Classifying seismograms using the FastMap algorithm and support-vector machines

Data

We assess the performance of FastMapSVM using seismograms from two data sets. All seismograms used in this paper record ground velocity at a sampling rate of 100 s⁻¹ and are bandpass filtered between 1 Hz and 20 Hz before analysis using a zero-phase Butterworth filter with four poles; we refer to this frequency band as our passband.

Stanford Earthquake Data Set (STEAD)

The first data set is the Stanford Earthquake Data Set (STEAD)⁵, a benchmark data set with >1.2 × 10⁶ carefully curated, three-component (3C), 60 s seismograms for training and testing algorithms in Earthquake…