Skip to main content
. 2021 Oct 4;18(19):10438. doi: 10.3390/ijerph181910438

Table 4.

Classification performance.

Code Classifier Class Precision Recall F-Score Accuracy
C1 MNB
n-gram: (1, 2)
features: 3000
against 67.26% 80.65% 73.28 70.53%
neutral 75.70% 68.14% 71.64
in favor 70.18% 62.82% 66.18
C2 MNB
n-gram: (1, 3)
features: all
against 66.01% 83.96% 73.88 69.77%
neutral 82.20% 59.71% 69.11
in favor 67.05% 66.97% 66.93
C3 RF
n-gram: (1, 2)
features: all
against 66.94% 77.53% 71.79 68.53%
neutral 68.93% 72.50% 70.59
in favor 70.77% 55.56% 62.10
C4 RF
n-gram: (1, 3)
features: all
against 66.93% 76.29% 71.23 67.86%
neutral 67.25% 72.57% 69.77
in favor 70.60% 54.73% 61.52
C5 SVM
n-gram: (1, 2)
features: all
against 73.31% 76.77% 74.90 72.19%
neutral 74.20% 73.19% 73.63
in favor 69.34% 66.62% 67.86
C6 SVM
n-gram: (1, 3)
features: all
against 70.35% 79.19% 74.39 71.73%
neutral 76.03% 70.29% 72.99
in favor 69.77% 65.72% 67.53
C7 BERT
cased: no
against 78.96% 77.30% 77.97 76.84%
neutral 77.82% 79.16% 78.35
in favor 74.29% 74.13% 74.07
C8 BERT
cased: yes
against 77.18% 76.25% 76.47 75.63%
neutral 77.07% 77.06% 76.88
in favor 73.60% 73.92% 73.45
C9 RoBERTa against 76.82% 83.65% 79.87 78.63%
neutral 81.82% 76.30% 78.84
in favor 78.23% 76.09% 76.99