Projection of the Impact of Climate Change on Crude Oil Prices Based on Relevance Vector Machine

Abstract

We propose an alternative algorithm referred to RVM (relevance vector machine) to circumvent the support vector machine’s (SVM) unnecessary use of basic functions, a large number of support vectors, lack of probabilistic prediction, and longer time computation complexity (TCC). Global annual land-ocean average temperature (GASAT) data and WTI oil market price data extracted from the National Aeronautic and Space Administration US and the US Energy Administration, respectively. The data were preprocessed and used to build RVM models. To evaluate the proposed RVM, its performance was compared to that of a SVM. The results were validated using ANOVA. A significant correlation between the two datasets was found. The relevance vectors for the RVM were significantly less than the support vectors for the SVM, and the TCC for the RVM was significantly better than the TCC for the SVM. The prediction accuracy of both the RVM and the SVM were found to be statistically equal. The RVM model was able to project the impact of GASAT on WTI crude oil prices from 2014 to 2023. The projection can be used by intergovernmental organizations to formulate a global response to combat WTI crude oil price negative impact, which is expected to worsen in the next decade.