地域分類法を用いた土地利用予測誤差の最小化

Abstract

Errors in a transition matrix of land-use cause very serious problems when we forecast the land-use by Markov chain models. In the previous papers we proposed the method for evaluating the error of estimate in the land-use forecasting. Also, we constructed the area dividing method using the neuralnetwork theory. In this paper we attempt to combine these two methods and decrease the error of estimate by dividing the objective area into some adequate zones. The result of case studies on the actual urban lattice data shows that the error of estimate can be minimized by the proposed method.