Random Forest Classification and feature importance score

s2tbx
#1

I applied RF for classification. don’t understand the following statement in the classifier text file
Testing feature importance score:
Each feature is perturbed 3 times and the % correct predictions are averaged

The importance score is the original % correct prediction - average

I want to find the feature importance score; File is uploaded
newClassifier26.txt (4.7 KB)

RF validation/OOB
#2

image
image.png1659×334 55.2 KB

Source

#3

I am enclosing text filed of classifier generated when applied Rainforest classification through SNAP
newClassifier26.txt (4.7 KB)
I used 14 features and RF gives some ranking in classifier for each feature. May I know is it the importance of the feature. Need guidance plz

#4

Basically the file you attached refers to the accuracy of RF, and the importance of each class, but be careful this doesn’t mean that your classification of land cover is correct, unless this RF classes could be compared using validation data, Similar issue is well discussed in the following thread,

Source of the thread

#5

Hi, do u know the mean of " correct prediction - average "?

#6

Random Forest Classification is widely used. There are a number of good documents such as Understanding Random Forest (previously mentioned) and Wikipedia’s entry and Brieman and Cutler’s web pagesas well as textbooks.