TClustVID: A novel machine learning classification model to investigate topics and sentiment in COVID-19 tweets | JoVE Visualize

TClustVID: A novel machine learning classification model to investigate topics and sentiment in COVID-19 tweets

Md Shahriare Satu¹, Md Imran Khan², Mufti Mahmud³

¹Department of Management Information Systems, Noakhali Science & Technology University, Noakhali, 3814, Bangladesh.

Knowledge-Based Systems|May 11, 2021

View abstract on PubMed

Summary

The TClustVID model effectively analyzes COVID-19 tweets, extracting public sentiment and prevalent topics with high accuracy. This intelligent approach outperforms traditional methods in understanding public opinion on the pandemic and prevention strategies.

Area of Science:

Computational Social Science
Infectious Disease Epidemiology
Natural Language Processing

Background:

COVID-19 (SARS-CoV-2) presents varied severity, impacting respiratory health, vascular systems, and older adults.
Public discourse on social media, particularly Twitter, significantly shapes opinions and can amplify misinformation.
Understanding public sentiment is crucial for managing the pandemic and its societal impacts.

Purpose of the Study:

To design an intelligent clustering-based classification and topic extraction model (TClustVID) for analyzing COVID-19 public tweets.
To accurately extract significant sentiments and prevailing topics from social media discussions.
To compare the performance of TClustVID against traditional classification methods.

Main Methods:

Collected COVID-19 Twitter datasets from IEEE Dataport.
Applied data preprocessing, tokenization, and word-to-index mapping.
Utilized clustering-based classification and topic modeling with the TClustVID model.

Main Results:

TClustVID demonstrated superior performance compared to traditional classification methodologies.
Successfully extracted significant topics from tweet clusters, categorized by sentiment (positive, neutral, negative).
Identified the most frequent topics reflecting public opinion on COVID-19 and prevention strategies.

Conclusions:

The TClustVID model offers a robust method for rapid analysis of public opinion on health crises like COVID-19.
This approach aids in identifying prevalent attitudes and misinformation trends across diverse populations.
Effective sentiment and topic extraction from social media can inform public health communication and policy.