Decoding Financial Markets: Unleashing the Power of Bi-LSTM in Sentiment Analysis for Cutting-Edge Stock Price Prediction

Predicting stock prices accurately is a challenging task due to various influencing factors. This paper addresses the need for precise stock price prediction by presenting an enhanced sentiment analysis methodology using deep learning, particularly focusing on the Bi-LSTM model. The challenge of accurate sentiment analysis in stock prediction is overcome through advanced data preprocessing techniques tailored to maintain sentiment context effectively. The proposed Bi-LSTM (Bidirectional long Short-Term Memory) and CNN (Convolutional Neural Network) model demonstrates superior performance in classifying sentiments with high precision, recall, F1-score, and accuracy. The numerical validation demonstrates the effectiveness of the Bi-LSTM model in achieving a balance across various metrics, highlighting its potential for practical implementation in stock market prediction. The Bi-LSTM and CNN model achieves remarkable precision 90.00%, recall 90.00%, F1-score 90.00%, and accuracy 91.00%. In comparison to established methods, such as Multivariate Sequential Long Short-Term Memory Autoencoder (MSLSTMA), long Short-Term Memory (LSTM) and Gated Recurrent Unit (GRU), the Bidirectional long Short-Term Memory (Bi-LSTM) model exhibits its competitive edge in stock price prediction as supported by evaluation metrics. The results emphasize the importance of tailored data preprocessing and deep learning techniques, showcasing the efficiency of the proposed methodology in enhancing stock price prediction accuracy while maintaining high percentages of precision, recall, and F1-score, along with impressive accuracy.