Sentiment analysis is beneficial for public opinion analysis in many disciplines. Emergency management, politics, health, consumer analytics, commercial valuation, financial market forecasts, and criminal prediction have all benefited from it. Several studies have examined Twitter sentiment analysis since so many individuals regularly offer their thoughts on a wide range of issues. This study examines the complexities of sentiment analysis in social networks. Layer-wise unsupervised pre-training and classification fine-tuning enhance feature learning in the Deep Belief Neural Network (DBNN). This architecture improves model generalization, especially for complex data patterns. The publication covers both the fundamentals of sentiment analysis and cutting-edge approaches. Performance comparison, the main uses of sentiment analysis, its limitations, and future prospects are also key objectives. According to our findings, lexicon/rules, machine learning (ML), and deep learning (DL) are the top methods for sentiment analysis. This includes a detailed comparison with other recent surveys. Comparing tools based on objective criteria identifies limitations in major tools that must be addressed to enhance the end-user experience.