In the domain of sentiment analysis, handling large amounts of unlabeled data, particularly from dynamic and mobile social media content, presents a significant challenge. Achieving high accuracy in sentiment classification requires the ability to effectively process and interpret this vast, constantly changing data. Recent research has shown that hybrid deep learning approaches deliver the best results for these tasks. This study introduces a novel hybrid model, Fuzzy-GAN, which integrates Fuzzy SVM and Generative Adversarial Networks (GAN) for multi-class sentiment analysis on social media, particularly addressing the dynamic nature of social content and mobility. The Fuzzy-GAN model resolves the ambiguity in fuzzy classification through fuzzy annotation, effectively capturing the fluid and context-sensitive nature of online sentiments. It generates multiple sentiment categories, including positive, very positive, negative, very negative, and neutral, providing valuable insights for decision-makers. Experimental results demonstrate that Fuzzy-GAN outperforms existing models, delivering superior performance metrics such as precision, recall, and F1-score. This approach successfully extracts sentiment features from large, unlabeled, and dynamic datasets, showcasing its potential for advanced sentiment analysis in the context of mobile and evolving social media content.