Prediction of South Africa’s Tourism Hotel Accommodation Monthly Income: Challenges in an Environment Characterised by a World Recession and a World Cup

Abstract

This paper uses official data to develop exponential smoothing models for predicting monthly total income in tourism hotel accommodation in South Africa for the period 2004 to 2011. The performance of the developed models is evaluated against a neural network model for in and out of sample predictions. Empirical results from the study show that the triple exponential smoothing model produces better forecast accuracy compared to the single and double exponential smoothing models. The neural network model outperforms all the three exponential smoothing models prior to the onset of recession in 1998. The triple exponential smoothing model produces better forecast accuracy after the onset of the world recession. This suggests that a time series undergoing structural changes may require different forecasting approaches. This study is meant to provide insight into the dynamics of the South African hotel accommodation income time series thereby stimulating future definitive work on methods of forecasting.

DOI: 10.5901/mjss.2014.v5n20p460