In recent years, deep learning techniques have emerged as powerful tools for analyzing and predict- ing complex patterns in sequential data across various fields. This study employs an ensemble of advanced deep learning models: Long Short-Term Memory (LSTM), Bi-Directional LSTM, Gated Recurrent Unit (GRU), LSTM Convolutional Neural Network (CNN), and LSTM with Self-Attention, to enhance prediction accuracy in time series forecasting. These models are applied to three distinct financial datasets: Tata Motors, HDFC Bank, and INFY.NS, we conduct a thorough comparative analysis to assess their performance. Utilizing K-fold cross-validation, we convert loss (MSE) into RMSE and MAPE, which help estimate accuracy .we achieved train accuracies of 97.46% for Tata Motors, 75.93% for INFY.NS, and 56.60% for HDFC Bank. Our empirical results highlight the strengths and limitations of each model within the ensemble framework and pro- vide valuable insights into their effectiveness in capturing complex patterns in financial time series data. This research underscores the potential of deep learning-based ensemble techniques for improving stock price forecasting and offers significant implications for investors and the development of sophisticated trading and risk management systems.