[Retracted] Sentiment Analysis on COVID-19 Twitter Data Streams Using Deep Belief Neural Networks

Table 3

Classifier accuracy comparison with the proposed method.

N-gram	Type of attribute	Classifier accuracy
		Proposed DBN (%)	Naïve Bayes (%)	SVM (%)	K-nearest neighbors (%)
Unigram	All twitter data	80.3	79.4	81.9	73.3
	Information gain >0	84.1	86.6	83.6	74.2
	Best 70% on ranking	88.1	88.0	83.2	73.5
Bigram	All twitter data	86.1	75.2	85.8	62.7
	Information gain >0	90.3	89.0	82.8	63.3
	Best 70% on ranking	89.5	83.0	87.8	62.7
1 to 3 gram	All twitter data	86.1	85.7	82.5	68.8
	Information gain >0	90.1	92.5	84.1	66.0
	Best 70% on ranking	89.0	88.3	83.8	66.4