ORAC-NT Phase F: Takens Embedding Morphology Classifier for Gravitational Wave Event Classification on GWTC-3

Phase F adds a morphology scoring layer to the ORAC-NT gravitational wave pipeline. Using Takens embedding on 0.5s strain windows (bandpass only, no whitening), we extract geometric features from phase space and train a Random Forest classifier on 654 GWTC-3 events. Results:- RF cross-validated AUC = 0.893 ± 0.098 (BBH vs NSBH)- Leave-one-out: 17/17 NSBH correctly ranked (100%)- Permutation test: p = 0.005 (N=10,000)- Blind O4 generalization: 419 events- Score correlates with SNR (ρ=-0.74) and distance (ρ=+0.82) Limitation: classifier captures detectability-driven waveform morphology, not intrinsic astrophysical differences.Suitable as post-detection ranker within ORAC-NT pipeline. Pipeline: raw strain → bandpass → Takens embed → eigenvalue geometry → RF score → p(BBH) | p(NSBH) CPU-only. Trained on Lenovo L14.Related: DOI 10.5281/zenodo.20129975 (Phase E)