land change modeler/accuracy


I have been trying to create a land prediction model of a specified region in Istanbul  in land change modeller tab. I have managed to do it( by choosing logistic regression)  but it is not accurate enough since the validation tab tells me so. How can I increase my hits and decrease false alarms and misses? 

I put the land use maps of 1997 and 2007 for predicting 2014.

The reason I try to predict 2014 is just a test if I am on the right track for predicting 2023 right ( for 2023, I included the maps of 2007 and 2014). 

my variables are :

distance to roads

distance to forests

distance to urban

distance to industrial places



Distance to Universities

Distance to Hospitals

Road Density

My land use map has 4 classes which are urban, industrial, barrenland, forest.

One more thing, I tried VALİDATE and I got the quality disagreement as 0.3158 and allocation disagreement( gridcell level) as 0.0838. 

For a detailed view, you can check : 
Thank you for your helps.



Please sign in to leave a comment.