Most induction algorithms for building predictive models take as input training data in the form of feature vectors. Acquiring the values of features may be costly, and simply acquiring all values may be wasteful or prohibitively expensive. Active feature-value acquisition (AFA) selects features...