LOW COST INFERENTIAL FORECASTING AND TOURISM DEMAND IN ACCOMMODATION INDUSTRY
Main Article Content
Keywords
Time series, forecasting models, Naïve models, ex-post and ex-ante forecasts, forecast accuracy and validation, tourism demand
Abstract
This paper establishes a low cost inferential model that allows reliable time series forecasts. The model provides a naive unique computationally straightforward approach based on widely-used additive models. It refers to the decomposition of every time series value in “random” components, which are compounded to constitute a “Fibonacci type” predictor random variable. The expected value of this predictor gives a forecast of a future time series value. The standard deviation of the predictor serves to construct a prediction interval at a predefined confidence level. The major features of our model are: forecasting accuracy, simplicity of the implementation technique, generic usefulness, and extremely low cost effort. These features enable our model to be adopted by tourism practitioners on various types of forecasting demands. In this paper, we present an application study to forecast tourism demand that exists in the Greek accommodation industry (i.e. in Greece and in the broad region of Athens). In the application study, two independent approaches have been adopted. In the first approach we implemented our model, and in the second approach we implemented the well-known Box-Jenkins method.