CMStatistics 2021: Start Registration
View Submission - CFE
A0247
Title: Real oil price forecasting: Gains and pitfalls of text data Authors:  Luigi Gifuni - University of Glasgow (United Kingdom) [presenting]
Abstract: New text-based measures are developed for assessing human sentiment and economic uncertainty in the oil market. Empirical experiments show that sentiment indexes are very responsive to historical geopolitical events that have affected the price of oil. In contrast, uncertainty indicators may hide structural pitfalls, which create problems when alternative measures of the real oil price are forecasted. We propose a linear kernelization of output forecasts, in order to achieve the best forecasting performance at any time horizon.