Improving the Accuracy of Hotel Reservations Forecasting: Curves Similarity Approach

Abstract

This study developed and tested a new type of an extrapolative forecasting model that improves the forecasting models of daily occupancy levels in a hotel by focusing on the shape of past booking curves. By applying a dissimilarity measure to identify past book ing curves with a similar shape, the identified similar curves were then used to project future reservations. The study examined 10 forecasting horizons ranging from 1 to 99 days in advance. The results indicate that the curves similarity model is considerably more accurate than the three benchmark models tested These results are statistically and prac tically significant. In addition, this study shows that under certain conditions a model that combines two independent models might produce a forecast that is more accurate than that of its components.