Unlocking Stock Price Prognostication Dual Attention-Infused Bi-Directional LSTM Empowered by Bayesian Optimization DA-Bi-LSTM-BO

Historical stock performance is a key component in stock trading decisions, and it takes skillful statistical data analysis to navigate the stock market’s unpredictability and maximize returns. It has been difficult to estimate stock prices accurately even with a great deal of study and the creation of machine learning (ML) and deep learning (DL) algorithms. For accurate prediction, a thorough grasp of the contextual factors influencing prior stock values is essential. To tackle this issue, it becomes essential to use long short-term memory (LSTM), which is well known for its capacity to record long-term data dependencies. A lot of models have been developed for stock price prediction; however, it is often forgotten to use Bayesian Optimization, a novel method for figuring out the best parameters while training deep learning models. The intricacy of parameter optimization increases, rendering conventional techniques less efficient and time-consuming. This paper presents a new optimization strategy for stock price prediction that combines a dual attention-based bi-directional long short-term memory (DA-Bi-LSTM-BO) model with a Bayesian optimizer. The dual attention mechanism (DA) enhances the model’s ability to capture both temporal dependencies and contextual relevance in the data, leading to more precise predictions. Bayesian optimization (BO) further refines the model by efficiently tuning its parameters, which is crucial for handling the increased complexity of deep learning models. Experimental results reveal that the proposed method outperforms existing models like ED-LSTM, ED-Bi-LSTM, and attention-based LSTM (AM-LSTM), demonstrating its potential to improve stock trading decision-making through more accurate forecasts. To minimize loss functions and forecast future values, evaluation measures like MAE, MSE, and RMSE are used.