Forecasting is of utmost importance for the Tourism Industry. The development of models to predict visitation demand to specific places is essential to formulate adequate tourism development plans and policies. Yet, only a handful of models deal with the hard problem of fine-grained (per attraction) tourism demand prediction. In this paper, we argue that three key requirements of this type of application should be fulfilled: (i) recency-forecasting models should consider the impact of recent events (e.g. weather change, epidemics and pandemics); (ii) seasonality-tourism behavior is inherently seasonal; and (iii) model specialization-individual attractions may have very specific idiosyncratic patterns of visitations that should be taken into account. These three key requirements should be considered explicitly and in conjunction to advance the state-of-the-art in tourism prediction models. In our experiments, considering a rich set of indoor and outdoor attractions with environmental and social data, the explicit incorporation of such requirements as features into the models improved the rate of highly accurate predictions by more than 320% when compared to the current state-of-the-art in the field. Moreover, they also help to solve very difficult prediction cases, previously poorly solved by the current models. We also investigate the performance of the models in the (simulated) scenarios in which it is impossible to fulfill all three requirements-for instance, when there is not enough historical data for an attraction to capture seasonality. All in all, the main contributions of this paper are the proposal and evaluation of a new information architecture for fine-grained tourism demand prediction models as well as a quantification of the impact of each of the three aforementioned factors on the accuracy of the learned models. Our results have both theoretical and practical implications towards solving important touristic business demands.
Copyright: © 2022 Khatibi et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.