Prediction algorithms are everywhere [11]. Based on knownor observable features of a set of objects, these algorithmsprovide informed guesses about unknown or unobservedoutcomes of the same set of objects. For this article, interestis limited to a single unknown outcome that can takeone of two states: positive or negative. In the context ofcredit scoring, positive may indicate that an applicant fora loan is deemed likely to default on the loan. In the contextof screening biomarkers, positive may indicate that abiomarker is strongly associated with a particular disease.In the context of information retrieval, positive may indicatethat the returned item is relevant for the request. Andin the context of drug discovery in the pharmaceutical industry,positive may indicate that a compound has scoredwell in a biological assay and has been cleared for followuptesting as an active or hit compound. For these and other applications, prediction algorithms typically provide a score (??) that represents the belief that an object is in the positive class; this score is based on attributes of the object.
展开▼
