A Novel Sine Cosine Optimization with Stacked Long Short-term Memory-enabled Stock Price Prediction

Abstract

Background: In the global financial market, the stock price index is used to analyse the performance of securities and the stock market. It can be obtained by accumulating stock price movements of every firm in the exchange market. A proper stock price prediction (SPP) model becomes essential for investors in turning the security market into a profitable place. Objective: Earlier works in the SPP models involve different approaches, such as statistical models, fundamental examination, time-series prediction, and machine learning (ML). Result and Method: Deep learning is a kind of ML model that tries to define high level conceptual concepts by the use of a learning process at distinct levels and stages. This study, in this view, provides a new sine cosine optimization (SCO) model with a deep learning-enabled stock price prediction (SCODL-SPP). The SCODL-SPP model intends to predict the closing prices of the shares using a deep learning model. The proposed SCODL-SPP model involves primary data pre-processing using a min-max normalization approach. A stacked long short-term memory (SLSTM) model is used to forecast stock values. Because hyperparameters in DL models are crucial, selecting them optimally can help improve prediction performance. Conclusion: The SLSTM Model's hyperparameters are optimised using the SCO algorithm in this research. According to the experiments, the SCODL-SPP model outperforms other models in terms of prediction accuracy.