<p>Stock market forecasting is about predicting future trends in financial markets, such as stock prices and trading volume levels, utilizing sentiment analysis tools. Sentiment analysis is a very popular technique nowadays to extract significant information based on suggestions and reviews of users from raw data available on the internet, which also classifies the sentiment as positive, negative, and neutral. The existing models suffer from the limitations of the market’s complexity, domain-specific language, non-availability of a large labelled dataset, volatility, noise, and unpredictability for stock prediction based on sentiment analysis. Moreover, this research developed a new model - Spider Feline Swarm Optimization based hybrid Gated Recurrent Unit and Bidirectional Long-Short Term Memory (SFS based GRU-BiLSTM) to overcome these challenges. The model integrates the SFS algorithm into the hybrid GRU and BiLSTM models, which is a revolutionary step in the application of deep learning techniques for stock market forecasting and time-series analysis. The capability of the SFS model to systematically search the problem domain and intelligently find out the most promising area is one of the reasons that the hybrid model method could successfully pinpoint the complex temporal patterns present in financial data. Moreover, due to the adaptivity of SFS, it can quickly adapt its search strategies according to the changes in market conditions, and this ensures its good performance even in turbulent times. In the usage of the SFS characteristics integrated with the hybrid GRU-BiLSTM model, the traders and investors obtain a sophisticated instrument that helps them to make correct decisions in a turbulent environment. The findings show that the performance is significantly better than traditional methods in terms of accuracy, MCC, precision, and recall when looking at a training percentage of 90 on the stock market sentiment dataset, with results of 96.95%, 93.91%, 96.36%, and 97.58% respectively.</p>