Active Learning And Confidence-Rated Prediction