Modeling Landuse changes using Bayesian Networks

Abstract : Bayesian networks are powerful tools for decision and reasoning under uncertainty. This paper proposes an example of application of Bayesian networks to geographical information systems. Our problem concerns the classification and prediction of land-use on the basis of spatial, geophysical and geometry parameters. The experimental study is done on data concerning French department of Gers, along the agricultural valley of Sousson. In the whole experimentation we compare results given by general Bayesian networks with the ones of nave bayes, which are simple Bayesian networks dedicated to classification tasks.