As a way for people to express their worries or complaints to political authorities, protests are an essential component of democracy. Protests have become more widespread due to increased awareness of civil rights, which has been exacerbated by the growth of social media and technological improvements. We looked at Twitter data to see how people felt about the farmers’ protests. More advanced techniques that make use of machine learning and deep learning models have replaced traditional sentiment analysis techniques, which only categorized tweets into either positive or negative categories. Several approaches were used in this work, including TF-IDF, CNN, RoBERTa, LSTM, BERT, Distil BERT, Logistic Regression, Linear SVC, and Naive Bayes. Remarkably, the CNN algorithm performed better than the others on a regular basis, with better accuracy, precision, recall, and F1-Score measures. Additionally, Distil BERT continued to perform well on all metrics, closely followed by RoBERTa and BERT. On the other hand, Naive Bayes performed worse across measures, suggesting space for improvement. The research included both categorization and prediction tasks using nine different machine learning models. CNN was shown to be the best option, closely followed by Distil BERT, whereas Naive Bayes performed poorly on this dataset.