Exploring Sentiment Analysis on Twitter: A Comprehensive Study Using Machine Learning and Deep Learning Approaches

These days, we essentially live on social media platforms, sharing thoughts and opinions about topics that are relevant to us all the time. As a result, information presented as viewpoints is flooding social media platforms, particularly Twitter. This study investigates sentiment analysis as a field in relation to social media, with an emphasis on Twitter data. Sentiment analysis (SA) which seeks to ascertain whether a text is positive, negative, or neutral in its context is an essential component of NLP and AI. Examining and contrasting these algorithm's performances considering scalability, accuracy, and performance into account is the main goal. The study uses efficient data balancing techniques to address the problems caused by imbalanced sentiment datasets. Text data is formatted for machine learning (ML) input using feature extraction techniques and TF-IDF representation. Confusion matrices, accuracy scores, and performance are evaluated using F1-score, recall, and precision of ML and Deep Learning (DL) algorithms that are put into practice and assessed. The study uses Naïve Bayes (NB), Multinomial NB, Linear SVC, Logistic Regression of ML algorithm, and RNN of DL algorithms to examine user sentiments as they are expressed in social interactions, customer feedback, and emotional data. We have evaluated the performance of those algorithms and identified the algorithm that achieves the highest accuracy during both the training and testing phases. RNN, which is deep learning, gives the highest accuracy 83.53%. Our suggested approach is working well, and the results are pleasing. The study concludes with insights into the limitations of the research, such as the exclusion of neutral tweets because of project goals, model complexity, and data imbalance.