The 20 top-ranked features, including ACF, Wavelet (DWT), FFT, and statistical moment-based parameters, for ν(A_max) (top panel), ν_max (middle panel), and Δν (bottom panel) obtained during 100 training/testing iterations. Feature subscripts indicate their position within each feature set. For example, ACF₆ is the sixth of twelve ACF features and shows the highest mean importance over 100 Random Forest training and testing runs.