Stock market forecasting is very important in constructing investment strategies and making a decision about putting money in stock or not. The approach presented in this study combines an ETL (Extract, Transform, Load) pipeline and a collection of LSTM (Long Short Term Memory) and GRU (Gated Recurrent Unit) models to improve prediction accuracy. The ETL pipe line made handling the historical and real - time data easier, and the ensemble model took advantage of LSTM’s ability to capture long term dependency s and GRU’s ability to reveal time series patterns. Real time sentiment analysis incorporated into the model helps it consider market moving news and social media trends which results in improved predictive performance. This entails automated alert and visualization tools with a decision support system framework. Results from the experiment suggest improved performance over standalone LSTM or GRU models, and provide a robust and flexible solution to stock market analysis.