Title: GRAPES: semi-automatic approach for forecasting models to predict GameStop prices using cloud computing and machine learning

Authors: Tan Van Vo; Sukhpal Singh Gill

Addresses: School of Electronic Engineering and Computer Science, Queen Mary University of London, London, England, UK ' School of Electronic Engineering and Computer Science, Queen Mary University of London, London, England, UK

Abstract: Since the Covid-19 pandemic, we have seen a surge of retail investors that now can easily trade anywhere in the world with just a Smartphone. Social media groups like Reddit's WallStreetBets have almost put a few hedge funds close to bankruptcy by driving GameStop share prices to the sky. In this work, we propose a framework called GRAPES which uses Cloud Computing and Machine Learning to explore various forecasting techniques in predicting GameStop prices. In addition to this, this work also provides light insight into semi-automating forecasting models using tools such as Google Cloud Platform (GCP), Airflow and Streamlit. Moreover, we monitored the investment funds from Ark Invest to provide additional insight into the market in general. Overall, the paper shows the Autoregressive Moving Average (ARMA) model gives the best accuracy based on the Mean Absolute Percentage Error (MAPE) of 1.12%. This means the predictive model is out with an average of 1.12% from the actual price.

Keywords: apache airflow; Google cloud platform; docker; stock; WallStreetBets; streamlit; forecasting; GEM; GameStop; ETF; exchange-traded fund.

DOI: 10.1504/IJGUC.2022.126191

International Journal of Grid and Utility Computing, 2022 Vol.13 No.5, pp.538 - 550

Received: 30 Dec 2021
Accepted: 12 Mar 2022
Published online: 14 Oct 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article