Forecasts and Reliability Analysis of Port Cargo Throughput in Hong Kong

Abstract

Hong Kong, the busiest container port in the world, has been using a regression analysis approach to forecast port cargo throughput for its port planning and development over the decades. In this paper, the neural network models are proposed and developed for forecasting 37 types of freight movements and hence Hong Kong port cargo throughput from 2002 to 2011. The historical data (1983–2000) of freight movements and explanatory factors are the input data used for model development. The models developed are used to forecast 1 year of freight movements for validation with actual data in 2001 and comparison with those forecasted by regression analysis. Using the same models, freight movements are then forecasted for the next 10 years based on projected explanatory factors and combined to form the predicted port cargo throughputs. The Monte Carlo simulation is used to assess the reliability of the forecasts due to projection error of explanatory factors and compare the results forecasted by regression analysis for three different growth rate scenarios. Results show that forecasts made by the proposed neural network models are more conservative, more reliable, and more comparable to reality.