Transforming Finance With Deep Learning Predictions

Abstract

Stock market predictions have always been an appealing area for the research community, and deep learning-based techniques have emerged as the newest addition to the domain. Through this chapter, the authors develop a novel multi-output stacked LSTM model for intraday predictions of the top three NIFTY stocks. While previous studies have utilized LSTM for stock market prediction, their focus has primarily been on generating a single output. In contrast, this study addresses the limitations of previous studies and offers a more comprehensive view of the price dynamics by attempting to predict the entire OHLC price range. The proposed model employed historical price and STIs as inputs and used PCA for dimensionality reduction, and obtained an average accuracy of above 91% on the test data. For the performance evaluation of the model, it was benchmarked against similar studies, and its superior prediction performance was measured using metrics such as MAE, MSE, and RMSE.