Reproducibility Assessment for Feature Selection in High-Dimensional Data

• Activity Number: 543
• Type: Contributed
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Chris Fraley* and Qunhua Li
• Companies: Insilicos LLC and Penn State
• Keywords: feature selection ; genomics ; high-dimensional ; support vector machines ; random forests ; classification

Abstract

We investigate the extension of a recent statistical approach to reproducibility in replicate high-throughput signals to feature-selection in data with large numbers of features, as is common, for example, in genomics. It is well known that, in high-dimensional data, some features will appear to affect an outcome by chance, and that subsequent predictions based on these features may not be as successful as initial results would seem to indicate. We apply this new approach to reproducibility to feature-selection methods based on support vector machines and random forests that are used for classification of genomic data.