Decision-Making Based on Predictive Process Monitoring of Patient Treatment Processes: A Case Study of Emergency Patients
Table 5
Evaluation metrics results of prediction models both for the gastroenterology (GA) and for the urology (UR) surgical units.
Models
GA
UR
MCC
AUC-ROC
MAE
RMSE
MCC
AUC-ROC
MAE
RMSE
Total triage
MLR + LR
0.622
0.758
8.961
12.631
0.804
0.784
9.227
12.763
RF + NN
0.850
0.840
5.061
10.948
0.920
0.827
4.851
10.549
LSTM
0.837
0.822
5.068
10.900
0.909
0.836
4.835
10.373
Red triage
MLR + LR
0.647
0.735
8.652
12.370
0.869
0.877
8.153
12.384
RF + NN
0.841
0.836
4.899
10.258
0.935
0.889
4.765
11.235
LSTM
0.815
0.793
4.953
10.585
0.928
0.895
4.415
10.500
Orange triage
MLR + LR
0.677
0.765
8.699
12.380
0.775
0.787
9.194
12.635
RF + NN
0.865
0.841
4.882
10.543
0.908
0.824
5.055
10.244
LSTM
0.851
0.813
4.842
10.695
0.888
0.824
5.071
10.249
Yellow triage
MLR + LR
0.627
0.763
9.089
12.778
0.826
0.792
9.199
12.768
RF + NN
0.849
0.845
5.220
11.081
0.934
0.838
4.889
10.633
LSTM
0.833
0.820
5.282
11.179
0.922
0.818
4.968
10.745
Green triage
MLR + LR
0.628
0.770
9.710
13.693
0.711
0.797
9.390
12.589
RF + NN
0.825
0.845
6.653
12.834
0.879
0.840
6.750
12.530
LSTM
0.785
0.838
6.663
13.080
0.755
0.821
7.134
12.409
AUC-ROC = area under the receiver operating characteristic; LR = linear regression; LSTM = long short-term memory; MAE = mean absolute error; MCC = Matthew correlation coefficient; MLR = multinomial logistic regression; NN = neural network; RF = random forest; RMSE = root mean square error.
