Feature selection fusion (FSF) for aggregating relevance ranking information with application to ZigBee radio frequency device identification
Trevor Bihl, Michael A. Temple, Kenneth W. Bauer · 2016
A Feature Selection Fusion (FSF) method is developed herein for aggregating the feature relevance ranking scores from Multiple Dimensional Reduction Analysis (DRA) methods. FSF methods are compared and contrasted using the raw relevance scores, ranks, concatenation, and concordance. Issues with discretizing scores are discussed. Results illustrate improved classification and verification performance for some situations considered.