The current implementation doesn't support sparse format. What do you mea blank values aren't missing values ? in all cases, the current implementation just skips any record with missing values.
On Tue, Nov 27, 2012 at 8:40 AM, Qinghao Dai <[email protected]> wrote: > Hello, > I am trying to train Random Forest in Mahout, by a arff file in sparse > format. > Some records contain blank values. > Not missing value(?). I can replace missing value with other values. > For example. A record like this: > {32 1, 45 0, 100 20} > This record in sparse format means, the 33th attribute is 1, the 46th > attribute is 0, and the 101st attribute is 20. And all others are blank > values. > I am transform this data file into non-sparse format, but Can I just take a > sparse format as input? > -- > Best Regards, > Qinghao >
