[R] sum epsilon-regression only works with certain epsilon values

Joao Quariguasi joao_quariguasi at hotmail.com
Wed Jun 19 15:17:32 CEST 2013


I am trying to run the following piece of command:

svm.linear<-svm(price~., subset = generation >0, data = trainset, cost = C, epsilon = 1, type = 'eps-regression')

If epsilon = 1, the program runs fine. For larger values, e.g. 10, I get the following error:

Error in predict.svm(ret, xhold, decision.values = TRUE) : 
  Model is empty!
Calls: RegressionAnalyses ... svm.formula -> svm.default -> na.action -> predict -> predict.svm
Execution halted

My impression is that sum.linear is returning empty. I initially believed that it could be coursed by an infeasible solution, but I think now that this may not be the case as we allow 
for slacks in both constraints, i.e. xi. Can someone help me here? 

I am a little stuck. 

(btw, C=300)

More information about the R-help mailing list