Dynamics of land-use change is always interesting and important to study because it is associated with a variety of global issues. This study aims to: (1) assess and predict changes in land use spatially using binary logistic regression model integration and GIS, and (2) assess the validity of the model in predicting changes in land use. Research located in six districts in the outskirts of the city of Yogyakarta daerarah.
Changes in land use is predicted based on the probability value is calculated using binary logistic regression models. Predictor variable changes determined a priori and then selected based on a statistical test method Spearman and Mann-Whitney. Binary logistic regression model used was: Y = 0.8963 to 0.0200 X1 + 0.3551 X2 - 0.0002 X3 - X4 + 0.0002 + 0.0007 0.0003 X5 X6. Six predictor variables in the model are: (1) the distance to the main road, (2) the distance to local roads, (3) the distance of the campus, (4) the distance to land up, (5) the distance to the center of the economy and (6) the density of the road network. The validity of the model in predicting changes in land use dianalisisis using the ROC (Relative Operating Charactristic) and cross tabulation. The validity of the model represented by the actual value of the coefficient of agreement and Kappa statistics ().
Model SIG-binary logistic regression generates predictions of land use changes that are spatial. Category changes the results predicted and actual change category actual value 81.8% agreement and Kappa statistics coefficient 0.24. Coefficient Kappa statistics showed agreement between the predictions and the actual conditions are included in the category of fair agreement. Model predictions generated from GIS-binary logistic regression tends to be over estimate.