A practitioners guide to time-series methods for tourism demand forecasting - a case study of Durban, South Africa

This paper compares a variety of time-series forecasting methods to predict tourism demand for a certain region, and is meant as a guideline for tourism forecasters at the commencement of any study who do not have access to large databases in order to create structural models. This study has been conducted at a metropolitan level to forecast the US demand for travel to Durban, South Africa. A brief description of the tourism attractions and context of this area is provided to give a qualitative feel of the system prior to the modelling process. A variety of techniques are employed in this survey, namely naïve, moving average, decomposition, single exponential smoothing, ARIMA, multiple regression, genetic regression and neural networks with the latter two methods being the non-traditional techniques. Official statistical data from 1992 to 1998 was used in this study. The actual and predicted number of visitors are then compared. The survey shows that the neural network method performs the best.