Bayesian Optimized Long-Short Term Memory Neural Network Framework in the Study of Stock Market Price Time Series Prediction
Download as PDF
DOI: 10.25236/iccem.2021.061
Corresponding Author
Sihang Zhao
Abstract
Stock market values are erratic and difficult to forecast. We present a Bayesian Optimized Long-Short Term Memory (BO-LSTM) Neural Network in this study to anticipate stock market values. We discovered that standard LSTM fails to generate reliable prediction results because to problems in tuning in hyperparameter space, therefore we developed a technique that uses a Bayesian approach to automatically determine the best hyperparameters. BO-predicted LSTM’s outputs have a relative error of 4% when compared to observed values, and it has significantly enhanced the LSTM approach in the field of time series prediction.
Keywords
Bayesian optimization, Lstm, Cnn, Stock market prediction