Skip to main content
. 2024 Aug 2;24(15):5009. doi: 10.3390/s24155009

Table 5.

Data augmentation and sampling results across all datasets: rare event prediction.

PM Dataset All (Aug. + Tomek Samp.) All (Aug. + ENN Samp.) All (Aug. + ADASYN Samp.) Aug. Only Samp. Only-tomek Samp. Only-ENN Samp. Only-ADASYN No Method
P P and P 0.87 0.72 0.88 0.9 0.86 0.6 0.7 0.7
FF-O 1
R P and P 0.72 0.74 0.7 0.78 0.73 0.64 0.63 0.7
FF-O 1 1 1 0.98 0.85 0.91 0.87 0.67
F1 P and P 0.78 0.73 0.76 0.83 0.77 0.62 0.65 0.7
FF-O 1 0.97 0.99 0.99 0.76 0.73 0.76 0.73
S P and P 3655 (3608 + 47)
FF-O 42,460 (42,407 + 53)
A P and P 0.99 0.99 0.99 0.99 0.99 0.98 0.99 0.98
FF-O 1 1 1 1 1 1 1 1

P and P—pulp-and-paper, BS—Bosch, APS—Air Pressure Systems, BR—Ball bearing, FF—S-Future factories sampled, FF—O-Future factories original, PM—Performance metric, P—Precision, R—Recall, F1—F1 Score, A—Accuracy, S—Support, Aug.—Augmentation, Samp.—Sampling. Bold numbers indicate the best performance.