Authors : Clement Asare; Derrick Asante; John Fiifi Essel

Volume/Issue : Volume 8 - 2023, Issue 7 - July

Google Scholar : https://bit.ly/3TmGbDi

Scribd : https://tinyurl.com/484xb4m7

DOI : https://doi.org/10.5281/zenodo.8224141

Abstract : Investors and financial professionals in today's dynamic stock markets attach great significance to precise forecasts of stock returns. This emphasis is not only on accurate forecasting models but also their reliability. While conventional machine learning models can predict nonlinear datasets with high accuracy, they often overlook uncertainties in their predictions, leading to unreliable outcomes. This study employed the Bayesian LSTM (Long Short-Term Memory) model for stock price prediction and examined its performance with that of the conventional LSTM model. The findings revealed that the Bayesian LSTM model produces better results than the conventional LSTM model considering the R 2 (R-squared), MAPE (Mean Absolute Percentage Error), and RMSE (Root Mean Squared Error) values.This study provides a more reliable approach for stock price prediction to help investors and financial professionals make informed decisions.

Keywords : Bayesian LSTM Model; Monte Carlo Dropout; Stock Price Prediction; LSTM Model; Machine Learning.