Accurately predicting the opening price of index NIFTY50 is important for enhancing trading strategies and financial decision-making within the Indian stock market. Traditional approaches rely solely on either historical prices or sentiment indicators, thus fail to capture the interrelation between market dynamics and public sentiments. To overcome this issue, we propose LGRNet, a stacked hybrid model to predict the opening price of NIFTY50 by combining the sentiment features with the time series data. This model has Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Linear Regression (LR) as learners, with LR as the meta-learner. To create a comprehensive feature set, two datasets are merged: news headlines dataset from Hugging Face, processed using FinBERT to extract the sentiment scores and financial dataset obtained via the Yahoo Finance API containing market indicator features—NiftyClose, NikkeiClose, HSIClose, SP500Close, SGXClose, and USDINR exchange rate. The performance of this model is evaluated against multiple baseline models including Bi-LSTM, Bi-GRU, and ARIMA. LGRNet outperformed other models with an R-squared (R2) value of 0.9862 and a Mean Absolute Percentage Error (MAPE) of 0.65%, demonstrating its effectiveness in capturing both temporal and sentiment-driven market signals for improved stock index forecasting.