Forecasting Baltic Dry Index using non-linear chaotic models artificial neural networks

摘要

We have tried to forecast a specific maritime index with two different forecasting methods, one based on chaos theory and the other on neural networks. The ability of making reliable forecasts of maritime indices can be very useful for maritime industry e.g. about freight market, the sale & purchase market and the demolition market. Our main purpose was to make forecasts with the above mentioned methods and compare and analyze the results so as to testify the reliability of the forecasts of these methods and find which one is the most efficient method of forecasting. The data we have used was from the General Baltic Dry Index (BDI) of all Bulk Carrier Vessels time series. The first forecasting method was based on chaos theory, with this method the time series of the General Baltic Dry Index (BDI) of all Bulk Carrier Vessels has been analyzed with a specific mathematical procedure that was followed and then has been made a forecast on this index through a specific software (NLTSA V2.0, Siri-opoulos C. & Leontitsis A. (2000)) which was created in C/C++ programming language. The second method that has been used was a forecasting method based on neural networks, at this method was made a forecast through specific software (Alyuda Neurointelligence), which was applied on a particular type of neural networks for every maritime index that we used. Finally was made a comparison between the forecasting results of the two methods so as to find which one gives the most accurate results and makes the most reliable approach of the future trend of the variable that was analyzed.