Sentiment analysis of COVID-19 social media data through machine learning

Area of Science:

• Computer Science
• Social Sciences

Background:

• Pandemics like COVID-19 pose significant global threats, impacting health, economies, and societies.
• Social media platforms, particularly Twitter, have become crucial for disseminating and analyzing pandemic-related information.
• Sentiment analysis of social media data offers insights into public perception and the effects of global health crises.

Purpose of the Study:

• To propose a novel approach, Sentimental Analysis of Twitter social media Data (SATD), for enhancing the accuracy of sentiment analysis on COVID-19 related tweets.
• To leverage machine learning techniques for effective classification and analysis of public sentiment expressed on Twitter during the pandemic.

Main Methods:

• Developed a novel approach (SATD) integrating five distinct machine learning models: Logistic Regression, Random Forest Classifier, Multinomial NB Classifier, Support Vector Machine, and Decision Tree Classifier.
• Utilized Twitter data to train and evaluate the performance of the proposed machine learning models.
• Performed experimental analyses to calculate metrics including precision, recall, f1-score, and support, visualized through confusion matrices, accuracy plots, and ROC graphs.

Main Results:

• The proposed SATD approach, utilizing a combination of machine learning classifiers, demonstrated high accuracy in classifying COVID-19 related tweets.
• Experimental results validated the effectiveness of the integrated models in analyzing sentiment from social media data.
• Performance evaluation metrics and graphical representations confirmed the robustness of the proposed method.

Conclusions:

• The SATD approach offers a highly accurate method for analyzing public sentiment towards COVID-19 on Twitter.
• Machine learning models integrated within SATD provide valuable tools for understanding the societal impact of pandemics through social media analysis.
• This research contributes to improved sentiment analysis techniques for public health monitoring during global crises.