University of Oulu

M. Oussalah and A. Zaidi, "Forecasting Weekly Crude Oil Using Twitter Sentiment of U.S. Foreign Policy and Oil Companies Data," 2018 IEEE International Conference on Information Reuse and Integration (IRI), Salt Lake City, UT, 2018, pp. 201-208, https://doi.org/10.1109/IRI.2018.00037

Forecasting weekly crude oil using Twitter sentiment of US foreign policy and oil companies data

Saved in:
Author: Zaidi, A.1; Oussalah, M.2
Organizations: 1Natural Language Processing Group, University of Cambridge, Cambridge CB3 0FD, UK
2Centre for Ubiquitous Computing, Faculty of Information technology, University of Oulu, Oulu, Finland
Format: article
Version: accepted version
Access: open
Online Access: PDF Full Text (PDF, 0.4 MB)
Persistent link: http://urn.fi/urn:nbn:fi-fe2020042119551
Language: English
Published: Institute of Electrical and Electronics Engineers, 2018
Publish Date: 2020-04-21
Description:

Abstract

The drop in crude oil price during late 2014 has had a significant impact on all nations. While some countries have reaped the benefits of low oil prices, others have suffered greatly. As a result, it is no surprise that many academics have attempted to develop reliable models to forecast crude oil price. In the age of information and social media, the role of Twitter and Facebook has become increasingly more relevant in understanding our environment. Many academics have exploited this wealth of data to extract features including sentiment and word frequency to build reliable forecasting models for financial instruments such as stocks. These methodologies, however, remain unexplored for the prediction of crude oil prices. The purpose of this investigation to develop a novel model that uses sentiment of United States foreign policy and oil companies’ to forecast the direction of weekly WTI crude oil prices. The investigation is divided into three parts: 1) a methodology of collecting tweets relevant to US foreign policy and oil companies’; 2) a statistical analysis of the novel features using Granger Causality Test; 3) the development and evaluation of three machine learning classifiers including Naïve Bayes, ANNs, and SVM to predict the direction of weekly WTI crude oil. The findings of the statistical analysis showed strong correlation between the novel inputs and WTI crude oil price. The results of the statistical tests were then used in the development of the predictive model. SVM was found to provide best forecasting performance. Furthermore, using these novel features, the predictive accuracy exceeded that of existing models mentioned in literature.

see all

ISBN: 978-1-5386-2659-7
ISBN Print: 978-1-5386-2660-3
Pages: 201 - 208
DOI: 10.1109/IRI.2018.00037
OADOI: https://oadoi.org/10.1109/IRI.2018.00037
Host publication: 2018 IEEE International Conference on Information Reuse and Integration (IRI)
Conference: IEEE International Conference on Information Reuse and Integration
Type of Publication: A4 Article in conference proceedings
Field of Science: 113 Computer and information sciences
213 Electronic, automation and communications engineering, electronics
Subjects:
Funding: We would like to thank the anonymous reviewers for their valuable suggestions because of which the technical quality of the work presented in this paper has improved. This work is partially supported by EU Marie Skodowska-Curie grant No 645706 and EU grant 770469-Cutler.
EU Grant Number: (770469) CUTLER - Coastal Urban developmenT through the LEnses of Resiliency
Copyright information: © 2018 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.