The evaluation of (big) data integration methods in tourism

Abstract

In view of many dynamic changes taking place in the modern world due to the pandemic, the migration crisis, armed conflicts, etc., it is a huge challenge for official statistics to provide good-quality information, which should be available almost in real time. In this context, integration of data from multiple sources, in particular big data, is a prerequisite. The aim of the article is to characterise and evaluate the following selected methods of data integration in tourism statistics: Natural Language Processing (NLP), machine learning algorithm, i.e. K-Nearest Neighbours (K-NN) using TF-IDF and N-gram techniques, and Fuzzy Matching, belonging to probabilistic methods. In tourism surveys, data acquired using web scraping deserve special attention. For this reason, the analysed methods were used to combine data from booking portals (Booking.com, Hotels.com and Airbnb.com) with a tourism survey frame. An attempt was also made to answer the question of how the data obtained from web scraping of tourism portals improved the quality of the frame. The study showed that Fuzzy Matching based on the Levenshtein algorithm combined with Vincenty’s formula was the most effective among all tested methods. In addition, as a result of data integration, it was possible to significantly improve the quality of the tourism survey frame in 2023 (an increase in the number of new accommodation establishments in Poland by 1.1% and in Bulgaria by 1.4%).