This class represents a decision-maker to decide if a query has been
correctly classified or not and to assign a cost to incorrectly
classified queries.
Tries to reduce the complexity of the greedy selection algorithm first selecting
the best b·k cases according to their similarity to the query and then applies
the greedy selection method to these cases.