{"id":"W4415820085","doi":"10.2139/ssrn.5688482","title":"Learning the Signature of Normal Driving to Predict Harm: A Variational Autoencoder Framework for Uncovering Safety-Critical Events in Multidimensional Data","year":2025,"lang":"","type":"preprint","venue":"SSRN Electronic Journal","topic":"Traffic and Road Safety","field":"Engineering","cited_by":0,"is_retracted":false,"has_abstract":false,"ca_institutions":"WSP (Canada)","funders":"","keywords":"Autoencoder; Hyperparameter; Baseline (sea); Anomaly detection; Crash; Scalability; Percentile; Poison control; Supervised learning; Unsupervised learning","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"metacan-v3-hybrid-931329e0061c","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.001171237,0.0006437328,0.0006824853,0.0005798314,0.0002202365,0.0006020708,0.001040516,0.001230858,0.0006032685],"category_scores_gemma":[0.002982856,0.0005717659,0.0006678349,0.0004360196,0.000780874,0.001040345,0.0009711448,0.001899487,0.0001806031],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0004444481,"about_ca_system_score_gemma":0.0008475169,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.006702457,"about_ca_topic_score_gemma":0.006080273,"domain_scores_codex":[0.9997848,0.00005254645,0.00001232839,0.0000688092,0.00003583934,0.00004572172],"domain_scores_gemma":[0.9990109,0.0006459147,0.00009525121,0.00008768544,0.0001027151,0.00005757724],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0001248809,0.0001370923,0.005962409,0.00005885738,0.0000925028,0.00008046799,0.00009277536,0.9051853,0.006952389,0.009723234,0.001104372,0.07048579],"study_design_scores_gemma":[0.00000146762,0.000005414348,0.0004124231,0.000002379554,0.000002481492,0.000003807523,0.000003201215,0.9975647,0.0001892364,0.001768708,0.0000433177,0.000002890755],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1640093,0.0004455597,0.8337655,0.0005436594,0.00006370153,0.00002910475,0.000242645,0.0002382632,0.0006622424],"genre_scores_gemma":[0.9454495,0.0002620541,0.05203958,0.00009844543,0.00006800432,0.00003424968,0.0004090463,0.00005431351,0.001584837],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.006702457,"threshold_uncertainty_score":0.01332688,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.009404366843308297,"score_gpt":0.2734164414460484,"score_spread":0.2640120746027401,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}