An Accurate Stock Prediction Using Ensemble Deep Learning Model

Abstract

Financial textual, statistical, and graphical data have all seen increased interest in machine learning and deep learning as analytical tools in recent years. Anticipating the trend of a stock price in the future is challenging applications of deep-learning(DL) in economics. Several factors influence the amplitde &incidence of ascend and descend of stocks at the alike time, making it difficult to predict their future movement. In addition, these are merely company-level considerations; industry-performance, investor-mood, and economic-factors also have a role in determining the direction of equities in the future. In this research, we offer a unique deep learning method for forecasting stock price behavior in the future. The model combines two recurrent-neural-networks(RNN) with a fully connected NN using a mixing ensemble learning technique. The Standard & Poor's 500 Index serves as a case study for our investigation. Our blending ensemble-DL strategy significantly outperforms the best existing anticipation desgn on the same dataset by decreasing the mean-squared-error(MSE) by 58%, growing the precision by 41%, the recall by 52%, the F1-score by 45%, and the movement-direction-accuracy by 35%. The goal study is to detail our intend ethos and demonstrate how EDL may more correctly anticipate upcoming stock price patterns and help investors make sound financial decisions than more conventional approaches.