{"id":"W2019441277","doi":"10.1109/icsm.2011.6080802","title":"Evaluating software clustering using multiple simulated authoritative decompositions","year":2011,"lang":"en","type":"article","venue":"","topic":"Software Engineering Research","field":"Computer Science","cited_by":10,"is_retracted":false,"has_abstract":true,"ca_institutions":"York University","funders":"","keywords":"Cluster analysis; Computer science; Data mining; Software; Decomposition; Correlation clustering; CURE data clustering algorithm; Machine learning; Artificial intelligence; Programming language","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":"codex-gemma-dda1882f352a","candidate_categories":[],"consensus_categories":[],"category_scores_codex":[0.0004167862,0.0001424918,0.0001321798,0.000192161,0.0002325252,0.0001008648,0.0006322216,0.00005944978,0.00008626201],"category_scores_gemma":[0.00165449,0.0001426756,0.00005541866,0.0005607382,0.0000300011,0.0005614199,0.0004826528,0.000183886,0.00007138726],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0001150782,"about_ca_system_score_gemma":0.00007543403,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0005260042,"about_ca_topic_score_gemma":0.00002891036,"domain_scores_codex":[0.9985918,0.00009013806,0.000218823,0.0003759005,0.0003235515,0.0003997479],"domain_scores_gemma":[0.9976074,0.00153528,0.00004276817,0.0004594582,0.0002157601,0.0001394096],"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":[0.00005104081,0.0004053598,0.0696869,0.0001330351,0.0002423114,0.0001424111,0.02945596,0.7874939,0.0217403,0.003800492,0.0001092714,0.08673894],"study_design_scores_gemma":[0.0002708175,0.00007931182,0.007726537,0.00003553508,0.000004449756,0.00001648326,0.00004315473,0.9886016,0.002571031,0.0004602746,0.000006186824,0.0001846254],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.193358,0.00001618679,0.8054571,0.00001254193,0.0002039748,0.0001659175,0.00000113379,0.000617457,0.0001677046],"genre_scores_gemma":[0.5021986,1.820948e-7,0.4977031,0.00001472003,0.00001401049,0.000004391967,9.850463e-7,0.00001106436,0.00005292559],"genre_candidate":"methods","genre_consensus":null,"teacher_disagreement_score":0.3088406,"threshold_uncertainty_score":0.5818144,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1972400476707733,"score_gpt":0.3931426085210296,"score_spread":0.1959025608502563,"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."}}