Stock price prediction using bidirectional simple recurrent neural network optimized with Optuna and Technical indicators
Author(s): Priya Sidhu, Himanshu Aggarwal and Madan Lal
Abstract: The stock price prediction is a challenging problem in the financial markets as it is always a volatile market subject to numerous economic and behavioral factors. This paper suggests the hybrid deep learning model that combines Bidirectional Simple Recurrent Neural Network (BiSRNN) that is optimized using the Optuna framework with technical indicators. The model is tested on five NIFTY 50 stocks such as Reliance industries, Tata Consultancy services (TCS), Infosys, ICICI bank and HDFC bank. The BiSRNN model is able to handle forward and backward temporal relationships of a time-series dataset, whereas Optuna is able to automate hyperparameter optimization in the network to achieve an optimal outcome. Experimental outcomes indicate that the model can achieve high levels of prediction with a high R2 of greater than 0.89 and a low MAPE of less than 1% in all the stocks chosen indicating the effectiveness and accuracy of the model. The results show that combining technical indicators with automated optimization can greatly improve the accuracy of the predictions, so the given approach can be utilized in financial forecasting when it comes to the utilization.
Research Journals