Implement data preprocessing
Closed this issue · 2 comments
florianfendt commented
Make a new tab where the user can see a quick summary of the data, like:
Any Na's, constant features etc.
Then make preprocessing available with help of impute, capLargeValues etc.
florianfendt commented
@Coorsaa
I really like how it turned out. Couple of minor things I noticed though:
- For Drop variables the slider input also says choose target. Should be choose columns/features.
Also setmultiple
toTRUE
. Then we can drop multiple columns at once without clicking the button again and again. - For impute we always select one imputation method for a whole class. Maybe the user wants different imputation techniques for specific features. In mlr this is achieved with the
cols
argument. Implementing this would be tedious I guess. But we could use the following workaround: Make a newselectInput
"exclude columns" and add these columns to thetarget
argument of impute. This way the user can use any imputation method for any column. Downside: Action Button needs to be clicked several times. - Rather a design perspective: I think we should separate the inputs "choose method" and "target" from the method specific inputs a bit. Because right now it's not immediately visible for the user what inputs belong to the currently selected preproc method. Either separate by columns or make some space between the rows. What do you think?