A non-linear forecasting of container traffic: the case-study of the Port of Piraeus, 1973-2008

The paper used data for 427 months of transhipping boxes in the Port of Piraeus between 1973 and 2008 to forecast 20 months ahead. The method used was the non-linear radial basis functions and the forecasting error was 0.19 < 1. Amazing is that the model captured the effect of the port's labour strike! The paper tested for the existence of chaos in the time series, and when this was affirmative applied the rescaled range analysis, due to Hurst and Mandelbrot, in a generalisation of Einstein's 1905 formula. Moreover, various tests checked the existence of normality and these were negative (Jarque and Bera, BDS, excess kurtosis and skewness). Time series after have been turned stationary gave Hurst exponent equal to 0.73 indicating persistence and long-term dependence (black noise) among observations. The system has shown a fractal dimension of 1.54 < 2. Together with Lyapunov exponent we detected chaos in the time series.