Comparative Analysis of Risk Hedging Models in the Petrochemical Industry: LSTM vs Linear Regression

Articles in Press

Document Type : Research Paper

Authors
Morteza Robatmili 1
Mohammad Ali Habibi Vostakolayi 2
1 M.A., Department of Accounting, Kharazmi University, Tehran, Iran.
2 PhD Student, Department of Accounting, Science and Research Branch, Islamic Azad University, Tehran, Iran.
10.22075/ijnaa.2025.36498.5373
Abstract
This article aims to use the neural network approach to model risk coverage for petrochemical companies and compare its results with the traditional linear regression model. In this regard, the type and characteristics of the contracts used for risk hedging in this market, the list of included goods and fees were identified in the Iran Commodity Exchange, which shows that the history of using this tool in the petrochemical market includes cases of seasonal forward contracts for products such as polystyrene, Heavy polyethylene, PVC, urea and urea granules are returned. In the following, using a short-term long-term memory model, risk hedging is done using this contract in a selected market (polystyrene), the results of which show the clearly better performance of the neural network model compared to the linear regression model, as well as the increasing trend of the optimal coefficient. Risk coverage is in the period of six years from the spring of 2017 to the winter of 2023 in both models.
Keywords
Risk hedging
risk management
petrochemical market
long short term memory model
commodity derivatives
Subjects
International Journal of Nonlinear Analysis and Applications

Articles in Press, Accepted Manuscript
Available Online from 20 August 2026

  • Receive Date 08 January 2025
  • Revise Date 16 February 2025
  • Accept Date 18 February 2025