I am using the Aylien API for Sentiment Analysis. It classifies text as positive, negative or neutral. For the sake of a training example, I am filtering out the neutral answers in order to leave just positive and negative, and then checking performance with the Performance operator. The label has just two classes, positive and negative.
The Performance operator is giving me an error "the prediction values do not match the label attributes". The label is positive and negative but the prediction still has three values, positive, negative and neutral, even though I have filtered out the neutral ones. The metadata for the prediction still includes 3 values (but zero instances of neutral).
How can I correct this?
Solved! Go to Solution.