Sentiment Analysis and Comprehensive Evaluation of Supervised Machine Learning Models Using Twitter Data on...

Area of Science:

• Computational Social Science
• Natural Language Processing
• Machine Learning

Background:

• The Russia-Ukraine War, escalating in February 2022, has generated significant global public discourse.
• Social media platforms like Twitter are crucial for disseminating public opinion during geopolitical crises.
• Understanding public perception is vital for analyzing the socio-political impact of the conflict.

Purpose of the Study:

• To examine public perceptions of the Russia-Ukraine War.
• To leverage machine learning for analyzing sentiment and opinions expressed in social media data.
• To evaluate the effectiveness of different machine learning models and feature extraction techniques for text analysis.

Main Methods:

• Utilized a dataset of 11,250 tweets related to the Russia-Ukraine War.
• Applied Natural Language Processing (NLP) techniques, including sentiment analysis and entity annotation.
• Developed and tested machine learning models using TF-IDF, Bag-of-Words (BoW), and N-gram feature extraction.
• Compared performance using metrics such as accuracy, precision, recall, and F1-score.

Main Results:

• The Extra Trees Classifier (ETC) model achieved the highest accuracy of 0.84 when combined with the Bag-of-Words (BoW) feature extraction method.
• Evaluated and compared various machine learning algorithms including Logistic Regression, Decision Tree, SVM, XGB, Gaussian Naive Bayes, ADA, and KNN.
• Demonstrated the efficacy of machine learning in classifying emotions and analyzing textual polarity and subjectivity.

Conclusions:

• Machine learning models, particularly the ETC with BoW, are effective for analyzing public sentiment on geopolitical events from social media data.
• NLP and text analytics provide powerful tools for understanding large-scale public opinion during international conflicts.
• The study highlights the potential of computational methods in social science research for crisis analysis.