Key points are not available for this paper at this time.
Protests are a crucial component of democracy and a way for people to express their requests and/or displeasure to the government. Around the world, protests have become more frequent for a variety of reasons as people become more aware of their rights. The usage of social media for information and idea sharing has increased dramatically along with technological improvement. In order to better comprehend the global public's feelings about farmers' protest, data was collected from Twitter for this research. Traditional sentiment analysis methods simply determine polarity and categorize tweets as positive, negative, or neutral. With the appearance of various machine learning and deep learning models, sentiment analysis tasks have once more evolved significantly. As an advanced attempt to determine the sentiment of tweets, a variety of machine learning and deep learning approaches were utilized, including BERT, Distil BERT, Logistic regression, Linear SVC, Naive Bayes, and TF-IDF. The algorithm "Distil Bert" consistently outperforms the others with the highest values for Accuracy, Precision, Recall, and F1-Score. "Bert" keeps up a solid performance across all measures while following closely. Additionally, "Linear SVC" exhibits competitive outcomes, especially in Accuracy and Recall. In contrast, "Naive Bayes" displays lower results across the board, indicating potential improvement areas. In this research, both categorization and prediction were performed using six machine learning models. It was discovered that Distil BERT produced the best outcome followed by BERT. Naive Bayes' performance on this type of dataset is poor.
Sresta et al. (Wed,) studied this question.