EEMD-LSSVR-Based Decomposition-and-Ensemble Methodology with Application to Nuclear Energy Consumption Forecasting

Abstract

Based on the principle of "decomposition and ensemble" and strategy of "the divide and conquer" [1,2], a hybrid Methodology integrating ensemble empirical mode decomposition (EEMD) and least squares support vector regression (LSSVR) is proposed for nuclear energy consumption forecasting. In the proposed EEMD-LSSVR-based Decomposition-and-Ensemble Methodology, the EEMD is first applied to decompose the original data of nuclear energy consumption into a number of independent intrinsic mode functions (IMFs). Then the LSSVR is implemented to predict all the extracted IMFs independently. Finally, the predicted IMFs are aggregated into an ensemble result as the final prediction using another LSSVR. The empirical results demonstrate that the novel methodology can strikingly outperform some other popular forecasting models both in level forecasting accuracy and in direction prediction accuracy.