I am performing a binary sentiment classification (positive/negative) with RapidMiner. My problem is that I have about 400 positive and 1350 negative documents. I get pretty good accuracy but therefore my precision and recall for the positive class is around 60-70%. Are there any operators or methods which can help me to get a better recall/precision (f1 score)?
I thought "Sample (Bootstrapping)" can help me. But this operator just sizes the whole dataset up or down?
