ON THE USE OF EXPECTED ATTAINABLE DISCRIMINATION FOR FEATURE SELECTION IN LARGE SCALE MEDICAL RISK PREDICTION PROBLEMS
D. R. Lovell, M. J. J. Scott, M. Niranjan, R. W. Prager, K. J. Dalton and R. Derom
August 1997
This report investigates the use of {\em expected attainable discrimination} (EAD) as a measure to select discrete valued features in two-class prediction problems. In essence, EAD tells us the performance we could expect to achieve with a simple histogram probability density model of a given dataset. For discrete valued features, this kind of density model is {\em bias-free} but can have large {\em variance}. Given insufficient training data, such a model's test set performance will be lower than that of a suitably biased model. In light of this, we explore the usefulness of EAD for feature selection.
If you have difficulty viewing files that end '.gz',
which are gzip compressed, then you may be able to find
tools to uncompress them at the gzip
web site.
If you have difficulty viewing files that are in PostScript, (ending
'.ps' or '.ps.gz'), then you may be able to
find tools to view them at
the gsview
web site.
We have attempted to provide automatically generated PDF copies of documents for which only PostScript versions have previously been available. These are clearly marked in the database - due to the nature of the automatic conversion process, they are likely to be badly aliased when viewed at default resolution on screen by acroread.