Title: Visualising model stability information for better prognosis based network-type feature extraction
Authors: Connor Smith - University of Sydney (Australia) [presenting]
Samuel Mueller - University of Sydney (Australia)
Boris Guennewig - University of Sydney (Australia)
Abstract: Findings to deliver new statistical approaches to identify various types of interpretable feature representations that are prognostically informative in classifying complex diseases are reported. Identifying key features and their regulatory relationships which underlie biological processes is the fundamental objective of much biological research; this includes the study of human diseases, with direct and important implications in the development of target therapeutics. We present new and robust ways to visualise valuable information from the thousands of resamples in modern selection methods that use repeated subsampling to identify what features predict best disease progression. The new method VIVID learns from feature importance measures via pairwise feature comparisons to identify significant features. We will show how the selected features are repeatedly ranked higher and are more stable than other features. We take advantage of cluster analysis to first construct a set of nested feature groups and to then select an optimal group of features. We highlight the computational speed and requirements of VIVID and how it is able to deal with data where the number of features is continually increasing.