• سیاستگذاریهای اقتصادی و مالی در حوزههای فوقالذکر در سطوح ملی، منطقهای و جهانی
reza taleblou; parisa Mohajeri
Abstract
This study investigates the application of recurrent neural network (RNN) models—specifically RNN, long short-term memory (LSTM), and gated recurrent unit (GRU)—in predicting the stock indices of the Iranian energy industry. Using daily time series data from May 1, 2020, to May 1, 2024, the ...
Read More
This study investigates the application of recurrent neural network (RNN) models—specifically RNN, long short-term memory (LSTM), and gated recurrent unit (GRU)—in predicting the stock indices of the Iranian energy industry. Using daily time series data from May 1, 2020, to May 1, 2024, the dataset was divided into a training period (80%) and a testing period (20%). In the first step, the optimal architectures of each model (estimating hyper-parameters) were determined for prediction horizons of 1, 2, 5 (one week), and 20 trading days (one month). Subsequently, prediction errors of the three machine learning models were compared with the linear econometric model (ARIMA) across various forecast horizons. The findings in two areas of cross validations of machine learning models as well as predication error reveal the following insights: First, as the forecast horizon increases, the batch size of optimal prediction decreases for all three machine learning models, and the larger the input training sample size leads to the smaller batch size. Second, in short-term forecast horizons (1, 2, and 5 trading days), machine learning models—particularly LSTM—demonstrate lower prediction errors than ARIMA, while in the 20-trading-day (1-month) forecast horizon, ARIMA's predictive accuracy approaches to the nonlinear machine learning models. Third, forecast accuracy decreases as the horizon lengthens, with accuracy dropping from approximately 98.5% (for a 1-day horizon) to 92.5% (for a 20-day horizon). Finally, selecting the appropriate forecasting method for the stock market indices of energy industries depends on the forecast horizon and data characteristics.
مطالعات اقتصادی مرتبط با حاملهای انرژی (فسیلی، تجدیدپذیر و برق)
Mandana Shiravand; mahdi zolfaghari; samaneh abedi; narges khosravinejad
Abstract
This study investigates the relationship between the Global Economic Policy Uncertainty (GEPU) index and global crude oil prices, employing a combination of classical econometric models—such as VAR, BEKK, and DCC—and advanced machine learning techniques including LSTM, GRU, and XGBoost. The ...
Read More
This study investigates the relationship between the Global Economic Policy Uncertainty (GEPU) index and global crude oil prices, employing a combination of classical econometric models—such as VAR, BEKK, and DCC—and advanced machine learning techniques including LSTM, GRU, and XGBoost. The results indicate a bidirectional and predictable relationship between GEPU and oil prices, along with volatility spillovers. However, the immediate and short-term effects of shocks between the two variables are generally weak and often statistically insignificant. The relationship is more evident at the level of volatility correlation, and in the long run, rising oil prices can contribute to increased global economic instability. Bidirectional Granger causality between the two variables is confirmed, yet impulse responses within the VAR framework are largely insignificant. Volatility-based models and wavelet analysis reveal that co-movements in volatility intensify during crisis periods, and volatility spillovers between GEPU and oil prices are substantial. The application of advanced machine learning methods enhances the accuracy of forecasting the long-term and noisy relationships between the variables. The findings underscore the complexity and multidimensional nature of the interaction between global economic policy uncertainty and oil prices.
• مطالعات اقتصادی مرتبط با حاملهای انرژی (فسیلی، تجدیدپذیر و برق)
Sajad Piri; Zahra Farshadfar
Abstract
High fluctuations in the price of crude oil, as the main source of energy and an important raw material of the global chemical industry, has doubled the importance of accurate estimation and forecasting of its price trend in recent years. The purpose of this applied research, is to increase the ability ...
Read More
High fluctuations in the price of crude oil, as the main source of energy and an important raw material of the global chemical industry, has doubled the importance of accurate estimation and forecasting of its price trend in recent years. The purpose of this applied research, is to increase the ability to predict crude oil prices using non-linear patterns by artificial intelligence. For this purpose, four artificial intelligence networks MLP, RNN, LSTM and GRU have been used and their capabilities compared to each other and the benchmark model, besides their prediction accuracy have been evaluated using the mean squared error method. The studied sample is North Sea Brent crude oil data from Aug 1st 2007 to May 31st 2024 on a daily, monthly and yearly basis.
The results of the research indicate that the network architecture in these models have several advantages in extracting information from the data in order to make more accurate predictions, and the time to obtain future prices is shorter and less error-prone. Also, among the selected non-linear models, GRU has more accurate predictions with less error in different frequencies and in a shorter time.
Introduction
As oil price fluctuations affect both oil exporting and importing countries in different ways, crude oil price is one of the most important key variables in international trade (Salik and Khorsandi, 2022), As a result, policymakers and oil market experts pay attention to its price and its fluctuations. The price of crude oil in the market is the result of many fundamental and non-fundamental factors (Shakri et al., 2018). Therefore, it is not simply possible to categorize and model all the factors affecting the price of crude oil. Since all the basic and non-basic factors that affect the price formation will finally appear in the price of crude oil, it is necessary to pay attention to the price and its fluctuations (Yadgari et al., 2022). Previous research indicate that the trend of oil price changes follows a non-linear pattern (Guo, 2019); and among the non-linear models used in predicting the price of oil, models based on artificial intelligence have shown better results (Gumus and Kiran, 2017; Zhao et al., 2017; Gao et al., 2022). Therefore, the purpose of this research is to improve crude oil prices out-of-sample prediction using non-linear machine learning algorithms. It is assumed that this non-linear long-short-term memory method has better performance than historical average method and multilayer perceptron network and recurrent network.
Methods and Material
The purpose of this applied research, is to increase the ability to predict crude oil prices using non-linear patterns by artificial intelligence. For this purpose, four artificial intelligence networks MLP, RNN, LSTM and GRU have been used and their capabilities compared to each other and the benchmark model, besides their prediction accuracy have been evaluated using the mean squared error method. The studied sample is North Sea Brent crude oil data from Aug 1st 2007 to May 31st 2024 on a daily, monthly and yearly basis.
Results and Discussion
The results of the research indicate that nonlinear neural network models have a better ability in predicting crude oil price in different daily, monthly and yearly frequencies with different volumes of training data compared to historical average linear model and it has less error. These findings are consistent with the results of Farshadfar and Prokopczuk (2019), Luo et al. (2022) and Zang et al. (2020).Calculations and estimation of the studied models show that the MSFE prediction criterion in all the samples used by the GRU is better than other networks. It also indicates that with the increase in training data amount, network prediction power increases.
Conclusion
It can be concluded that the network architecture in these models have several advantages in extracting information from the data in order to make more accurate predictions, and the time to obtain future prices is shorter and less error-prone. Besides that, among the selected non-linear models, GRU has provided more accurate predictions with less errors in different frequencies and in a shorter time.
Acknowledgments
Authors would like to appreciate Eng. Behzad Alipour for his kind collaboration in program coding.
