The stock market provides a gateway for analyzing a company's present and future financial value to the world of investors as a potential ground for receiving better returns than traditional savings or bonds. This paper presents a deep learning model-BiLSTM (Bidirectional Long Short-Term Memory) models predicts the closing prices of main stocks: Meta, Amazon, Apple, and Tesla. The study leverages socioeconomic data as predictors to improve the accuracy of the forecasts. Comparative analysis is done on the performance of six well-known optimization algorithms in the Keras machine learning library—Adam, Adagrad, stochastic gradient descent, RMSprop, and others—that are utilized to train the LSTM model. Performance metrics, such as MAE and RMSE, are used to evaluate model performance. The results allow one to understand the most efficient method for optimizing the prediction of stock prices and practical implications in terms of offering data-driven strategies for decision-making by financial analysts and investors.