Skip to main content

View full-text article in PMC

. 2020 Aug 12;22(8):e17478. doi: 10.2196/17478

Table 8.

Performance of sentiment classifiers.

Classifiers	Area under the receiver operating characteristic curve (95% CI)	Precision	Recall	F1
Logistic regression	0.78 (0.71-0.84)	0.73	0.88	0.82
Random forest	0.78 (0.70-0.83)	0.78	0.79	0.82
Support vector machine	0.69 (0.64-0.78)	0.66	0.98	0.75
Naive Bayes	0.75 (0.66-0.82)	0.75	0.79	0.80
CNN^a (vaping-related word vectors)	0.74 (0.66-0.81)	0.73	0.85	0.80
LSTM^b (vaping-related word vectors)	0.74 (0.69-0.82)	0.75	0.81	0.81
LSTM-CNN (vaping-related word vectors)	0.75 (0.71-0.84)	0.74	0.91	0.83
BiLSTM^c (vaping-related word vectors)	0.74 (0.68-0.81)	0.72	0.91	0.82
CNN (GloVe^d word vectors)	0.81 (0.75-0.87)	0.72	0.96	0.86
LSTM (GloVe word vectors)	0.78 (0.71-0.84)	0.76	0.82	0.84
LSTM-CNN (GloVe word vectors)	0.80 (0.74-0.86)	0.83	0.84	0.84
BiLSTM (GloVe word vectors)	0.83 (0.78-0.89)	0.79	0.79	0.88

^aCNN: convolutional neural network.

^bLSTM: long short-term memory.

^cBiLSTM: bidirectional long short-term memory.

^dGloVe: Global Vectors for Word Representation.