Requirements

1. Create Contingency Table? of known v. predicted class.
2. Sort dataset according to scores
3. For each example, add a new example to the threshold curve containing the right info. Add a jump to the curve if the scores are the same.
4. Takes score attribute, dataset, and Contingency Table?

Changes

• DONE Change Contingency Table? to allow appending examples.
• Change Contingency Table? to allow zero-sized tables, zero probs.

Unit Tests

• Class is a DatasetFactory
  • get/set evaluators
  • initial evaluators: True Positive, False Positive

• Errors:
  • Null utilities
  • Class index is invalid, not in range of attribute
  • Class att is not nominal
  • Score attribute is not numeric
  • Null Dataset
  • Empty dataset

• Algorithm
  • Call utils to sort based on prediction att.
  • Made evaluators accept iteratively adding examples
  • curve is equal to that created by ThresholdCurve by Weka (Code is shorter).