Forecasting Energy Market Dynamics with ARIMA approaches and Complex Network Indicators
No Thumbnail Available
Date
2025-12-30
Journal Title
Journal ISSN
Volume Title
Publisher
Scientific and practical journal "Economics and technical engineering"
Abstract
This study develops and evaluates a hybrid forecasting framework for energy market dynamics that combines classical econometrics with complex network science. Using daily West Texas Intermediate (WTI) spot prices from May 23, 1990 to October 30, 2025, we target short-horizon risk by forecasting 7-day forward returns, standardized via a rolling 50-day mean and volatility to mitigate heteroskedasticity. The univariate price series is mapped into Natural Visibility Graphs (NVGs) on rolling windows ( = 25, 50, 75, 100 days), and a vector of topological descriptors is extracted at each step, including spectral measures (e.g., algebraic connectivity, spectral gap, natural connectivity, graph energy) and centrality/efficiency and clustering indicators (e.g., global efficiency, harmonic centrality, betweenness, maximum degree, clustering, assortativity). These metrics serve as exogenous regressors in an ARIMAX model, enabling the forecasting engine to incorporate structural information embedded in the geometry of price history. Empirical results show that NVG topology exhibits regime-dependent signatures: crisis episodes are associated with sharp shifts in centralization and connectivity, and several network indicators become statistically significant predictors of standardized 7-day returns. In particular, global efficiency and harmonic centrality repeatedly emerge as dominant covariates, with coefficient signs varying across window lengths, consistent with multi-timescale market behavior. The sample is split into a pre-2020 training set and a post-2020 testing set to stress-test robustness under extreme events. The findings highlight heterogeneous horizons in oil price dynamics and the value of NVG features for practical forecasting and monitoring.
Description
Keywords
WTI crude oil, energy market forecasting, ARIMA, natural visibility graph, complex network indicators, spectral graph measures, harmonic centrality, global efficiency, early-warning signals
Citation
Bielinskyi, A., Solovieva, V., Radko, V., Seleznov, M., & Pavlysh, T. (2025). Forecasting Energy Market Dynamics with ARIMA approaches and Complex Network Indicators. Scientific and practical journal "Economics and technical engineering", Vol.3 No.2 (2025), 55–65.