Stock market prediction based on sentiment analysis using deep long short-term memory optimized with namib beetle henry optimization

Abstract

Stock price prediction is a recent hot subject with enormous promise and difficulties. Stock prices are volatile and exceedingly challenging to predict accurately due to factors like investment sentiment and market rumors etc. The development of effective models for accurate prediction is extremely tricky due to the complexity of stockdata. Long Short-Term Memory (LSTM) discovers patterns and insights that weren’t previously visible, and they can be leveraged to make incredibly accurate predictions. Therefore, to perform an accurate prediction of the next-day trend, in this research manuscript, a novel method called Updated Deep LSTM (UDLSTM) with namib Beetle Henry optimization (BH-UDLSTM) is proposed on historical stock market data and sentiment analysis data. The UDLSTMmodel has improved prediction performance, which is more stable during training, and increases data accuracy. Hybridization of namib beetle and henry gas algorithm with the UDLSTM further enhances the prediction accuracy with minimum error by excellent balance of exploration and exploitation. BH-UDLSTM is then evaluated with several existing methods and it is proved that the introduced approach predicts the stock price accurately (92.45%) than the state-of-the-art.