Richness-independent feature set results at selected τ values (Section 4.3): τ = 2.0 Gyr for the past- and future-merger scores and τ = 1.0 Gyr for the full-merger score. Top: Predicted versus true merger scores for past-mergers (left), future-mergers (middle), and full-merger samples (right); the red dashed line indicates perfect prediction. Bottom: Permutation feature importances for the corresponding models; bars show the mean importance, and error bars show the standard deviation over 10 permutations. The red dashed line marks the 0.01 threshold used for feature selection. Features are labeled F1–F11, ranked by importance within each model. Their definitions are as follows: F1: Δv, F2: σ_r, 1, F3: σ_r, 2, F4: σ_v, 1, F5: σ_v, 2, F6: Δbic, F7: λ₂/λ₁, F8: z, F9: f_m, F10: σ_r, F11: σ_v.