HiRID-ICU-Benchmark -- A Comprehensive Machine Learning Benchmark on High-resolution ICU Data

Benchmark Model Rank Results
circulatory-failure-on-hiridLGBM#2AUPRC: 0.389±0.003
circulatory-failure-on-hiridLGBM ( + hand crafted features)#3AUPRC: 0.388±0.002
circulatory-failure-on-hiridGRU#4AUPRC: 0.368±0.005
circulatory-failure-on-hiridTCN#5AUPRC: 0.358±0.006
circulatory-failure-on-hiridTransformer#6AUPRC: 0.352±0.006
circulatory-failure-on-hiridLSTM#7AUPRC: 0.322±0.008
circulatory-failure-on-hiridLR#8AUPRC: 0.305±0.000
icu-mortality-on-hiridLGBM ( + hand crafted features)#1AUPRC: 0.626±0.000
icu-mortality-on-hiridTransformer#2AUPRC: 0.610±0.008
icu-mortality-on-hiridGRU#3AUPRC: 0.603 ±0.016
icu-mortality-on-hiridTCN#4AUPRC: 0.602±0.011
icu-mortality-on-hiridLSTM#5AUPRC: 0.600±0.009
icu-mortality-on-hiridLogistic Regression#6AUPRC: 0.581±0.000
icu-mortality-on-hiridLGBM#7AUPRC: 0.546±0.008
kidney-function-on-hiridLGBM ( + hand crafted features)#1MAE: 0.45±0.00
kidney-function-on-hiridLGBM#2MAE: 0.45±0.00
kidney-function-on-hiridTransformer#3MAE: 0.48±0.02
kidney-function-on-hiridGRU#4MAE: 0.49±0.02
kidney-function-on-hiridLSTM#5MAE: 0.50±0.01
kidney-function-on-hiridTCN#6MAE: 0.50±0.01
patient-phenotyping-on-hiridLGBM ( + hand crafted features)#1Balanced Accuracy: 45.8±2.0
patient-phenotyping-on-hiridTransformer#2Balanced Accuracy: 42.7±1.4
patient-phenotyping-on-hiridTCN#3Balanced Accuracy: 41.6±2.3
patient-phenotyping-on-hiridLGBM#4Balanced Accuracy: 40.4±0.8
patient-phenotyping-on-hiridLSTM#5Balanced Accuracy: 39.5±1.2
patient-phenotyping-on-hiridGRU#6Balanced Accuracy: 39.2±2.1
patient-phenotyping-on-hiridLogistic Regression#7Balanced Accuracy: 39.1±0.0
remaining-length-of-stay-on-hiridLGBM#1MAE: 56.9±0.4
remaining-length-of-stay-on-hiridLGBM ( + hand crafted features)#2MAE: 57.0±0.3
remaining-length-of-stay-on-hiridTransformer#3MAE: 59.5±2.8
remaining-length-of-stay-on-hiridTCN#4MAE: 59.8±2.8
remaining-length-of-stay-on-hiridGRU#5MAE: 60.6±0.9
remaining-length-of-stay-on-hiridLSTM#6MAE: 60.7±1.6
respiratory-failure-on-hiridLGBM ( + hand crafted features)#1AUPRC: 0.604±0.002
respiratory-failure-on-hiridTransformer#3AUPRC: 0.594±0.003
respiratory-failure-on-hiridGRU#4AUPRC: 0.592±0.003
respiratory-failure-on-hiridTCN#5AUPRC: 0.589±0.003
respiratory-failure-on-hiridLGBM#6AUPRC: 0.585±0.001
respiratory-failure-on-hiridLSTM#7AUPRC: 0.569±0.003
respiratory-failure-on-hiridLogistic Regression#8AUPRC: 0.530±0.000