Unsupervised Feature Selection for Pattern Discovery in Seismic Wavefields

摘要: This study presents an unsupervised feature selection approach for the discovery of significant patterns in seismic wavefields. We iteratively reduce number features generated from time series by first considering significance individual features. Significance testing is done assessing randomness with Wald-Wolfowitz runs test and comparing observed theoretical variability In a second step in-between dependencies are assessed based on correlation hunting subsets using Self-Organizing Maps (SOMs). show improved discriminative power our procedure compared to manually selected cross-validation applied synthetic wavefield data. Furthermore, we apply method real-world data aim define suitable earthquake detection phase classification recordings.