{"id":"W2014820189","doi":"","title":"An Industrial Case Study on the Automated Detection of Performance Regressions in Heterogeneous Environments","year":2015,"lang":"en","type":"article","venue":"","topic":"Software System Performance and Reliability","field":"Computer Science","cited_by":40,"is_retracted":false,"has_abstract":true,"ca_institutions":"Queen's University; Polytechnique Montréal; York University; Blackberry (Canada)","funders":"","keywords":"Computer science; Weighting; Automation; Benchmark (surveying); Regression testing; Machine learning; Data mining; Software; Artificial intelligence; Software system; Engineering; Operating system","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":[0.005689933,0.001080097,0.0005596234,0.001177845,0.0008348937,0.001254773,0.002200332,0.001455385,0.001523089],"category_scores_gemma":[0.01738531,0.0004714677,0.0007626783,0.001252137,0.001103696,0.001481653,0.001349191,0.001539236,0.0004550703],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001366172,"about_ca_system_score_gemma":0.0008937977,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.007349811,"about_ca_topic_score_gemma":0.008125897,"domain_scores_codex":[0.994904,0.002339632,0.0002725048,0.0008681232,0.001181818,0.0004338241],"domain_scores_gemma":[0.9747122,0.01764225,0.001561995,0.002999182,0.002397555,0.0006868021],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"observational","study_design_scores_codex":[0.001199488,0.002695107,0.171848,0.0007645308,0.0003374063,0.008896582,0.003185756,0.6063059,0.02689794,0.006751283,0.007077463,0.1640406],"study_design_scores_gemma":[0.0001282315,0.0007858016,0.0228113,0.00005866472,0.0000848847,0.0007790204,0.0007556205,0.9516125,0.0152806,0.003025979,0.004625103,0.00005232732],"study_design_candidate":"observational","study_design_consensus":null,"genre_codex":"empirical","genre_gemma":"empirical","genre_scores_codex":[0.9292257,0.0002462958,0.06492653,0.0005444482,0.00003173832,0.0003065279,0.0003859107,0.00191251,0.002420319],"genre_scores_gemma":[0.9518743,0.00008256581,0.04629507,0.0001057081,0.00001795371,0.00009860107,0.0004622269,0.0001812827,0.000882343],"genre_candidate":"empirical","genre_consensus":"empirical","teacher_disagreement_score":0.007349811,"threshold_uncertainty_score":0.03009158,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.07025538241745381,"score_gpt":0.2929690648829657,"score_spread":0.2227136824655119,"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."}}