Research On The Prediction Of Baltic Dry Index Based On Support Vector Machine And Chaotic Time Series

Dry-bulk shipping market is an important component of international shipping market. Baltic Dry Index (BDI) reflects the freight level and is considered to be the barometer in dry-bulk market. Affected by numerous uncertain factors, BDI fluctuates violently and gets out of trend in an unpredictable direction. However, the traditional forecasting methods are difficult to get sound prediction result, which brings much difficulty to the market owners to make decisions.Given that BDI contains the long-term evolution information of dry-bulk shipping market and is a complicated nonlinear time series, this paper conducts the research on the forecasting of BDI with Support Vector Machine (SVM) and Chaotic Time Series Analysis.Firstly, this paper analyses the supply and demand of international dry bulk cargo shipping market, revealing the root cause of the fluctuation in BDI. Next, through the description of the BDI formation and impact factors, the historical tendency and fluctuation rule have been deeply analyzed in a qualitative point of view. Secondly, this paper discusses the modeling idea and develops a hybrid forecasting model based on Support Vector Machine (SVM) and chaotic time series theory. For the parameters selection in hybrid forecasting model, this paper establishes a nonlinear mathematical model of parameters joint optimization and employs Genetic Algorithm for solution. Finally, on the basis of chaotic character identification and data preprocessing, the hybrid prediction model is tested for forecasting monthly average BDI both on single-step and multi-step. In the single-step prediction, simulation experiment results shows that among the methods of model parameters selection, the one based on Genetic Algorithm greatly improved the prediction ability of SVM hybrid model. Compared with the ARIMA model and BP neural network model, the numerical analysis shows that the SVM hybrid forecasting model has higher accuracy both in BDI single-step and multi-step prediction and is better able to grasp the development trend of BDI.