Hello,
Currently I'm working on Transition Potentials in Setiu, Terengganu, Malaysia. My problem is, i keep getting low accuracy rate even though I increase the iterations and sample size.
At first, I grouped the transition (5 transitions : changes from classes Water, Bare land, Mixed forest, Wetland and Agriculture to Developed area) into 1 same sub model : Disturbance. The accuracy were low< somewhere around 50%. after adjusting the sample size and the iterations number ( i did it for many times, and i ncrease both little by little from 10000, 20000,30000 etc), i keep getting low accuracy rate. After that i decided to not group all the transitions under the same sub model. So, i run the model separately.
for the sub model water to developed area, the accuracy rate produced were high (98.23%). however whenever i run the sub model for bare land to developed area, (though using the same variables as in the first sub model) the accuracy rate were again low (somewhere around 50%). so i pay attention on the variables used in the model and remove the variable that give the least influence to the model, but still the accuracy rate produced did not reaching the satisfied rate (it only increase to 62%).
the same thing happen again when i run the transition potentials for transition of wetland to developed area.
one more thing, is it normal that i keep gaining different result on the variables that have the most and least influence on the transition modeled (same transition, eg: wetland to developed area) even though i've been run it several times?
note:
-Variables in my research are:
a. Slope
b. Elevation
c. Distance from roads
d. Distance from streams
e. Distance from developed area (earlier landcover : 1991)
f. Changes land cover map for different transition.
- land cover maps were from 1991 and 1995, using landsat image.
Please help me and thank you in advance.
