This article examines the use of Long Short-Term Memory (LSTM) networks for stock prices. Stock market movements are especially difficult because they are nonlinear. High volatility in the stock market and off-season stock price movements. LSTM is a type of Recurrent Neural Network (RNN) that can handle and predict time-series data. Conventional financial models suffer from nonlinearity and high volatility associated with fluctuating stock prices. Therefore, it cannot effectively predict future trends. However, deep learning has shown the ability to facilitate sequential switching by automatically learning and interrupting complex functions to create navigation. Presented in a sequential distribution. This research uses real-time financial data (i.e., volume, technical indicators, and previous stock prices) to train Long Short-Term Memory (LSTM)-based models that predict future closing stock prices. In addition, the accuracy and robustness of predicted results by the LSTM-based model under real-time conditions are verified using an experiment based on predicting stock price direction. We train our LSTM models based on historical closing stock prices after applying feature extraction techniques that help to improve prediction accuracy. To evaluate performance of the proposed models, we rely on different error measurements reported in literature such as Mean Absolute Percentage Error (MAPE), Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), and accuracy. Next, we conduct real-time testing and validation for different stocks’ closing prices short-term predictions using LSTM-based model with proposed window size. Finally, we conclude the paper emphasizing to enhance generalization capability of our model.