This study analyzed a deep learning strategy based on “Bi- Directional LSTM network” for forecasting stock values. The suggested method takes in past stock prices and outputs hypothetical new stock values. Utilizing 30-day window width for prediction, the model got trained and tested on the “New York Stock Exchange”, the Nikkei 225, and the Nasdaq Composite. The suggested method indicated a MAPE for the NYSE of 0.014, for the Nikkei 225 of 0.01, and for the Nasdaq Composite of 0.018. These results were compared with the results of a base paper, which showed significant improvement in terms of prediction accuracy. Future work includes testing the studied approach on other stock exchanges and exploring the use of additional features such as news sentiment analysis for further improvements. Overall, this approach showed promise in the field of stock price prediction and could potentially benefit investors and financial analysts in making informed decisions