MULTICLASS CLASSIFIER-INDEPENDENT FEATURE ANALYSIS

In (Holz and Loew, 1994a,b), we presented a metric for use in classifi er-independent feature analysis called relative feature importance (RF I). RFI was shown to correctly rank features on a variety of two-class multi-cluster, mixed-distribution problems, including problems that c annot be solved using the marginal distributions of the features. We p resent here a complete design for RFI, including new results on parame ter settings and calculation details determined on two-class problems. We then show that, using the design arising from exploration of two-c lass problems, RFI extends naturally and successfully to multi-class p roblems. (C) 1997 Published by Elsevier Science B.V.