This paper introduces a new algorithm for cost-sensitive classification, called ICET
(Inexpensive Classification with Expensive Tests — pronounced “iced tea”). ICET uses a
genetic algorithm (Grefenstette, 1986) to evolve a population of biases for a decision tree
induction algorithm (a modified version of C4.5, Quinlan, 1992). The fitness function of the
genetic algorithm is the average cost of classification when using the decision tree, including
both the costs of tests (features, measurements) and the costs of classification errors. ICET
has the following features: (1) It is sensitive to test costs. (2) It is sensitive to classification
error costs. (3) It combines a greedy search heuristic with a genetic search algorithm. (4) It
can handle conditional costs, where the cost of one test is conditional on whether a second