I am trying to use multi label classification(Binary relevance). Many predictor variables are factor variables. It seems that for BR all the predictors has to be numeric. So how do I deal with this? Should I convert factors to integers?
Asked
Active
Viewed 139 times
0
-
Use dummy binary variables for factors or one hot encoding with feature hashing. – Sandipan Dey Jan 22 '17 at 20:34
-
Thanks Sandipan! I tried dummy binary variable, but I face memory issue as my data is relatively bigger with lot of factor labels. I don't know much about feature hashing.. would that be less memory intensive? – Ravi Yadav Jan 22 '17 at 20:42
-
Yes feature hashing will be – Sandipan Dey Jan 22 '17 at 21:15