Integrating Textual Analytics with Time Series Forecasting Models: Enhancing Predictive Accuracy in Global Energy and Commodity Markets
Innovations in Applied Engineering and Technology · 2023
DOI: 10.62836/iaet.v2i1.265 ↗Kurz gefasst
Fachartikel in der Zeitschrift Innovations in Applied Engineering and Technology, 2023. Verfasst von Jiarui Rao und 4 weiteren Autor:innen. Die Arbeit wurde bisher 1-mal zitiert (Crossref).
Abstract (Originalsprache)
This study presents a comprehensive framework for predicting crude oil prices by integrating textual features extracted from news headlines into a time series forecasting model. The rationale for using headlines instead of full articles is twofold: headlines encapsulate the essence of the news, and the approach aligns with previous research by Li et al. The focus on futures news over gold news is justified by the larger dataset and the complex interrelations between futures prices, including gold, natural gas, and crude oil. The methodology involves extracting thematic and sentiment information from news headlines using text mining techniques, constructing daily topic strength indices, and developing an emotional strength index that accounts for the decay effect of news influence over time. The study employs a vector autoregression model to determine the optimal lags for various exogenous sequences, including topics and sentiment indices, relative to crude oil prices. The forecasting model is trained using machine learning techniques such as Random Forest Regression (RF), Support Vector Regression (SVR), Autoregressive Integrated Moving Average (ARIMA), and their extended versions with exogenous variables (ARIMAX). The performance of the models is evaluated using metrics like Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Mean Absolute Percentage Error (MAPE). The results indicate that incorporating textual features significantly improves the prediction accuracy of RF, SVR, and AdaBoost models, while the traditional ARIMA model performs well without textual features. The study also introduces a novel approach combining Ensemble Empirical Mode Decomposition (EEMD) with Independent Component Analysis for analyzing non-linear and non-stationary time series data, specifically applied to gold price analysis. The EEMD-BPNN-ADD model is identified as the most accurate for forecasting, with interval predictions provided for gold prices. This research contributes to the field by demonstrating the effectiveness of integrating textual analysis with traditional financial models for improved market forecasting.
Zitieren
APA
Rao, J., Zhang, Q., Ke, Z., Liu, S., & Liu, X. (2023). Integrating Textual Analytics with Time Series Forecasting Models: Enhancing Predictive Accuracy in Global Energy and Commodity Markets. Innovations in Applied Engineering and Technology. https://doi.org/10.62836/iaet.v2i1.265
Vancouver
Rao J, Zhang Q, Ke Z, Liu S, Liu X. Integrating Textual Analytics with Time Series Forecasting Models: Enhancing Predictive Accuracy in Global Energy and Commodity Markets. Innovations in Applied Engineering and Technology. 2023. doi:10.62836/iaet.v2i1.265
BibTeX
@article{Rao2023,
author = {Rao, Jiarui and Zhang, Qian and Ke, Zong and Liu, Shaoyu and Liu, Xinqi},
title = {Integrating Textual Analytics with Time Series Forecasting Models: Enhancing Predictive Accuracy in Global Energy and Commodity Markets},
journal = {Innovations in Applied Engineering and Technology},
year = {2023},
doi = {10.62836/iaet.v2i1.265}
}
EUVIMED