Abstract:

This article presents a comparison of the most popular models in Sentiment Analysis. The following algorithms were used for comparison: Naive Bayes, Support Vector Machine, Random Forest, Gradient Boosted Trees, Logistic Regression. The comparison was based on data obtained from Twitter, which included positive, neutral and negative classifications. We presented results that contain several measurements of classification quality which include: precision, recall, F1-score and accuracy. Based on our research results, the best algorithm to perform such analysis is XGBoost except for a few exceptional cases where Naive Bayes achieves better results.