Another missing values problem

[Daniel Schwarz]

Hi,
First of all, thanks for a great tool. I am also a bit confused about
missing values - to be precise, I am not sure how Random Jungle
handles them. I apologise if my questions were answered in this
thread, so I might need a simplified answer.

First, if I got it right, missing values need to be represented either
as a number (specified by the -x option), or as 'NA'. In my case, as
missings occur in a numerical attribute, I would go for 'NA'-s, and I
don't want to impute missing values. Does that mean I don't need to
specify -x command line option?

What does RJ actually do to instances with missing values? I am
classifying 300000+ instances (amino acids), each represented with 8
attributes. In one of the attributes, ~70% of instances miss a value.
Should I just try building a model using 30% of my dataset if I don't
want the missing ones to be imputed?

Many thanks,
Anya

[Daniel Schwarz]

Hi Anya,

Unfortunately, there is a short answer: RJ can handle imputed data
only. It imputes the missing value on-the-fly.

Best regards,
Daniel

[Daniel Schwarz]

Dear Daniel,
thanks for the clarification.

Best wishes,
Anya