Forecasting Price of Crude Oil Using the Weight Markov Chain (WMC) and ARIMA Model Techniques

Abstract

Crude oil prices play a pivotal role in the global economy, influencing everything from energy expenses to inflation rates. Accurately predicting these prices is crucial for governments, businesses, and investors. In our paper, we introduce two forecasting models: the first is a weighted Markov chain (WMC), and the second is an autoregressive integrated moving average (ARIMA) model. The weighted Markov chain is employed to capture the probabilistic transitions of crude oil prices across different states or regimes. The weights assigned to these transitions are determined based on historical price data, allowing us to give more importance to recent price movements. This approach offers a versatile and adaptive framework for modeling the dynamic behavior of oil prices. We collected data from the period of January 1, 2000, to September 29, 2023, for crude oil prices, spanning 285 months, from the website (https://sa.investing.com/commodities/brent-oil-historical-data). We compared two models using metrics such as mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and AIC. Our study led us to conclude that the ARIMA (2,0,0) model outperforms the weighted Markov Chain (WMC) model. This conclusion is supported by lower values for mean absolute error, root mean square error, mean absolute percentage error, and AIC. Additionally, the predictive performance of the chosen model for crude oil prices showed an increasing trend for the months of October and November, followed by a decline for the months of December, January, February, March, April, May, June, July, August, and September.

Keywords

• Forecasting, Weighted Markov Chain, ARIMA, MAPE, MAE, AIC