I have unbalanced data now what

You can either perform cost-sensitive classification or try a resampling filter to get a more balanced class distribution:

• Resample filter (supervised version, to take class distribution into account)
• SMOTE (available through SMOTE package)