Enhancing inflation forecasting across short- and long-term horizons in IRAN: a hybrid approach integrating machine learning, deep learning, ARIMA, and optimized nonlinear grey Bernoulli model

Reza Roshanpour (), Amirreza Keyghobadi (), Ali Abdi () and Mohammad Ehsanbakhsh ()
Additional contact information
Reza Roshanpour: Iran University of Science and Technology
Amirreza Keyghobadi: Islamic Azad University
Ali Abdi: Islamic Azad University
Mohammad Ehsanbakhsh: Islamic Azad University

SN Business & Economics, 2025, vol. 5, issue 6, 1-21

Abstract: Abstract Inflation forecasting remains a critical challenge in economic research and policymaking, requiring accurate and adaptive models to capture both short-term volatility and long-term trends. This study proposes a hybrid forecasting framework integrating Machine Learning (ML), Deep Learning, ARIMA, and the Nonlinear Grey Bernoulli Model (NGBM) optimized using the Imperialist Competitive Algorithm (ICA) to enhance inflation prediction accuracy across different time horizons. The research employs Bidirectional Long Short-Term Memory (BILSTM), Gradient Boosting Machines (GBM), Elastic-Net regression, and ARIMA to model inflation by all datasets for the long-term horizon (ten-step-ahead), while NGBM, optimized via ICA, Krill Herd (KH), and Political Optimization Algorithm (POA), is used for short-term forecasting (three-step-ahead) by recent data. A k-fold cross-validation technique ensures robustness and model generalizability. Empirical validation of Iran’s inflation data from 1937 to 2020 for the long-term and 2010 to 2020 for short-term forecasts reveals that BILSTM achieves the lowest RMSE (0.0532), while ICA-optimized NGBM (RMSE: 0.118) outperforms alternative short-term models. Additionally, the study reveals notable differences between forecasting with all datasets and recent data to predict ten-step-ahead and three-step-ahead. We conducted a Wilcoxon signed-rank test to determine whether the differences in predictive accuracy between the short-term (NGBM-ICA) and long-term (BILSTM) forecasting models are statistically significant. This analysis confirms that forecasting short-term inflation trends using recent data fundamentally differs from long-term forecasting based on the complete historical data set. This highlights the importance of selecting an appropriate modeling approach based on forecasting objectives. The findings offer actionable insights for policymakers and financial analysts.

Keywords: Machine learning models; Deep learning; ARIMA; NGBM model; Inflation; Optimization (search for similar items in EconPapers)
Date: 2025
References: Add references at CitEc
Citations:

Downloads: (external link)
http://link.springer.com/10.1007/s43546-025-00830-x Abstract (text/html)
Access to the full text of the articles in this series is restricted.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:snbeco:v:5:y:2025:i:6:d:10.1007_s43546-025-00830-x

Ordering information: This journal article can be ordered from
https://www.springer.com/journal/43546

DOI: 10.1007/s43546-025-00830-x

Access Statistics for this article

SN Business & Economics is currently edited by Gino D'Oca

More articles in SN Business & Economics from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().