Title: Forecasting inflection points: Hybrid methods with machine learning algorithms
Authors: Julien Chevallier - IPAG Business School (France) [presenting]
Bangzhu Zhu - Jinan University (China)
Shujiao Ma - Jinan University (China)
Yi Ming Wei - Beijing Institute of Technology (China)
Abstract: Hybrid time-series forecasting models are investigated which are based on combinations of ensemble empirical mode decomposition (EEMD) and least square support vector machines (LSSVM). Several algorithms are considered: genetic algorithm (GA); grid-search (GS); particle swarm optimization (PSO) and uniform design (UD). Theoretical guarantees of prediction accuracy are tested with sine curves. From a numerical testing perspective, we are interested to show the superiority of one approach over another based on multiple arbitrary time series in engineering (HTTP requests to the NASA servers), finance (MCD ticker on the NYSE), macroeconomics (China GDP) and commodities (ECX CO2 Futures). Out-of-sample forecasting superiority is assessed in a horse race based on Diebold-Mariano test statistics.