I attempted to do land use classification (CART classifier in GEE) using the 2020 South African National Land Cover Map as training data (73 classes) (although only 53 of the classes occur in my study area).
The overall accuracy is only 40 - 45 %.
I saw in the map publisher's accuracy report that they decreased their classes and understand that my low accuracy is ascribed to the large quantity of classes I am trying to use.
What is your experience with land use classification in GEE and how can I attempt to improve the overall classification accuracy?
This student thanks you in advance.
Dear Cindy Viviers
These links might be helpful, have a look:
1. https://www.mdpi.com/2072-4292/13/17/3501/pdf
2. https://www.analyticsvidhya.com/blog/2021/04/google-earth-engine-machine-learning-for-land-cover-classification-with-code/
Kind Regards
Qamar Ul Islam
- Badges
- Science topic