0.04, respectively. Neural network results were relatively better than regression. Also, according to the results of sensitivity analysis of input variables, four criteria of slope, bedrock, erosion susceptibility, and soil texture showed the highest influence in estimating the model, respectively.