Unsupervised feature selection and general pattern discovery using Self-Organizing Maps for gaining insights into the nature of seismic wavefields

摘要: This study presents an unsupervised feature selection and learning approach for the discovery intuitive imaging of significant temporal patterns in seismic single-station or network recordings. For this purpose, data are parametrized by real-valued vectors short time windows using standard analysis tools data, such as frequency-wavenumber, polarization, spectral analysis. We use Self-Organizing Maps (SOMs) a data-driven selection, visualization clustering procedure, which is particular suitable high-dimensional sets. Our method based on significance testing Wald-Wolfowitz runs test individual features correlation hunting with SOMs subsets. Using synthetics composed Rayleigh Love waves real-world we show robustness improved discriminative power that compared to subsets manually selected from wavefield parametrization methods. Furthermore, capability techniques investigate discrimination wave phases shown means synthetic waveforms regional earthquake