臨床薬理の進歩　No.47

195/342

ModelNaive BayesLinear Discriminant AnalysisLogistic RegressionK Neighbors ClassifierDecision Tree ClassifierSVM - Linear KernelRidge ClassifierRandom Forest ClassifierQuadratic Discriminant AnalysisAda Boost ClassifierGradient Boosting ClassifierExtra Trees ClassifierExtreme Gradient BoostingLight Gradient Boosting MachineCatBoost ClassifierDummy ClassifierAUC：Area Under the Receiver Operating Characteristic Curve、Prec.：Precision、Recall：Sensitivity、Kappa：Cohen’s Kappa、MCC：Matthews Correlation Coefficient、F5-score：Recall 重視型 F 値、TT：Training TimeModelExtreme Gradient BoostingGradient Boosting ClassifierCatBoost ClassifierLight Gradient Boosting MachineDecision Tree ClassifierK Neighbors ClassifierExtra Trees ClassifierNaive BayesRandom Forest ClassifierLinear Discriminant AnalysisLogistic RegressionAda Boost ClassifierSVM - Linear KernelRidge ClassifierQuadratic Discriminant AnalysisDummy ClassifierAUC：Area Under the Receiver Operating Characteristic Curve、Prec.：Precision、Recall：Sensitivity、 Kappa：Cohen’s Kappa、MCC：Matthews Correlation Coefficient、F5-score：Recall 重視型 F 値、TT：Training Time表 2　未調整データに基づく ICIM 発症予測モデル各分類器の性能比較表 3　クラス不均衡補正として SMOTE を適用した場合の各分類器の性能比較0.5410.6980.0030.6500.0280.0140.6710.0000.0000.5080.0000.0000.4980.0000.0000.5500.0000.0000.6490.0000.0000.5300.0000.0000.5160.0000.0000.6070.0000.0000.5860.0000.0000.5400.0000.0000.5700.0000.0000.5980.0000.0000.5810.0000.0000.5000.0000.0000.925 0.515 0.996 0.677 0.895 0.494 0.853 0.623 0.913 0.487 1.000 0.653 0.917 0.413 0.984 0.580 0.703 0.411 0.437 0.422 0.904 0.330 0.404 0.363 0.940 0.295 0.939 0.447 0.699 0.881 0.016 0.031 0.956 0.147 1.000 0.255 0.844 0.038 0.303 0.068 0.867 0.029 0.600 0.054 0.847 0.019 0.933 0.037 0.746 0.000 0.000 0.000 0.844 0.000 0.000 0.000 0.604 0.000 0.000 0.000 0.500 0.000 0.000 0.000 0.0050.000-0.0050.0180.0150.0160.0000.0000.0000.0000.0000.0000.000-0.003-0.0030.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000-0.001-0.0010.0000.0000.0000.0000.0000.0000.0000.000-0.0010.0000.0000.0000.0000.0000.0000.675 0.714 0.620 0.645 0.651 0.695 10.114 0.577 0.634 0.417 0.417 0.357 0.359 0.444 0.522 0.012 0.065 0.253 0.379 0.065 0.104 0.054 0.127 0.037 0.128 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.2000.5881.7623.0960.3220.2420.1481.2961.1260.8742.5761.7780.5000.3587.8800.1220.498 3.670 0.448 0.406 2.126 2.312 0.270 1.762 0.660 0.396 1.036 0.248 0.172 1.196 0.310 医療情報を活用した免疫療法誘発心筋炎予測デバイスの日米共同開発F5-score Accuracy AUC Recall Prec. F1-score Kappa MCC TT (Sec)0.0610.2810.0270.9920.0000.9980.0000.9980.0000.9930.0000.9980.0000.9980.0000.9980.0000.9980.0000.9980.0000.9970.0000.9980.0000.9980.0000.9970.0000.9980.0000.998F5-score Accuracy AUC Recall Prec. F1-score Kappa MCC TT (Sec)0.525 0.995 0.502 0.994 0.496 0.995 0.422 0.994 0.412 0.989 0.332 0.989 0.302 0.993 0.283 0.448 0.152 0.992 0.039 0.990 0.030 0.990 0.020 0.990 0.000 0.990 0.000 0.990 0.000 0.990 0.000 0.990 179