Twitter Sentiment Analysis Under COVID-19
Analyzing political sentiment and tweet popularity during the pandemic.
Check out the full report here.
This project analyzes the popularity of political tweets during the COVID-19 pandemic. Using 10,000 political tweets extracted with Twitter’s API, we built a Naive Bayes classifier to predict tweet popularity, incorporating retweets, favorites, and follower counts.
Key Achievements:
- Naive Bayes Classifier: Classified tweets into “Not Popular”, “Popular”, and “Very Popular” categories, achieving a 48% improvement in precision for predicting “Very Popular” tweets compared to the baseline.
- Popularity Score: Engineered a custom metric combining retweets and favorites normalized by follower count to better capture tweet popularity trends.
- Exploratory Data Analysis (EDA): Performed detailed analysis of skewed engagement metrics, handling over 1,500 outliers and visualizing insights using log transformations and correlation analysis.