Using ensemble margin to explore issues of training data imbalance and mislabeling on large area land cover classification.
Andrew MellorSamia BoukirAndrew HaywoodSimon D. JonesPublished in: ICIP (2014)
Keyphrases
- training data
- land cover classification
- training set
- class distribution
- multispectral
- supervised classification
- learning algorithm
- decision trees
- generalization error
- supervised learning
- data sets
- base classifiers
- neural network
- remote sensing data
- test data
- training samples
- ensemble methods
- hyperspectral
- feature selection
- land cover
- remote sensing
- infrared
- satellite images
- support vector machine
- color images
- hyperspectral images
- computer vision
- high spatial resolution
- machine learning