Stock market prediction is a complicated task due to market volatility, unessential leverages, and non-linear dependencies in fiscal data. Authentic machine learning methods like (Extreme Gradient Boosting) XGBoost and Deep Neural Network (DNN) have been used for stock forecasting, but they always fail to capture long-term anchors in successive data. In this paper, Long Short-Term memory (LSTM) network is a kind of Recurrent Neural Network (RNN) has been validated to be more efficient in timeseries forecasting due to their ability to learn mundane patterns. We use LSTM for stock price prediction and compare its interpretation with XGBoost and DNN models. The validation is done with the experimental task on six publicly available datasets (Kotak Bank, Axis bank, Yahoo Finance, TSLA, ICICI Bank, Infosys), together with stock price from suggested standard economic markets, and integrate social media and fiscal news sentiment analysis to help predicting accuracy. The finding indicates that LSTM gains maximum accuracy in all the six datasets equated to the traditional models.