{"id":"W2282861635","doi":"10.1007/s10618-015-0444-8","title":"On the evaluation of unsupervised outlier detection: measures, datasets, and an empirical study","year":2016,"lang":"en","type":"article","venue":"Data Mining and Knowledge Discovery","topic":"Anomaly Detection Techniques and Applications","field":"Computer Science","cited_by":732,"is_retracted":false,"has_abstract":false,"ca_institutions":"University of Alberta","funders":"Conselho Nacional de Desenvolvimento Científico e Tecnológico; Canadian Network for Research and Innovation in Machining Technology, Natural Sciences and Engineering Research Council of Canada; Fundação de Amparo à Pesquisa do Estado de São Paulo; Teknologi og Produktion, Det Frie Forskningsråd","keywords":"Outlier; Anomaly detection; Benchmark (surveying); Computer science; Data mining; Artificial intelligence; Ground truth; Pattern recognition (psychology); Set (abstract data type); Machine learning","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":"metacan-v3-hybrid-931329e0061c","candidate_categories":["metaresearch"],"consensus_categories":[],"category_scores_codex":[0.03498443,0.001710232,0.001651135,0.00700916,0.001441702,0.003051224,0.00281065,0.002473269,0.0006181667],"category_scores_gemma":[0.1873556,0.000367881,0.001387332,0.005792591,0.00221568,0.005608084,0.002598465,0.001759048,0.0002410891],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.002109847,"about_ca_system_score_gemma":0.001821054,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.004075689,"about_ca_topic_score_gemma":0.004933161,"domain_scores_codex":[0.9454237,0.02698848,0.006062921,0.005060516,0.01550229,0.00096213],"domain_scores_gemma":[0.658196,0.2746069,0.01788623,0.02211152,0.02525189,0.001947556],"domain_codex":null,"domain_gemma":"methods","domain_candidate":"methods","domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.006365941,0.004265401,0.3466473,0.002784813,0.003256323,0.000417447,0.001608432,0.189528,0.009163713,0.01165585,0.01500079,0.409306],"study_design_scores_gemma":[0.0003832132,0.004315083,0.07509547,0.0003935689,0.0008865569,0.001528291,0.001402172,0.8831693,0.01294639,0.01522497,0.004442294,0.0002127602],"study_design_candidate":"simulation_or_modeling","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.7978086,0.005405053,0.1867845,0.0009972693,0.0003056825,0.0004534209,0.00341261,0.001684273,0.003148376],"genre_scores_gemma":[0.9221168,0.0006867843,0.070075,0.00014896,0.0001161745,0.0001700518,0.00605825,0.0001733945,0.000454681],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.9650156,"threshold_uncertainty_score":0.1850175,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1553202106037584,"score_gpt":0.3896393330565236,"score_spread":0.2343191224527651,"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."}}