search

AreteY / post-wildfire-recovery

An Earth Lab Certificate project studying post-wildfire recovery.
MIT License
0 stars 1 forks source link

Compare vegetation indices using Random Forests algorithm #2

Open AreteY opened 2 years ago

AreteY commented 2 years ago

Determine which index best discriminates between burned and unburned areas.

AreteY commented 2 years ago

Hi @Chathu84 - I've tried this using a cross validation method with the training data being classified dnbr data. I'm unable to get an oob score; I think it's because I have too much training data for bagging (using entire tile 1000x1000). I'm going to try again with a much smaller tile, so that each class only has 100s of counts rather than 100000 counts.

Chathu84 commented 2 years ago

Hi @AreteY . Let's discuss about this today. I want more information about this process. So are you using all the pixels in 1km x 1 km tile as the training data? And to classify other tiles or just the validation of some pixels in the same tile?