A new approach for Baltic Dry Index forecasting based on empirical mode decomposition and neural networks
- PDF / 459,750 Bytes
- 19 Pages / 442.205 x 663.307 pts Page_size
- 1 Downloads / 187 Views
A new approach for Baltic Dry Index forecasting based on empirical mode decomposition and neural networks Q i n g c h e n g Z e n g a, C h e n r u i Q u a, A d o l f K . Y . N g b a n d X i a o f e n g Z h a o a a
School of Transportation Management, Dalian Maritime University, 1 Linghai Road, Dalian 116026, PR China. E-mail: [email protected]; [email protected]; [email protected] b Department of Supply Chain Management, I.H. Asper School of Business, University of Manitoba, Winnipeg, Canada MB R3T 5V4. E-mail: [email protected]
Abstract
In this article, a method based on empirical mode decomposition (EMD) and artificial neural networks (ANN) is developed for Baltic Dry Index (BDI) forecasting. The original BDI series is decomposed into several independent intrinsic mode functions (IMFs) using EMD first. Then the IMFs are composed into three components: short-term fluctuations, effect of extreme events and long-term trend. On the basis of results of decomposition and composition, ANN is used to model each IMF and composed component. Results show that the proposed EMD-ANN method outperforms ANN and VAR. The EMD-based method thus provides a useful technique for dry bulk market analysis and forecasting.
Maritime Economics & Logistics advance online publication, 19 February 2015; doi:10.1057/mel.2015.2
Keywords: dry bulk shipping market; empirical mode decomposition; artificial neural networks; forecasting; Baltic Dry Index (BDI)
Introduction The dry bulk shipping market is the major component of international shipping market and it has the characteristics of seasonality, cyclicality, high volatility and capital intensiveness. Owing to the magnitude of investments required and the frequent fluctuations of freight rates, the forecasting of freight rates attracts much attention both from scholars and business practitioners. However, due to the © 2015 Macmillan Publishers Ltd. 1479-2931 Maritime Economics & Logistics www.palgrave-journals.com/mel/
1–19
Zeng et al
complexity of the bulk shipping market and the non-stationary and non-linear nature of freight rates series (Goulielmos and Psifia, 2009), their accurate prediction presents researchers with certain challenges (Goulielmos and Psifia, 2013). In the past decades, econometric and statistical methods, such as ARIMA, VAR (vector auto-regression), GARCH and VECM (vector error correction model) models have been used in the analysis and forecasting of the shipping market. For example, Kavussanos and Nomikos (2003) found that VECM generated the most accurate forecasts of spot prices, but not of future prices. Batchelor et al (2007) compared ARIMA, VAR and VECM in predicting spot and forward freight rates. Their results demonstrated that ARIMA provided better forecasts of forward prices than spot prices, and VAR, VECM slightly outperformed ARIMA in predicting spot prices. Jing et al (2008) investigated the characteristics of volatility in dry freight rates by GARCH model, and the fficiency of GARCH model was verified. Finally, Chen et al (2012) applied ARIMA and VAR
Data Loading...
