Closed desecnd closed 10 months ago
Personally, I think the dataset class should accept a list of strings as a way to select attributes. For now, let's just define the attribute subsets as class constants e.g. CICIDS2017.Attributes.ALL
or CICIDS2017.Attributes.YULIANTO_25
.
Currently
CICIDS2017
dataset allows only for selecting 10 hardcoded attributes or full set. Dataset should probably allow for one of: a) selecting larger set of attibutes (e.g. 20) for experiment purpose__docstring__
of algorithmsBut if implementation will be simple hardcoded variable / dictionary, then plain comment with reference is definitely fine
b) holding attributes in order (or performing some kind of sorting?) with respect to specified paper / work.
top_n
attributes, which might be helpful and interesting to test during experiments