Forecasting LNG prices with the kernel vector autoregressive model

Abstract

ABSTRACT LNG prices in the Northeast Asian countries are closely related multivariate time series, because they are traded with similar contracts. For the analysis of multivariate time series data, the vector autoregressive model is one of the most successful tools to use. But the vector autoregressive model assumes a linear relationship between the present and previous data, which sometimes provides unreliable results. To address this problem, we applied the weighted version of the least squares support vector machine to the vector autoregressive model. In numerical studies with liquefied natural gas importing prices in four Asian countries, comparisons with other methods indicated that the proposed kernel vector autoregressive model provides more satisfying results on fitting and forecasting for multivariate time series.