Problem

We needed to track real-time public opinion and sentiment trends about the Russia–Ukraine conflict from social media.

Approach

Collected and preprocessed tweets through the Twitter API, used VADER for initial labelling and TF-IDF for features, then trained and tuned Logistic Regression, Random Forest and Naive Bayes classifiers and compared them. Models were serialised with Pickle behind an interactive Streamlit dashboard that tracks sentiment polarity over time.

Results

94% accuracy on the test set. The dashboard tracks polarity trends over time for real-time public-opinion analysis.