I have a simple question about filtering attributes in WEKA.
Let's say I have 500 attributes 30 classes and 100 samples for each class which equals 3000 rows and 500 columns. This causes time and memory problems a you can guess.
How do I filter attributes that occur only once or twice (or n times) in 3000 rows. And is it a good idea开发者_StackOverflow社区?
Thank you
Use the following filter
weka.filters.unsupervised.attribute.RemoveUseless
This filter removes attributes that do not vary at all or that vary too much. All constant attributes are deleted automatically, along with any that exceed the maximum percentage of variance parameter.
精彩评论