A performance comparison of supervised machine learning models for Covid-19 tweets sentiment analysis.

Area of Science:

• Natural Language Processing
• Machine Learning
• Public Health

Background:

• The Covid-19 pandemic necessitated real-time situation assessment.
• Social media platforms became crucial for information dissemination regarding Covid-19.
• Accurate sentiment analysis of public opinion on Covid-19 is vital for resource allocation and pandemic management.

Purpose of the Study:

• To perform sentiment analysis of Covid-19 related tweets using supervised machine learning.
• To evaluate the performance of various machine learning classifiers with a novel feature set.
• To compare the proposed approach with existing sentiment analysis techniques.

Main Methods:

• A dataset of Covid-19 tweets was collected from Twitter.
• Tweets were preprocessed, and sentiments were extracted using TextBlob.
• A feature set was created by concatenating Bag-of-Words and TF-IDF.
• Machine learning classifiers (including Extra Trees) and a Long Short-Term Memory (LSTM) model were evaluated.
• Performance metrics included accuracy, precision, recall, and F1 score.

Main Results:

• Extra Trees Classifiers achieved the highest accuracy (0.93) using the proposed concatenated feature set.
• The proposed feature set outperformed the Vader sentiment analysis with GloVe features.
• The Long Short-Term Memory (LSTM) deep learning model showed lower accuracy compared to the machine learning classifiers.

Conclusions:

• The proposed feature set combining Bag-of-Words and TF-IDF significantly enhances sentiment analysis of Covid-19 tweets.
• Extra Trees Classifiers are highly effective for identifying public sentiment towards Covid-19 from social media data.
• Machine learning approaches demonstrate superior performance over deep learning (LSTM) for this specific task.