Forecasting oil inventory changes with Google trends: A hybrid wavelet decomposer and ARDL-SVR ensemble model

The current crude oil inventory is still at a historical high, and the destocking of crude oil has become a long-term pattern. In the context that changes in crude oil inventories have attracted much attention from the market, a hybrid Wavelet-ARDL-SVR (WAS) model is proposed to predict the change in the oil inventory. 11 The abbreviations and definitions of field-specific terms used in the paper are shown in Appendix A. First, this paper constructs a new indicator to express the correlation between investor behavior and inventory through Google Trends. Then, aiming at the problem that the relationships between inventory and influencing factors are not significant in the time domain, the application of wavelet finds the driving factors and frequency characteristics of inventory changes. We innovatively find that the buffering effect of inventory is reflected in the long-term, while the speculation effect is mainly superimposed in the short-term, especially the speculation on the supply side is more likely to cause market risks. Finally, the empirical results show that the proposed method provides better prediction accuracy. Especially, it improves sign consistency by 19% compared to the predictions of the research institution.