{"id":"W4309321702","doi":"10.1148/ryai.220028","title":"Generalizability of Machine Learning Models: Quantitative Evaluation of Three Methodological Pitfalls","year":2022,"lang":"en","type":"article","venue":"Radiology Artificial Intelligence","topic":"Radiomics and Machine Learning in Medical Imaging","field":"Medicine","cited_by":131,"is_retracted":false,"has_abstract":true,"ca_institutions":"Jewish General Hospital; Montreal General Hospital; McGill University Health Centre; University of Calgary","funders":"National Institutes of Health; Fondation de l'Association des radiologistes du Québec","keywords":"Generalizability theory; Computer science; Machine learning; Artificial intelligence; Wilcoxon signed-rank test; Random forest; Overfitting; Feature selection; Artificial neural network; Data mining; Statistics; Mathematics","routes":{"ca_aff":true,"ca_fund":true,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["insufficient_payload"],"consensus_categories":[],"category_scores_codex":[0.01018057,0.0001437283,0.000624999,0.0001734635,0.0001697599,0.000003197282,0.0002175497,0.00008780934,0.001321336],"category_scores_gemma":[0.004977545,0.0001262492,0.0001713369,0.0003601596,0.0005753281,0.00004843346,0.0001415746,0.0008010471,0.000003901308],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001138245,"about_ca_system_score_gemma":0.0001847124,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0002843335,"about_ca_topic_score_gemma":0.00001583192,"domain_scores_codex":[0.9953931,0.002587946,0.0007916331,0.0003826714,0.0006037423,0.0002408982],"domain_scores_gemma":[0.9977637,0.001160996,0.0003704182,0.0002753542,0.000348423,0.00008113873],"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":[0.000738796,0.0003888506,0.01340681,0.00005955976,0.0001578186,0.00001104432,0.001379677,0.7387374,0.04099403,0.1172967,0.00004119835,0.08678812],"study_design_scores_gemma":[0.0001305366,0.0009517589,0.001693627,0.00001206115,0.0001677717,0.00007545032,0.0004379895,0.8822901,0.005764395,0.108295,0.00008241535,0.00009890389],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.6120318,0.001431336,0.3846522,0.0007371339,0.0002205247,0.0003307601,0.000009071054,0.0000274512,0.000559691],"genre_scores_gemma":[0.9668601,0.00007201919,0.0327674,0.0001394747,0.00004650734,0.00004595107,0.00003446432,0.00001436812,0.0000196816],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.3548283,"threshold_uncertainty_score":0.9995916,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.383065366064314,"score_gpt":0.4425682224172016,"score_spread":0.0595028563528876,"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."}}