Forecasting financial volatility is crucial for informed decision-making in financial management. Realized volatility (RV) in financial time series exhibits complex and nonlinear characteristics that pose challenges for accurate prediction. This study presents a comparative analysis of hybrid models that integrate Long Short-Term Memory (LSTM) networks with mode decomposition techniques to forecast RV for Apple, Google, and Netflix stocks for high-frequency volatility forecasting. The empirical evaluations were performed using metrics including RMSE, MAE, and RMSPE. The paper evaluates three mode decomposition methods: Empirical Mode Decomposition (EMD), Variational Mode Decomposition (VMD), and complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) for volatility forecasting. The empirical results evidence the superior performance of the VMD utilizing the LSTM over other decomposition techniques utilizing LSTM and LSTM variants. This research underscores the effectiveness of hybridizing LSTM with advanced mode decomposition techniques, highlighting their potential to enhance volatility forecasting in financial markets.