METHODS COMPARATIVE ANALYSIS WHILE LIQUEFIED NATURAL GAS SALES FORECASTING

Abstract

In the paper comparative analysis of the most well-known forecasting methods on liquefied natural gas sales data are presented. The methods using exponential smoothing, ARIMA model and gradient boosted regression trees were chosen for comparison. Their comparative analysis was carried out using the R language and the RStudio IDE. According to the obtained results the exponential smoothing methods showed the best forecast accuracy while short-term forecasting (1-3 days forward). And all the used methods showed the same accuracy while long-term forecasting (28 days forward).