A Compressed Sensing-Based Denoising Approach in Crude Oil Price Forecasting

摘要

Crude oil price forecasting has been a difficult challenge for years. To improve the forecasting performance, a novel forecasting method is proposed through combining compressed sensing based denoising (CSD) approach and least square support vector regression (LSSVR) forecasting model. In the forecasting model, the grid search algorithm is used to optimize the parameter of LSSVR. Compared with the wavelet denoising method, the compressed sensing based denoising method shows better performance when they are applied to forecasting models. The entire forecasting model CSD-LSSVR also shows its superiority in direction accuracy prediction which is of great significance in business decision making.