Table 3.

Classification performance.

	Without sample balance			With sample balance
	Class 1	Class − 1	Class 0	Class 1	Class − 1	Class 0
Precision
Textblob	0.33	0.15	0.60
Vader	0.33	0.19	0.58
TF-IDF + DT	0.46	0.29	0.68	0.41	0.26	0.67
TF-IDF + RF	0.91	0.91	0.66	0.70	0.57	0.79
TF-IDF + NB	0.69	0.58	0.70	0.58	0.35	0.84
TF-IDF + SVM	0.61	0.49	0.76	0.56	0.40	0.81
TF-IDF + LR	0.68	0.58	0.75	0.58	0.41	0.81
FastText + LSTM	0.65	0.43	0.68	0.44	0.26	0.78
GloVe + LSTM	0.56	0.40	0.72	0.40	0.32	0.75
Recall
Textblob	0.57	0.26	0.31
Vader	0.52	0.52	0.24
TF-IDF + DT	0.36	0.21	0.78	0.39	0.31	0.66
TF-IDF + RF	0.24	0.04	0.99	0.61	0.49	0.85
TF-IDF + NB	0.43	0.11	0.92	0.63	0.73	0.63
TF-IDF + SVM	0.56	0.38	0.82	0.62	0.66	0.67
TF-IDF + LR	0.54	0.34	0.88	0.61	0.64	0.71
FastText + LSTM	0.33	0.21	0.90	0.55	0.40	0.61
GloVe + LSTM	0.43	0.35	0.82	0.58	0.34	0.58
F1-score
Textblob	0.41	0.19	0.41
Vader	0.40	0.28	0.34
TF-IDF + DT	0.40	0.25	0.72	0.40	0.28	0.66
TF-IDF + RF	0.38	0.08	0.79	0.65	0.52	0.82
TF-IDF + NB	0.53	0.19	0.79	0.61	0.47	0.72
TF-IDF + SVM	0.59	0.43	0.79	0.59	0.50	0.73
TF-IDF + LR	0.60	0.42	0.81	0.60	0.50	0.75
FastText + LSTM	0.44	0.28	0.78	0.49	0.31	0.68
GloVe + LSTM	0.50	0.37	0.76	0.47	0.33	0.65
Accuracy		Training	Testing		Training	Testing
Textblob			37.4%
Vader			34.9%
TF-IDF + DT		94.6%	59.6%		94.5%	54.3%
TF-IDF + RF		92.8%	67.6%		99.2%	74.4%
TF-IDF + NB		83.9%	69.1%		85.6%	64.4%
TF-IDF + SVM		89.0%	69.7%		92.9%	65.5%
TF-IDF + LR		85.6%	72.1%		94.6%	67.4%
FastText + LSTM		94.2%	66.3%		86.9%	56.8%
GloVe + LSTM		89.4%	65.9%		92.2%	55.3%