Using Neural Network to Evaluate Construction Land Use Suitability

Abstract

Construction land suitability evaluation is essential for urban development decision. Back Propagation Neural Network (BPNN) is suitable for the non-linear issue. In this study, BPNN architecture has been set up, with 9 neurons of input layer and 4 of output layer. The neurons of input layer include indices related to topography, engineering geology, hydro-geology, and geo-hazard, which are determined based on analysis of Hangzhou land use conditions, suitable for Hangzhou and related urban region construction land suitability research. As the most important basis, learning and testing dataset are determined through Delphi and K-Means Clustering evaluation. The evaluation conclusion shows that conditions of topography features, layer of saturated soft soil, engineering geology, and salinity of groundwater, influence construction land suitability as predominant factors in Hangzhou. And the BPNN model has obvious advantages for land use suitability issues and related researches.