Machine Learning-based Sentiment Analysis of Twitter Data on the Russia-Ukraine Conflict
Abstract
The Russia-Ukraine conflict has emerged as a focal point of global discourse, with Twitter serving as a significant platform for real-time public sentiment ex- pression. In this study, sentiment analysis techniques applied to Twitter data regarding the Russia-Ukraine conflict are explored, with machine learning models employed to derive insights. This study compares the efficacy of Bag-of-Word and term frequency-inverse document frequency (TFIDF) feature engineering strategies using a bespoke dataset of 87,547 tweets and a regular Kaggle dataset of 43,398 tweets. Accuracy, precision, recall, and F1 score measures are used to assess seven machine learning models: gradient descent classifier, stochastic decision forest classifier, binary split classifier, margin maximization classifier, probability distribution classifier, and sigmoid regression classifier. The analysis reveals TF-IDF, particularly in conjunction with the decision tree classifier, as a promising approach for sentiment analysis on Twitter concerning the Russia-Ukraine conflict. These findings contribute to understanding sentiment dynamics and provide valuable insights for future research in this domain.