Project using sentiment analysis to understand twitter's behavior

The present work aims to analyze tweets in English, using a model of sentiment analysis, to see if, in fact, as common sense tells us, twitter is a social network with negative trends, which becomes a worrying fact, since exposing yourself to environments with this load can affect users' mood. To train the CatBoost classification model, used in the sentiment analysis task, we use data obtained through the dataset Emotions dataset for NLP from the kaggle website and, with the trained model, we obtained data from twitter through the Twarc tool. we had as conclusion that, in fact, twitter, in the American context, is a more negative social network, with rates of 44.62% higher than positivity.

Distribution of positive and negative expressions

Wordcloud for negative text

Wordcloud for positive text