A Comparative Analysis of LSTM, ARIMA, XGBoost Algorithms in Predicting Stock Price Direction

Apply

A Comparative Analysis of LSTM, ARIMA, XGBoost Algorithms in Predicting Stock Price Direction

Journal Title: Engineering and Technology Journal - Year 2024, Vol 9, Issue 08

Abstract

This research report presents a comprehensive investigation into the prediction of Google's stock prices using advanced machine-learning techniques. The study focuses on assessing the predictive capabilities of three distinct algorithms: XGBoost, LSTM, and ARIMA, applied to historical stock price data with a specific emphasis on close prices. The primary goal is to develop accurate univariate models to forecast the closing stock price for the next day, a crucial aspect of financial decision-making. The evaluation of model performance utilizes a range of metrics including R-squared, Mean Absolute Error (MAE), and Root Mean Square Error (RMSE) to provide insights into predictive accuracy. Furthermore, the study explores the effectiveness of hyperparameter tuning and ensemble methods in optimizing model performance. The findings highlight the strong performance of the XGBoost model, which achieves a notable R-squared value and effectively minimizes error metrics. While ensemble techniques exhibit potential, they do not consistently outperform all individual models. The subsequent hyperparameter tuning of the XGBoost algorithm achieves a higher R-squared value of 99.47%, accompanied by an MAE of 15.98 and an RMSE of 27.34. This research contributes valuable insights into the potential of machine learning for stock price prediction, emphasizing the importance of thoughtful model selection and parameter optimization.

Authors and Affiliations

Aiyegbeni Gifty , Dr. Yang L,

Keywords

Mathematical Analysis of Energy and Exergy loss in Air Conditioning system as Affected by Ambient Temperature

The article provides a Mathematical analysis of exergy in air-conditioning as affected by ambient temperature. The Mathematical simulation model employs an integrated air conditioning system subjected to varying ambient...

Developing Laboratory Experiments to teach Undergraduate Engineering Students Multi-User Massive MIMO Technology for 5G Cellular Networks

An effective method of facilitating student learning in a laboratory environment is “practice by doing”. The study of modules related to signal-processing for digital communications requires deep mathematical and theoret...

EP ID EP743492
DOI 10.47191/etj/v9i08.50
Views 77
Downloads 0