Tourism demand forecasting using ensembles of regression trees

In this study, sophisticated ensemble models are empirically examined and investigated for their success and applicability in the case of forecasting tourism demand with the multi time series data. This study makes the contributions to the tourism demand forecasting by developing Ensemble learners, which is one of the most recent area in the data mining. Ensemble learners have been developed, implemented and examined based on the M5P and M5-Rule model trees, and Random Forest algorithms in the framework of the bagging, boosting, randomization, stacking and voting to make tourism demand forecasting for Turkey. Virtually it is not seen that the most of ensemble models proposed in this study have been employed in the context of the tourism demand forecasting. Comparison of forecasting accuracies in terms of relative absolute error (RAE) and root relative squared error (RRSE) measurements shows that the proposed ensemble models achieve a better performance in terms of accuracy than the single use of the M5P, M5-Rule model trees, and Random Forest algorithms. The use of ensemble models related to single models in data mining could be a better alternative. In tourism management, it is important reliably to estimate the tourism demand for the sector.