Table 1.
A summary of the related work.
Table 2.
A sample of the dataset available at IEEE data port [30].
Table 3.
A comparison of sentiment scores between those provided by IEEE data port (SS1) and those determined by our approach after preprocessing (SS2).
Table 4.
Counts of tweets within the three classes.
Fig 1.
Term counts in Covid-19 tweets dataset [30].
Fig 2.
The word-cloud to highlight the topics after preprocessing.
Table 5.
The parameter settings of five machine learning models.
Fig 3.
Our proposed approach based on concatenation of BoW and TF-IDF.
Fig 4.
Our proposed methodology.
Table 6.
Three sample tweets.
Table 7.
Sample tweets after removing usernames and links.
Table 8.
Sample tweets after removing punctuation marks and conversion to lower case.
Table 9.
Sample tweets after removing stopwords and numeric values.
Table 10.
Sample tweets after stemming.
Table 11.
The TextBlob performance on original data.
Table 12.
TextBlob performance after preprocessed data.
Table 13.
Training and testing tweets count for SS1 and SS2.
Table 14.
TF-IDF features on two sample tweets.
Table 15.
BoW features on two sample tweets.
Table 16.
Concatenated features on two sample tweets.
Table 17.
Models performance for SS1 using TF-IDF.
Table 18.
Models performance for SS1 using TF-IDF.
Fig 5.
ETC confusion matrix using SS1 and SS2 under TF-IDF features.
Table 19.
Models performance for SS1 using BoW.
Table 20.
Models performance for SS2 using BoW.
Fig 6.
ETC confusion metric for SS1 and SS2 under BoW features.
Table 21.
Models performance for SS1 using TF-IDF and BoW concatenation.
Table 22.
Models performance for SS2 using TF-IDF and BoW concatenation.
Fig 7.
ETC confusion matrix for SS1 and SS2 using concatenated features.
Table 23.
Models accuracy performance comparison for SS2 and Vader.
Fig 8.
Models accuracy performance for SS1, SS2 and Vader using concatenated feature engineering technique.
Table 24.
Model accuracy performance comparison on SS2 with GloVe, CNN-LSTM, DNN and concatenated features.
Fig 9.
Models accuracy performance for SS2 using the concatenated feature engineering technique.