Due to the fluctuations in stock markets, making price predictions is a complex but necessary task for an investor or financial analyst. Stock price movements assume many dynamic patterns based on the market analysis; these patterns are ever-changing at high volatility, so traditional statistical models and conventional machine learning (ML) techniques help to a limited extent. This study uses an innovative deep learning (DL)-based framework capable of detecting and predicting real-time stock prices. We combine the patterns from training model Recurrent Neural Networks (RNNs) using transfer learning and sent the patterns to proposed Long-Short Term Model (LSTM) networks, and Transformer-based network analyzers to capture long-term correlations in stock price data and help our model learn temporal dependencies most efficiently. In addition, attention mechanisms are applied to select features for better prediction performance. It then analyzes historical market data, technical indicators, and sentiment analysis from broader financial news and social media to augment the predictive capabilities of real-time stock market data. The result analysis shows that the proposed approach obtains high prediction values based on the dynamic hikes in the market.