A Feature Selection Algorithm Performance Metric for Comparative Analysis

This study presents a novel performance metric for feature selection algorithms that is unbiased and can be used comparative analysis across problems. The baseline fitness improvement (BFI) measure quantifies the potential value gained by applying selection. BFI to compare of datasets measuring change in classifier as result selection, with respect where all features are included. Empirical results presented show there complementarity suite on variety real world datasets. normalised correlate problem characteristics algorithm performance, multiple ability paves way towards describing space per-instance algorithms.