Feature selection is one of the fundamental problems in nearly every application of statistical modeling, and hyperspectral data analysis is no exception. We propose a new methodology for combining unsupervised and supervised methods under classification accuracy and computational requirement constraints. It is designed to perform not only hyperspectral band (wavelength range) selection but also classification method selection. The procedure involves ranking hands based on information content and redundancy and evaluating a varying number of the top ranked bands. We term this technique Rank Ordered With Accuracy Selection (ROWAS). It provides a good tradeoff between feature space exploration and computational efficiency. To verify our methodology, we conducted experiments with a georeferenced hyperspectral image (acquired by an AVIRIS sensor) and categorical ground measurements.