A big data analytics method for forecasting tourism flows.

This paper was motivated by the fact that there is an increasing share of government and commercial organizations’ profit from tourists in European countries. The model used to analyze tourist flows considers a city as a dynamic flow network where the nodes of it are represented by sightseeing and other points of attractions including hotels, shops, etc. Dynamic generalized linear model specification of Bayesian multivariate time-series analysis methodology was implemented, where the tourist flows are conditional Poisson distributed observable variables which are expected to be determined by latent parameter changing over time, so there is a hidden Markov chain. More sophisticated and evolved versions of the model allowing to investigate seasonal trends present in the tourism industry were estimated as well. A computational part of the case-study is done using the R programming language. A possibility to make real-time predictions within specific confidence interval allowing to approximate demand level and to increase the efficiency of services is an example of practical benefits of the work.