2004 UC Proceedings Abstract
 Using a Neural Network to Evaluate Land Use Change Track: Modeling Author(s): Dave Nouza, Verne LaClair, Matthew Schwab By applying an Artificial Neural Network to a selection of raster spatial layers (GRIDS) it is possible to model build-out within ArcGIS. Inputs include two land use time steps, a mask layer, and user-selected socioeconomic, political, and environmental inputs-driving variables. Cell values for driving variables are normalized between 0.0 and 1.0. For each cell in the study area the real-change between the two time steps is determined, and analyzed vis-à-vis the provided driving variables in order to produce a probability of land-use change layer. The spatial accuracy of model predictions is evaluated in Excel by charting the proportion of correct predictions (# predictions/# cells transitioning in the observed database) against a floating cell-window. ArcObjects and Visual Basic automate integration with the neural network. Dave Nouza PAR Government Systems Corporation 314 South Jay Street Rome , NY 13440 US Phone: 315-339-0491 Fax: 315-339-4771 E-mail: dave_nouza@partech.com Verne LaClair PAR Government Systems Corporation 314 South Jay Street Rome , NY 13440 US Phone: 315-339-0491 Fax: 315-339-4771 E-mail: verne_laclair@partech.com Matthew Schwab New York City Dept. of Env. Protection 71 Smith Avenue Kingston , NY 12401 US Phone: 845-340-7550 Fax: 845-340-8494 E-mail: mschwab@dep.nyc.gov |
|||||
