{"id":"W3184473617","doi":"10.1007/s10489-021-02662-2","title":"Evaluation of interpretability methods for multivariate time series forecasting","year":2021,"lang":"en","type":"article","venue":"Applied Intelligence","topic":"Explainable Artificial Intelligence (XAI)","field":"Computer Science","cited_by":43,"is_retracted":false,"has_abstract":false,"ca_institutions":"Toronto Metropolitan University","funders":"","keywords":"Interpretability; Computer science; Machine learning; Artificial intelligence; Fidelity; Time series; Series (stratigraphy); Focus (optics); Multivariate statistics; Data mining","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":[0.01578503,0.001192366,0.0008490623,0.003857956,0.0004277448,0.002496723,0.0008846898,0.001360716,0.002808825],"category_scores_gemma":[0.08454087,0.0002581715,0.001243009,0.001501126,0.0006143718,0.002483837,0.001108153,0.001254317,0.0002376811],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.001126165,"about_ca_system_score_gemma":0.0008844323,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.002249565,"about_ca_topic_score_gemma":0.001478746,"domain_scores_codex":[0.9941116,0.003706405,0.000386561,0.0004461357,0.001191921,0.0001573547],"domain_scores_gemma":[0.9004217,0.08945514,0.002504932,0.002920674,0.004214331,0.0004831236],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"design_other","study_design_gemma":"bench_or_experimental","study_design_scores_codex":[0.002769655,0.0006434635,0.02618049,0.000740326,0.0008909527,0.0002580107,0.0005501898,0.3734517,0.006319199,0.03269411,0.002849619,0.5526523],"study_design_scores_gemma":[0.00006946625,0.0002366568,0.004148408,0.00006478198,0.0001149096,0.00004319993,0.00008292736,0.97794,0.002351311,0.01432096,0.0006059347,0.00002148908],"study_design_candidate":"bench_or_experimental","study_design_consensus":null,"genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.2443258,0.00291427,0.7432967,0.001023284,0.0001800128,0.0003037065,0.0007273656,0.001184862,0.006043916],"genre_scores_gemma":[0.8436717,0.0005271816,0.1534527,0.00009401008,0.0001543055,0.0001372575,0.0008700857,0.0001767492,0.0009160606],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.01578503,"threshold_uncertainty_score":0.08348024,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1252773207783487,"score_gpt":0.4005119010539609,"score_spread":0.2752345802756122,"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."}}