Exchange rates forecasting with decomposition-clustering-ensemble learning approach

This paper proposes a new EEMD-LSSVR-K-based decomposition-clustering-ensemble learning approach for foreign exchange rates forecasting by integrating ensemble empirical mode decomposition (EEMD), least square support vector regression (LSSVR) and K-means clustering algorithm. Clustering strategy is used to extend the fixed-weighted meta-synthetic in decomposition-ensemble learning approach to weighted with local data characteristics meta-synthetic. Our proposed approach can effectively solve the shortcoming of fixed-weighted meta-synthetic in decomposition-ensemble learning approach. Meanwhile, our proposed approach is applied to four type exchange rates forecasting. The empirical results show that our proposed approach significantly improves the level and directional accuracy of exchange rates forecasting, and verify the importance of clustering strategy in decomposition-ensemble learning approach.