Forecasting Natural Gas Prices in Malaysia Using Double Exponential Smoothing, Arima and XGBoost Models

Abstract

Natural gas is an integral component of Malaysia’s energy structure, accounting for a significant share of overall primary energy consumption, which makes the market prone to price volatility due to the influence of changes in policies, demand, and global factors. As reliable natural gas price forecasts play a pivotal role in making informed energy policies and other related decisions, this research work examines and compares the ability and performance of three models: Double Exponential Smoothing (DES), Autoregressive Integrated Moving Average (ARIMA), and eXtreme Gradient Boosting (XGBoost) in making accurate forecasts of the prices of natural gas in Malaysia using historical price time series data on the basis of Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) on the training and testing datasets. Findings indicate the poor performance of the DES model in terms of accuracy in conditions of price volatility, while XGBoost yields an impressive performance in terms of accuracy on the training dataset but performs poorly on the testing dataset; hence, the ARIMA(2,1,0) model is shown to provide the lowest forecast errors on the testing dataset, making it the model of superior performance on the test datasets in terms of precision and accuracy. The forecasts of the selected model indicate a stabilisation of prices towards the end of the forecast time period, ranging between the values of RM 37-38/MMBtu.