consider a two$-$class case with classes fraudulent and nonfraudulent. A data mining routine has been applied to a transaction dataset and has classified $85$ records as fradulent $(50$ correctly so$)$ and $915$ as nonfraudulent $(890$ correctly so$).$