{"id":"W7117253960","doi":"10.6000/1929-6029.2025.14.75","title":"Dataset-Specific Bootstrap-Stability Weighting for Calibrated and Clinically Useful Ensemble Prediction in Medical Diagnosis","year":2025,"lang":"en","type":"article","venue":"International Journal of Statistics in Medical Research","topic":"Machine Learning in Healthcare","field":"Computer Science","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"","funders":"","keywords":"Weighting; Calibration; Medical diagnosis; Random forest; Resampling; Decision tree; Complement (music); Probabilistic logic","routes":{"ca_aff":false,"ca_fund":false,"ca_venue":true,"about_ca":false,"invisible_to_affiliation_only":true},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.0174033,0.0001059317,0.0002955291,0.0007336493,0.00008067987,0.0001669017,0.00178542,0.0002319784,0.0003030288],"category_scores_gemma":[0.04839328,0.00009399033,0.0000338071,0.0005365827,0.0003232642,0.0002570759,0.0005734673,0.002198828,0.000002645401],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0003108569,"about_ca_system_score_gemma":0.00172826,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002596943,"about_ca_topic_score_gemma":0.0004477601,"domain_scores_codex":[0.9930176,0.001078669,0.001575061,0.0003800275,0.003564617,0.0003840495],"domain_scores_gemma":[0.9798858,0.01806467,0.0002162117,0.0002383109,0.001210214,0.0003848509],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"observational","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0003895733,0.0005156875,0.4985762,0.0001687564,0.00005123745,0.001324272,0.0004079366,0.00005491495,0.00001362823,0.06304185,0.03534406,0.4001119],"study_design_scores_gemma":[0.004494052,0.0007628488,0.2319251,0.002283948,0.000005178299,0.0001559356,0.0001978393,0.6121286,0.0000955733,0.08921676,0.05854103,0.0001932353],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.07804654,0.0005037868,0.8692878,0.04927095,0.001670583,0.0003751594,0.0005907011,0.00001470906,0.0002397892],"genre_scores_gemma":[0.9428704,0.003012659,0.05247372,0.000939095,0.0004448392,0.00005824064,0.0001545222,0.00001350083,0.00003306389],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8648238,"threshold_uncertainty_score":0.9596225,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1355572804016163,"score_gpt":0.5023661450127471,"score_spread":0.3668088646111308,"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."}}